2003-11-18 Matti Aarnio * configure.in: Mark the version as 2.99.56 (finally!) 2003-11-16 Matti Aarnio * proto/cf/aliases.cf: More patches from Eugene Crosser (actually something I omitted by accident when applying things manually; as in "retyping" ..) * smtpserver/policytest.c: Moved target specific rejects early -- so that we can reject some targets even for cases where e.g. message source IP address carries "relay all thru us". * transports/smtp/smtp.c: More of wrong return code hunting. It turns out, that reopening has been done without resetting MX index, and that omission caused in some case at some targets the reopen to fail with fatal error code. 2003-11-07 Matti Aarnio * transports/smtp/smtp.c, transports/smtp/smtp.h: Tracked connection restart problems, specifically against HOTMAIL, which runs things in highly questionable manner, but that may happen with other systems, too. (Apparently HM hangs up the inbound smtp connection right after receiving end of previous message.) * compat/sendmail/sendmail.c, smtpserver/smtpserver.h, smtpserver/smtpserver.c, smtpserver/smtpcmds.c: - When 'sendmail' is called with '-bs' option, let it call the subsequently started smtpserver as: "sendmail-smtp-in". - Don't call policy database functions, when the program isn't being run under a network socket. 2003-11-07 Eugene Crosser * proto/cf/aliases.cf: Define "sysuser" variable that is either "user" or "user@domain". Use this as argument to 'homedirectory' and 'login2uid' functions that ultimately call getpwnam() * proto/cf/server.cf: Strip off domain part of "user" variable only if getpwnam() is not domain-aware. As far as I can tell, this variable is unly used in 'login2uid' and 'fullname' functions. * transports/mailbox/fmtmbox.c: If getpwnam() is domain-aware, pwd->pw_name may contain username@domain; but we need to compute hash of local part only, because otherwise hash("user@mydomain.com") and hash("user") would differ, and "defaultdomain" feature would break. "address" is always the local part (domain has been cut off before). * transports/mailbox/mailbox.c: If DOMAIN_AWARE_GETPWNAM=1, do not rfc822localize() the user, just strip off quotes in case they are present (no syntax check here, just delete any double quote that is not backslash- escaped). Thus further zgetpwnam() will be called on user with domain. 2003-11-06 Matti Aarnio * smtpserver/mxverify.c, smtpserver/policytest.c: More of Mariano's question analysis. Found out that I were slightly too paranoid about T_TXT object length, and actually cut back one char from all objects. Now will log lookup results into smtpserver's commentary side-stream. * smtpserver/mxverify.c, smtpserver/policytest.c: Mariano Absatz wanted DNS RBL tests (results, at least) to be logged into smtpserve log. * smtpserver/readpolicy.c: Looked deeper into policy database compilation process. Behold; sysadmin can create attributes that are TOO LONG. Now the sysadmin can't. Pattern compilation will fail at such. 2003-11-05 Andrey Blochintsev * smtpserver/readpolicy.c: Allow '_' prefixed user names to be used. 2003-11-04 Matti Aarnio * router/daemonsub.c: Well, things weren't proper for tasks in the primary router directory. My testing was with alternates... * smtpserver/policytest.c: Roy Bixler reported that his smtpserver does hang up when processing (rather long) DNS BL datasets. It turned out that policy compiler uses unsigned char for various things, including attribute length (thus up to 255-2 is ok for an attribute data), but my interpreter referred to it as signed char, and things larger than 127 became negative in length all the sudden... * configure.in, SiteConfig.in: From Eugene Crosser: Configure system-wide SYSLOGFLG=, SNMPSHAREDFILE=, and DOMAIN_AWARE_GETPWNAM= ZENV variables. * configure.in: Test and determine, how one can find out the FD associated with opendir() target object. To be used in router process, when a fchdir() is needed. * libc/mail.c, libc/sfmail.c: A wee bit of adaptations to compile the beast in Windows (cygwin, that is..) * proto/cf/process.cf, router/daemonsub.c, router/rfc822.c: Support input files going into ROUTERDIRHASH= sublevel, and don't do unnecessary rename()s if the original file has proper name ( = i-node number in decimal ) already. The process function may get scalled as: process B/123456 which is now perfectly legal. NOTE: The daemon subsystem will do file locking on the original file for the duration of file staying in the processing! 2003-10-26 Matti Aarnio * router/libdb/ldap.c: Patches to support LDAPv3 (via option setting) thru a "protocol 3" line in the configuration file. 2003-10-07 Matti Aarnio * transports/smtp/smtptls.c: In some rare occasions the discipline processor loops forever in errno processing, when it has received an EINTR during a syscall. 2003-10-03 Matti Aarnio * smtpserver/smtpserver.c: Compressed logged commentary lines a bit. Remote host reversed name, and its actual IP address are listed in same line. 2003-10-01 Matti Aarnio * smtpserver/smtpcmds.c: When outputing commentary data to spool file, quote all chars outside normal printable ASCII range, and some inside it... 2003-09-27 Matti Aarnio * configure.in: Changed the --without-ldap to --with-ldap and worked a bit in other internal related things. 2003-09-26 Matti Aarnio * include/mail.h.in, libc/mail.c, libc/sfmail.c: New variants of *mail_close*() functions, now can control the asynchronity of the close() thing. Essentially controlling use of fsync(2) on outgoing spool file descriptor. * man/scheduler.8.in, proto/scheduler.conf.in, scheduler/msgerror.c, scheduler/prototypes.h, scheduler/readconfig.c, scheduler/scheduler.c: Added "PARAMmsgwriteasync = 1" option that can control the asynchronity in mail.c/sfmail.c library routines. * transports/smtp/smtp.c, man/smtp.8.in: From Andrey Blochintsev an option to allow connection to IP addresses of the same host. 2003-09-25 Matti Aarnio * configure.in, router/db.c, router/libdb/gdbm.c, router/libdb/ldap.c: - A bunch of "--without-*" things: --without-ndbm --without-gdbm --without-ldap To disable some autodetections. - Removed traces of SDBM autoconfig; a once-upon-a-time replacement for NDBM. * smtpserver/smtpcmds.c: When reporting various rejections, do report also peer IP address. This to help some analysis, and to some extent, to help users. * configure.in: - Default builtin value for SCHEDULEROPTIONS is now changed to "-S -H". - Fixed a missing comma from --with-perl-installdirs option parameter processing. - Added some AC_MSG_RESULT()s about NDBM library pickup processing. * proto/cf/fqdnaliases.cf: One instance of $(listexpand .. ) call needed "-N -" parameter as not to do DSN expansions. * proto/forms/delivery - Added "ADR Bcc: " to drain errors-on-error reports. - Made small URL change * utils/perl/Makefile.in: Removed (from 'install' target) explicite PREFIX= value setup. Possible PREFIX= value can be set thru configure time --with-perl-installdirs="..." parameter. 2003-09-09 Matti Aarnio * configure.in, utils/perl/Makefile.in: have --with-perl-includedirs="INCLUDEDIRS=vendor" option at configure time. It seems to be usefull (or at least working) thing at RedHats, but the same doesn't hold true with all other perl environments that I use... 2003-09-08 Matti Aarnio * man/Makefile.in: Replaced (in 'make install' target) literal 'zm' suffix in files with '$Z' substitution. * man/aliases.5.in, man/newaliases.1.in, man/smtpserver.8.in: Man-page text updates. 2003-09-05 Matti Aarnio * transports/smtp/smtp.c: Patch from Jerzy Sobczyk about outgoing SMTP not always honouring '-h' option set local hostname in outbound EHLO/HELO greetings. 2003-09-03 Matti Aarnio * smtpserver/contentpolicy.c: Patch from Eugene Crosser to run be able to run content-filter as an AF_UNIX socket server. 2003-09-02 Matti Aarnio * transports/smtp/appendlet.c: Oops, readalready processing bug. * transports/sm/sm.c, transports/mailbox/mailbox.c: Code rearrangements during verification job that similar bugs don't exist as in appendlet.. 2003-09-01 Matti Aarnio * scheduler/prototypes.h: Change 'const char *log;' to 'const char *logfn;' * utils/perl/Makefile.in: Deeper study of what to do in perl module compilation. Now "make install DESTDIR=.." parameter should have real effect in the underlying Makefile.PL files. * transports/libta/ctlopen.c: One test of ta_use_mmap was slightly wrong, and end result was faulty let_buffer() initialization. 2003-08-29 Matti Aarnio * debian/* Debian packet builder scripts from Xose Vazquez Perez * include/libz.h, lib/zgetbindaddr.c, scheduler/transport.c, smtpserver/smtpserver.c, smtpserver/smtpserver.h, smtpserver/cfgread.c, transports/smtp/smtp.c: Added third parameter into zgetbindaddr() function, namely: what address family should be used ? Entry parameter is: AF_INET / AF_INET6 (0 == AF_INET) That is needed for "iface:..." binding mode. * utils/smtpserver-log-parser.pl.in: Found a bug in the 'pid-parser', changed the RE patterns, and system that formerly leaked a lot of memory does now only a little bit.. * man/scheduler.8.in: Fixed a typo in PARAMauthfile= document. * proto/smtpserver.conf.in: Documentary comment changes in the end of the file in the 'style-flags' section. 2003-08-28 Matti Aarnio * man/Makefile.in, SiteConfig.in, man/zmailer-conf-generate.pl, man/mailq-m-generate-pl: New generator (zmailer.conf.5zm), and typo fix in the old. Also source changes in SiteConfig.in which is used as input for the zmailer.conf.5zm file. * configure.in, transports/libta/ctlopen.c, SiteConfig.in: Change the --with-ta-mmap to control what SiteConfig file (and thus zmailer.conf file) gets about TA_USE_MMAP= variable value. * man/*: Massive rework of man-pages; all cross-references will be using suffix names like: (8zm). Man-sections got "zm" tagged to them to ease the placement into various systems, like with 'inn' that has 'sm.8' ... * smtpserver/contentpolicy.c, smtpserver/smtpdata.c: - Typo in debug texts of contentpolicy filter harness. - Small default-text code rearrangements in smtpdata.c * man/man-to-html.sh: Apparently we can be without 'groff -c' option, and still not get colour output to produced HTML.. Oh yes, that was when using '-Thtml' output driver. Present HTML generation uses '-Tascii' and magic state processor to convert produced (with occasional escapes) text file into fixed width courier font HTML. 2003-08-26 Matti Aarnio * proto/smtpserver.conf, smtpserver/smtpcmds.c: In prototype smtpserver.conf: comment away 'localhost' case with 'f' and 't' flags. People do get surprises when using those, and skipping some notes in INSTALL. ALSO: Made 'enable_router' to semi-silently to control even looking into router controlling flags. If router is not enabled, it is not complained about. * SiteConfig.in, include/ta.h, transports/libta/ctlopen.c, transports/libta/lockaddr.c, transports/mailbox/mailbox.c, transports/sm/sm.c, transports/smtp/smtp.h, transports/smtp/smtp.c, transports/smtp/appendlet.c: Changed compile time configured TA_USE_MMAP feature to runtime configured ZENV variable TA_USE_MMAP, whose existence uses the thing, and nonexistence does read message into limited size internal buffer... ACTUALLY TESTED CODES ARE: 'mailbox' AND 'smtp'. THE 'sm' IS NOT TESTED AT THIS POINT! 2003-08-25 Matti Aarnio * utils/makedb/makedb.c, utils/policy-builder.sh.in: Code crashed in case the input wasn't like it was wanted.. Oops. * smtpserver/policytest.c, doc/guides/smtp-policy, man/smtpserver.8.in, proto/db/smtp-policy.src: Document editing, and at policytest.c: localizing pbuf[] variable to its actual use. * transports/smtp/smtp.c: Conditionalize one ESMTP_STARTTLS referral, that was added recently. 2003-08-22 Matti Aarnio * proto/db/smtp-policy.src: Couple comments on input syntax. * utils/policy-builder.sh.in, utils/makedb/makedb.c: "makedb -p" crashed in some situations, fixed NULL referral. Also report about exit code of the makedb in policy-builder.sh script. * configure.in: Version 2.99.56.pre9 * smtpserver/smtpserver.h, smtpserver/cfgread.c, smtpserver/contentpolicy.c, smtpserver/smtpdata.c, man/smtpserver.8.in: Support multiline replies from the content-policy program. Separator in between lines is '\r' (CR) character! * Makefile.in: Added target "make autoconf" -- to help myself... 2003-08-22 gettextize / Matti Aarnio * configure.in (AC_CONFIG_FILES), aclocal.m4, config.sub, config.guess. config.rpath, include/gettext.h, po/*, m4/* ABOUT-NLS, intl/*: Ran "gettextize --c --intl", and did what it told to do. Introduced also m4/ directory. 2003-08-21 Matti Aarnio * smtpserver/contentpolicy.c: One extra newline in comment logging into a file. * transports/mailbox/mailbox.c: Didn't always yield diagnostics for the outgoing messages. My bad.. (EDQUOT processing in particular.) * transports/libta/diagnostic.c, include/ta.h: Make sure we always have NOTARY 'action' value set for something! 2003-08-20 Matti Aarnio * libc/Makefile.in, libresolv/Makefile.in, man/Makefile.in, utils/Makefile.in, utils/makedb/Makefile.in, utils/mxverify/Makefile.in, utils/perl/Makefile.in, utils/vacation/Makefile.in: DESTDIR= defined where it overrides inherited thing from environment. Don't define it! 2003-08-19 Matti Aarnio * smtpserver/smtpserver.c: Moving variable 'int s' into places where it is actually used. 2003-08-15 Matti Aarnio * sfio2002/* Removing the entire tree, won't be using that one, way too much demanding stuff, like threads... * man/smtpserver.8.in, smtpserver/cfgread.c, smtpserver/smtpauth.c, smtpserver/smtpserver.h: A bit more of SASL related things. Added "PARAM sasl-mechanisms ...", copied a bit more of sendmail code into here, and used it. Still my tests failed to login... * scheduler/scheduler.c: Jerzy Sobczyk reported bug in 'Z' option processing * smtpserver/smtpserver.c, smtpserver/smtpauth.c, smtpserver/smtpserver.h, man/smtpserver.8.in, proto/smtpserver.conf.in, smtpserver/Makefile.in, smtpserver/cfgread.c: Complete the Cyrus-SASL-2 code integration, quite experimental yet! In early tests the SASL library appears to use /dev/random, which is very bad news... * smtpserver/smtpserver.h, smtpserver/smtpserver.c, smtpserver/cfgread.c, man/smtpserver.8.in: Support multiple instances of "PARAM BindAddr" Also massive code rearrangement around the primary accept() call. * smtpserver/smtpdata.c: Use smtp_tarpit() calls instead of local own response delayer.. * include/policy.h, smtpserver/readpolicy.c, smtpserver/policytest.c, smtpserver/policytest.h, smtpserver/smtpserver.c, smtpserver/smtpserver.h: Eugene Crosser's IP source configurable maxparallelipconnects * proto/zmailer.sh.in: Couple typos in -Z/--zconfig parameter passing * router/db.c, router/libdb/ldap.c, configure.in: Synced to be in par with OpenLDAP 2.1.25, should also work with older ones. * proto/cf/aliases.cf, proto/cf/fqdnaliases.cf: Support (convert!) (.+)--(.*) in place of ([^+]+)\+(.*) * scheduler/scheduler.c: Move the first getzenv() call to be after readzenv() call, if any. * smtpserver/rfc821scn.c: Comment away an unused variable. 2003-07-23 Matti Aarnio * scheduler/scheduler.c: Found a likely reason for scheduler's file descriptor leakage.. The sfio-package does something wrong in odd case used in log-file re-opening, and filehandle recycling. * scheduler/scheduler.c: Don't unlink the message file in case its opening fails. (Like when scheduler/system has ran out of fd's) * scheduler/mailq.c: Count also files in the $POSTOFFICE/queue/, and show it, if it _differs_ from $POSTOFFICE/transport/ count. Changed also slightly the output texts, which _MAY_ affect some management scripts! 2003-07-01 Matti Aarnio * doc/guides/smtp-policy, proto/smtpserver.conf.in, proto/db/smtp-policy.src: Typo fixes * transports/libta/ctlopen.c, transports/libta/mimeheaders.c, transports/smtp/smtp.c: Semi-meaningless changes to assist trace debugging of memory leaks. No idea if that finds my last 1-2 % of memory leaks that I still see... 2003-06-27 Matti Aarnio * README.LICENSES, LICENSES/README, LICENSES/ARTISTIC, LICENSES/BSD, LICENSES/GPL-2, LICENSES/LGPL-2: Clarifying the license issues of ZMailer. 2003-06-25 Matti Aarnio * configure.in: Changed version number to more RPM-friendly one, e.g. no hyphen in it. * transports/smtp/smtp.c: The getmxrr() calls in cname_lookup() did leak AI and NAME objects. Freeing collected (and unneeded) data there. 2003-06-24 Matti Aarnio * configure.in, scheduler/mailq.inc, scheduler/mailq.c, scheduler/mq2.c: Auto-config test to find getloadavg() function (BSD, Solaris 8, Linux, ... a de-facto API for this data, in libc !) and to print 1 and 5 minute load-averages into the scheduler's "mailq -M" report. * transports/libta/diagnostic.c, transports/smtp/getmxrr.c: Moving variable declarations around a bit, and not doing initializations for NULLs/zeroes, thus letting things to be placed into BSS instead of initialized read/write data segment. 2003-06-22 Matti Aarnio * transports/libta/mimeheaders.c: Last detected memory leak in header parsers. 2003-06-21 Matti Aarnio * include/ta.h, transports/libta/ctlopen.c, transports/libta/diagnostic.c, transports/libta/mimeheaders.c: More memory leak tracking related work. Most fumbling in the dark, I am afraid.. 2003-06-20 Matti Aarnio * include/ta.h, transports/libta/ctlopen.c, transports/libta/Makefile.in, transports/sm/sm.c, transports/smtp/smtp.c, transports/mailbox/mailbox.c, transports/hold/hold.c, transports/errormail/errormail.c, transports/expirer/expirer.c, transports/reroute/reroute.c: Removed rp->routermxes field, data format parsing, and support calls for it from system. It never materialized fully, and is weaker method, than original (not materialized either) X-or recipient addresses.. * transports/mailbox/mailbox.c, transports/sm/sm.c, transports/smtp/appendlet.c: Still tracking memory leaks. Apparently the SFIO library doesn't free up memory in all cases. Namely sfnew() call has a bit of problems, especially when sfdisc() sets processing discipline, and likely reallocs buffer.. Now altering code to use heap based static buffers. 2003-06-19 Matti Aarnio * router/rfc822.c, router/functions.c: - Removing dead code (lots of it) - Reworked a bit the reject() function address processing, now also the originally intended error recipient will receive the message. (Thanks to Peter Ip to raise my attention at it.) - All router internal mail_open() calls are wrapped with runastrusteduser()/runasrootuser() calls, so that created message file's owner will be 'trusted'. Is THAT an ok thing in all situations ? No definite idea... * proto/cf/standard.cf: Rearranged the list of known top-level domains, and added 'info', 'biz', 'name', 'eu'. * transports/mailbox/mailbox.c, transports/libta/mimeheaders.c, transports/libta/ctlopen.c: Been hunting memory leaks... Have found a bunch more.. * transports/smtp/smtptls.c: When there was no logging file of any kind, this used to send TLS related diagnostics to the scheduler's log, which didn't work very nicely in one of my systems.. Now changed so that if there is verbose-log, printouts go there, and independent of that, if there is smtp transport-agent log, printouts go also there. If there are neither, nothing is printed. 2003-06-17 Matti Aarnio * smtpserver/policytest.c: If MAIL FROM test had 'rejectsource -' and/or 'freezesource -' attributes in 'user@do.main' lookup, disable lookups of those same things from fallbacks of 'do.main' lookup. * transports/mailbox/mailbox.c, transports/libta/mimeheaders.c, include/ta.h, transports/smtp/smtp.c: Track a bit more of allocated memory chunks, and provide explicite free() routines for them. (Hunting originally memory leaks in mailbox.) * transports/mailbox/mailbox.c, man/mailbox.8.in: Added option "-F edquot" to instantly fatalify quota-exceed condition of the baseline code. * libc/getaddrinfo-test.c: Another test-harness.. 2003-06-09 Matti Aarnio * configure.in: Mark version 2.99.56-pre7 * include/libz.h, include/shmmib.h, lib/fdstatfs-test.c, lib/fdstatfs.c, lib/zshmmibattach.c, router/daemonsub.c, scheduler/mailq.inc, scheduler/scheduler.c, smtpserver/smtpcmds.c, smtpserver/smtpdata.c, smtpserver/smtpserver.h: - When wanting to add used spool space gauges into SHM segment header, realized that it does not have 32 spare uint variables.. Oops.. Had to change block layout -> new MAGIC, full rearrancement of all so far accumulated small counters, etc.. - renamed fd_statfs() into free_fd_statfs(), and added a cousin for it: used_fd_statfs(). Both return space in kilobytes (e.g. can describe up to 2 TB...) - Removed RT.StoredRecipients gauge, which the router can't maintain anyway. * lib/fdstatfs.c, lib/zshmmibattach.c, scheduler/scheduler.c, smtpserver/smtpserver.h, smtpserver/smtpcmds.c, smtpserver/smtpdata.c, smtpserver/smtpserver.c: Changed the fd_statfs() to report KILOBYTES of free space, while still returning it via mere 'long' variable in all systems (e.g. may have limit of LONG_MAX a.k.a. around 2 G, which in kilobytes means: around 2 Terabytes ..) Also re-juggled codes using this free-space data. * man/Makefile.in, man/mailq-m-generate.pl, scheduler/mailq.inc, man/man-to-html.sh, man/mailq.1.in: Generating "mailq-m(5)" man-page from within mailq.inc stored XMLish comments. * include/shmmib.h, router/daemonsub.c, router/rfc822.c, smtpserver/smtpserver.c, smtpsserver/smtpdata.c: - Count router process faults explicitely - Let smtpserver to count attempted incoming recipients (similarly with ss.ReceivedMessages counter) - Let smtpserver to count instances, when arriving connection has been summarily closed due to too high a load. (Limit criterias are PARAM parametrizable.) - Track router actual job processing instances failing somehow, and count those separately. 2003-06-05 Matti Aarnio * scheduler/mailq.inc: Print added seen EHLO-capability counters, and sent SMTP options counters. 2003-06-04 Matti Aarnio * include/shmmib.h, smtpserver/smtpcmds.c, transports/smtp/smtp.c: Count several ESMTP options seen, (and used) by the smtp-ta, and used by the smtp clients. * include/shmmib.h, scheduler/mailq.inc, smtpserver/smtptls.c, transports/mailbox/mailbox.c, transports/smtp/smtp.c, transports/hold/hold.c, transports/errormail/errormail.c, transports/sm/sm.c, transports/expirer/expirer.c, transports/reroute/reroute.c: Changed names of a number of variables, and added basic counters into all transport agents, which didn't have anything. * Makefile.in, compat/Makefile.in, doc/Makefile.in, lib/Makefile.in, libc/Makefile.in, libident/Makefile.in, libmalloc/Makefile.in, libresolv/Makefile.in, libs/Makefile.in, libsh/Makefile.in, man/Makefile.in, proto/Makefile.in, router/Makefile.in, scheduler/Makefile.in, sfio/Makefile.in, smtpserver/Makefile.in, ssl/Makefile.in, utils/Makefile.in, transports/Makefile.in: "make mostlyclean" -- clean everything except SFIO library. (Which is stand-alone in itself..) This is side-result from SHMMIB-block tunings. * configure.in, proto/zmailer.sh.in, transports/sm/sm.c, transports/smtp/smtp.h, transports/smtp/smtp.c, transports/mailbox/mailbox.c: While going over some unrelated things, spotted mentions about mailbox "From " header syntax details. Its presumed syntax is very particular, and needs care with correct values. Particular danger is of getting some odd system-wide locale setting to wreck proper date presentation rules. Best solution would most likely be to open-code the presentation of time, instead of using ctime(3), but ... 2003-06-02 Matti Aarnio * transports/smtp/smtp.c: Wondering, if we are sometimes missing closing of smtp sessions in SMTP transport agents. Just in case we do, added an atexit() mechanism to handle it. Latter experiments of shutting down the scheduler, and observing that eventually the outgoing socket gauges do go to zero.. all is thus well. 2003-05-30 Matti Aarnio * scheduler/mailq.c: Just before closing the MAILQ-v2 command socket, do send there "QUIT\n" command too! Sure the EOF handles the cleanup too, but ... * scheduler/mq2.c, scheduler/mailq.pl, scheduler/update.c: Andrey Blochintsev sent patch to implement "KILL MSG" type functionality via the MAILQ-v2 socket command. Client to implement it is in (surprise), the perl-testbench of MQ2... Yours Truly did small style edits (like C99/C++ style "//" comment change to "/*" ) 2003-05-29 Matti Aarnio * include/shmmib.h, scheduler/scheduler.c, scheduler/mailq.inc: Count for scheduler's time-server starts, and last start-time. * transports/smtp/smtp.c: Count pipelining transmits only when doing connection open, not for every sent message... 2003-05-27 Matti Aarnio * smtpserver/smtpdata.c: Examples in RFC 2034 weren't the best possible, practice has shown, that "250 2.6.0 .." isn't quite what is intended. Sending now (like sendmail) "250 2.0.0 ..." * include/shmmib.h, scheduler/Makefile.in, scheduler/mailq.inc, scheduler/scheduler.h, smtpserver/smtpcmds.c, smtpserver/smtphelp.c, smtpserver/smtpserver.h, smtpserver/smtpserver.c, transports/smtp/smtp.c: More simpleish counters and gauges... * transports/smtp/smtp.c, include/shmmib.h, scheduler/mailq.inc: Count outgoing sessions where we use PIPELINING. * scheduler/scheduler.c: Fix the LogFreeSpace-kB-G to be filled (instead of overwriting SpoolFreeSpace-kB-G !) * include/shmmib.h, transports/smtp/smtp.c, scheduler/mailq.inc: Added smtp TA process creation counter, and a gauge to observe how many there are running in parallel. 2003-05-26 Matti Aarnio * bin/mkdep.in, bin/mklibdep.in, compat/rmail/Makefile.in, compat/sendmail/Makefile.in, lib/Makefile.in, libc/Makefile.in, libident/Makefile.in, libresolv/Makefile.in, libsh/Makefile.in, router/Makefile.in, router/libdb/Makefile.in, scheduler/Makefile.in, sfio/Makefile.in, smtpserver/Makefile.in, transports/errormail/Makefile.in, transports/expirer/Makefile.in, transports/fuzzyalias/Makefile.in, transports/hold/Makefile.in, transports/libta/Makefile.in, transports/mailbox/Makefile.in, transports/reroute/Makefile.in, transports/sm/Makefile.in, transports/smtp/Makefile.in, utils/makedb/Makefile.in, utils/vacation/Makefile.in: Fixed the "make depend" processing, now it works also when compiling in "adjacent directory" (tested only in that, though..) * include/libz.h, include/shmmib.h, lib/fdstatfs.c, router/daemonsub.c, lib/zshmmibattach.c, router/rfc822.c, scheduler/mailq.c, scheduler/mq2.c, scheduler/scheduler.c, scheduler/threads.c, scheduler/transport.c, scheduler/update.c, smtpserver/smtpchild.c, smtpserver/smtpcmds.c, smtpserver/smtpdata.c, smtpserver/smtpetrn.c, smtpserver/smtpserver.c, smtpserver/smtpserver.h, smtpserver/smtptls.c, transports/smtp/smtp.c, transports/smtp/smtp.h: Reworked a bit more of the monitor counters and gauges. (Actually a lot.. Added first glimmers even to SMTP-ta.) 2003-05-25 Matti Aarnio * smtpserver/smtpserver.c: Account incoming SMTP protocol command lines, and some part of bad inputs. * lib/zshmmibattach.c: Fixes around the file locking mechanisms. flock/fcntl/lockf supported. * scheduler/scheduler.c: Fixed the StoredReceipients and ReceivedRecipients accounting counters. Primarily that gauge, secondarily moved the received counter next to the gauge. (The gauge was incrementing twice for each received recipient, and decremented once for each sent ...) * router/rfc822.c: Track the number of recipient addresses before routing those addresses (and possibly expanding them). 2003-05-24 Matti Aarnio * include/shmmib.h, lib/zshmmibattach.c, router/daemonsub.c, router/router.c, scheduler/mailq.c, scheduler/mq2.c, scheduler/scheduler.c, scheduler/scheduler.h, scheduler/threads.c, scheduler/transport.c, scheduler/update.c, smtpserver/smtpchild.c, smtpserver/smtpserver.c, smtpserver/smtpserver.h: - Added fields of primary daemons start times, and counters of how many times they have been started in this block.. - Fixed the scheduler process accounting, and for a good measure, threw in thread, vertex, and recipient gauges. * utils/perl/Makefile.in: As the "mailq" needs to be copied from its original location, when processing building in "adjacent directory", the "make clean" will need to remove it under same conditions. * router/router.c, router/daemonsub.c, scheduler/scheduler.c, smtpserver/smtpserver.c: Moved Z_SHM_MIB_Attach() into a place where these programs have _not_yet_ detached themselves from the console, and will loudly complain if the SHM attachement fails (presuming that SNMPSHAREDFILE ZENV variable is set at all.) * scheduler/mailq.c, scheduler/mq2.c, utils/perl/mailq/mailq.pm, man/mailq.1.in: A new primitive: "SHOW COUNTERS", which can be invoked with "mailq -QQQQ" and is same text as local memory attached "mailq -M" -- but can be at remote system. * man/Makefile.in, man/man-to-html.sh, man/zmailer.3.in: Using GNU Makefile "pattern rules" to do things that had been open-coded into the makefile. Also support referring to e.g. ZMailer::mailq(3) in hyperlinks. * configure.in, lib/zshmmibattach.c, include/libz.h, scheduler/mailq.c: Added test for header, Added flock()/lockf() locking of the shared segment storage, and detailed error status returning from Z_SHM_MIB_Attach() so that "mailq -M" can report what exactly happened, when things didn't work. * include/libz.h, lib/zshmmibattach.c, smtpserver/smtpdata.c, scheduler/scheduler.c, smtpserver/smtpcmds.c, scheduler/mailq.c: More text changes, implemented two instances of spool free space gauge. (That thing peaks at about 2 GB, only when space goes BELOW it, there is a reason to watch for it..) * smtpserver/fdstatfs.c, smtpserver/fdstatfs-test.c, smtpserver/Makefile.in, lib/fdstatfs.c, lib/fdstatfs-test.c, lib/Makefile.in: Moved the fdstatfs*.c files from one directory to another. Did that also in CVS repository, and thus preserved revision data.. * scheduler/mailq.c: More 'mailq -M' printouts. * include/shmmib.h, router/daemonsub.c, router/rfc822.c, scheduler/scheduler.c, scheduler/transport.c, smtpserver/smtpcmds.c, smtpserver/smtpdata.c, smtpserver/smtpetrn.c, smtpserver/smtptls.c: - New counter variables, changed the magic version (mapping will be rejected if there is old persistent file in the system!) - More code to do SHM counting/tracking * utils/webtools/mailq-m.php, utils/webtools/mailq-q.php, utils/webtools/mailq-v.php: An experimental set of web-tools to do ZMailer status display in web.. Developed for friends to look into a busy system without using command-line tools.. Also as concepts for doing some advanced tinkering. 2003-05-23 Matti Aarnio * lib/zshmmibattach.c, include/libz.h, include/shmmib.h, scheduler/scheduler.c, router/daemonsub.c, scheduler/update.c, scheduler/threads.c, lib/Makefile.in, SiteConfig.in, scheduler/mailq.c, man/mailq.1.in, router/router.c, smtpserver/smtpserver.h, smtpserver/smtpchild.c, smtpserver/smtpserver.c: - Added new library function -- attach global MTA-instance specific shared memory mapped file to some particular running process. - Added new counters and gauges, far in excess of official RFC listed gauges.. SNMP shall coalesc subsystem counters and gauges. - Using this global shared segment for the time-server, too. * utils/perl/Makefile.in: Doing "make clean" shall not crash even if previous "make" hasn't been run in a subdirectory, where things are done with 'Makefile.PL' way... * transports/smtp/smtp.c: Additional "once" counter that breaks the loop, if other ways don't stop it -- when doing "more recipients" retry things. 2003-05-22 Matti Aarnio * configure.in, transports/smtp/smtp.h, transports/smtp/smtp.c: Had a deeper look into "NAGLE" code, and realized that we didn't #include , which carries the necessary socket-option tokens, thus no "NAGLE" control was ever done. * sfio/bin/iffe, sfio/src/lib/sfio/Stdio_b/features/stdio, sfio/src/lib/sfio/features/common, sfio/src/lib/sfio/features/sfio: Picked the 'iffe' script from SFIO-2002, as well as 'features/sfio', and 'features/common', the 'features/stdio' did need a bit of manual fixing, as there is no equivalent in newer version. (SFIO-2002 usage is -- seriously in doubt due to its overflowing of thread stuff.) 2003-05-21 Matti Aarnio * smtpserver/smtprouter.c: When encountering "*INTERNAL*ERROR*" do report with '400' status, instead of '500' status... * proto/db/smtp-policy.src: In boilerplate texts, say "Your address is not liked .." instead of "Your domain is not liked ..". * transports/smtp/smtp.c: The SMTP "NOOP" shall timeout in 30 seconds.. Otherwise it lingers around with DOT-timeout (20 minutes).. Also report this process state in proctitle data. * transports/smtp/smtp.c: In EZMLM mode the SMTP TA didn't unlock all recipients, which did cause considerable amount of "#resync" activity in between the scheduler, and SMTP TA. Replaced a ton of 'return's with 'goto more_recipients;' in order to coherently handle the EZMLM mode in all of its forms. Also unbroke the rp-chain (which is broken in EZMLM detection) * sfio/src/lib/sfio/sfsetbuf.c, sfio/src/lib/sfio/sftable.c, sfio/src/lib/sfio/sfvprintf.c, sfio/src/lib/sfio/Sfio_dc/sfdcdio.c: Small fixes to silence the compilation warnings. * scheduler/scheduler.c, scheduler/update.c: Moved global_wrkcnt, and MIBMtaEntry->mtaStoredMessages counter managements around so that they are essentially tracking on how many memory instances there are of the 'struct ctlfile' objects. Strictly speaking this isn't the same as count of jobs, but a lot simpler to do.. * scheduler/update.c: When reporting 'multiple processing of address ...', do mention also channel & host data for the _process_ which is reporting. 2003-05-20 Matti Aarnio * configure.in: Mark version: 2.99.56-pre6 * router/rfc822.c: Eugene Crosser reported that envelope header "with" without a value does cause router process to drop core. Indeed it does that. Oops. (denial of service attack is possible from inside the system.) * man/smtpserver.8.in, man/zmailer.3.in: Reworked/updated environment variable pickups of these programs. * smtpserver/smtptls.c: Instead of picking the POSTOFFICE location all by itself, use smtpserver process global value. * smtpserver/smtpserver.c: Andrey Blochintsev sent a patch to move WHOSON lookup upwards a bit so that the data is available, when it is being logged. * INSTALL: Updated to match the reality a bit better, than what it has been for a long time... 2003-05-15 Matti Aarnio * smtpserver/contentpolicy.c, smtpserver/smtprouter.c, smtpserver/smtpserver.c, smtpserver/smtpserver.h: From Andrey Blochintsev spotted process deaths by NULL pointer referral during state cleanup. * transports/mailbox/mailbox.c: Turning EDQUOT failure to instant permanent failure. 2003-05-13 Matti Aarnio * smtpserver/policytest.c, smtpserver/smtpserver.c: Removed printf()s from policy-test debug outputs, and use type() instead. * smtpserver/policytest.c, proto/db/smtp-policy.src, doc/guides/smtp-policy: From Andrey Blochintsev a bit more code, and commentary updates in boilerplate. * transports/smtp/smtptls.c: Oops, when "HAVE_OPENSSL" is inactive, still supply msg_info() function to be used, as well as FILE *vlog. * libc/strsignal.c: Matching tests with earlier autoconf rewrites * configure.in, man/Makefile.in, man/*: Changed ALL man-page files to be autoconf generated so that @MAILBIN@ substitutions (et.al.) can be done into the files. (Also renamed files in the CVS repository to reflect changed names in order to preserve the change history.) * proto/post-install.sh.in: Added '--destdir' option. * Makefile.in: DESTDIR value dependent reminder notes referring to "postinstall" tool. 2003-05-10 Matti Aarnio * transports/smtp/smtp.c: Increase the timeout_data to 5 minutes by default to match timeout_cmd -- in PIPELINING mode we had a bit too low timeouts.. * proto/scheduler.conf.in: By default, comment away the "ageorder" flag. When a troublesome message jams the queue, allow the queue to be randomized so that the troublesome message won't stay in its head. * smtpserver/policytest.c: From Andrey Blochintsev : Is it possible to setup smtp-policy to block almost all mail (exept to: abuse@ or to: postmaster@) from some networks? It seems for me settings like: _rbl0 rcpt-dns-rbl +:dul.maps.vix.com.... _rbl1 test-rcpt-dns-rbl + doing that for RBL-listed network[s], but it is not possible to have analogous setup for locally listed (in smtp-policy.*) networks. The patch below solves this problem. To activete you need add rcpt-dns-rbl attributes starting with '_' ('_' symbol is not allowed in dns names). With this patch settings [10.10.10.25]/32 rcpt-dns-rbl "_+" [10.10.10.0]/24 rcpt-dns-rbl "_TEST reject msg" postmaster@my.domain = _full_rights means that: Mail from 10.10.10.0/24 to postmaster@my.domain -- accepted from 10.10.10.25 accepted/rejected by other rules (relaytarget +/- ...) from 10.10.10.0/24 to other will be rejected with "550 5.7.1 Policy analysis reported: TEST reject msg rcpt=" diagnostic 2003-05-09 Matti Aarnio * libsh/io.h, libsh/io.h, libsh/builtins.c: While trying to implement a sh_read(), with a lot less incomprehensible code, had to introduce fgetc() redirecting. * proto/cf/fqdnalias.cf, proto/cf/aliases.cf, router/functions.c: Correcting a bit of the 'listexpand' treatment of DSN data in expanded addresses. When expanding thru LISTS, new ORCPT (and new Return-Path) will result.on each recipient address, but when expanding thru ALIASES (and .forward), ORCPT and Return-Path will stay unmodified. Disabled former (now unused) "listaddresses" function, which does a small part of the job... * scheduler/scheduler.c: Andrey Blochintsev reported typo in setproctitle() call. (***BSD specific thing.) * proto/zmailer.sh.in: Eino Tuominen spotted that I still have sloppy style to write scripts without testing them before publishing.. ( e.g. users test :-/ ) Missing double-quote in one sh embedded perl script. * proto/cf/fqdnalias.cf: Modification typo lurked in the FQDN aliasing script. OUTCH! Found it when testing local virtualized email service failing to failure, and producing 'deferred' messages... * lib/sleepycatdb.c, smtpserver/policytest.c, utils/makedb/dblook.c, utils/makedb/makedb.c: Support compiling against SleepyCat DB 4.1 (the C-API has again been changed; db->open() got additional parameters...) 2003-05-08 Matti Aarnio * libsh/builtins.c: Debugging sh_read()'s internal buffer re-allocation/extension code. Something is wrong in there, and it grows without bounds.. * include/listutils.h, libsh/listmalloc.c, router/db.c: "function with parameter that is a pointer to a function with ..." .. found correct thing. I wish gcc would tell as much as the Sun WorkShop C compiler tells... * configure.in, include/zresolv.h: Solaris 2.6 oops.. u_int16_t isn't known type there.. * transports/sm/sm.c, transports/mailbox/mailbox.c, router/daemonsub.c, scheduler/transport.c, smtpserver/smtpserver.c, smtpserver/smtpserver.h: Worked over the wait4()/wait3()/waitpid()/wait() usage. Essentially addded wait4() to the mix. * configure.in, Makefile.in: Syncing things to be in par with autoconf 2.57, including having it as a pre-req.. Also producing BUGREPORT and VERSION literals from configure.in file ! * configure.in, smtpserver/smtpserver.c, scheduler/scheduler.c, transports/smtp/smtp.c: Andrey Blochintsev reports that FreeBSD 4.x needs to use setproctitle() function. Added autoconfig tests for that. * lib/rfc822date.c: Just to be on the safe side, enlarging some buffers, and adding explicite sprintf() string lenth limits. * proto/zmailer.sh.in: Refer to inlined perl-script executor with @PERL@ autoconfig token. * proto/post-install.sh.in: Typo fix in autoconf patterns. Should actually rename this either to 'post-install', or 'post-install.pl', as it is a perl script... * lib/sleepycatdb.c: Andrey Blochintsev reported that things didn't work quite smoothly with only BSD DB 1.x in system. (No SleepyCat's latter versions.) * scheduler/conf.c: Lower the global_report_interval from 15 to 5 minutes. 2003-05-06 Matti Aarnio * lib/zgetbindaddr.c, transports/mailbox/mboxpath.c, router/rtsyslog.c: Looking deep into "gcc -Wall" diagnostic reports. Added includes for missing prototype declarations. * router/libdb/bind.c, smtpserver/mxverify.c, transports/libta/dnsgetrr.c, transports/smtp/getmxrr.c, utils/mxverify/mxverify-cgi.c, include/zresolv.h: Modern resolver libraries don't have _getshort(), and _getlong() as public prototyped entities. Took model off modern resolver libraries, and used macroes NS_GET16() / NS_GET32() / NS_INT16SZ / NS_INT32SZ. Where macroes are not available, supplying our own versions. Centralized all related defines. * proto/zmailer.sh.in: Reworked 'resubmit' command internal mechanism to handle cases, where the 'deferred' directory has thousands of messages. A bit perl required.. * lib/mail.c: realloc()ing uninitialized variable -- OOPS! (Since 2003-4-17) * INSTALL: Small updates, more is needed... * Makefile.in, compat/rmail/Makefile.in, compat/sendmail/Makefile.in, libc/Makefile.in, man/Makefile.in, packaging/solaris/Makefile.in, proto/Makefile.in, router/Makefile.in, scheduler/Makefile.in, smtpserver/Makefile.in, transports/errormail/Makefile.in, transports/expirer/Makefile.in, utils/Makefile.in, transports/fuzzyalias/Makefile.in, transports/hold/Makefile.in, transports/mailbox/Makefile.in, transports/reroute/Makefile.in, transports/sm/Makefile.in, transports/smtp/Makefile.in, utils/makedb/Makefile.in, utils/mxverify/Makefile.in, utils/perl/Makefile.in: Rewriting the virtual-root installation support to use "DESTDIR=" instead of "prefix=". The Proper Way ... * include/mail.h.in, scheduler/readconfig.c, scheduler/scheduler.c, smtpserver/debugreport.c, transports/errormail/errormail.c, transports/expirer/expirer.c, transports/hold/hold.c, transports/sm/sm.c, transports/libta/lockaddr.c, transports/libta/tasysyslog.c, transports/mailbox/mailbox.c, transports/reroute/reroute.c: Further cleanup after inserting Eugene's patch-cluster. Adding #include "libc.h" usually, and various other merryments it involves. 2003-05-05 Matti Aarnio * SiteConfig.in, configure.in, include/libz.h, scheduler/transport.c, smtpserver/cfgread.c, transports/smtp/smtp.c, lib/Makefile.in, lib/zgetbindaddr.c: Patch cluster from Eugene Crosser to librarize common code, and to introduce 'BINDADDR' ZENV variable. * compat/sendmail/sendmail.c: If process environment has ZCONFIG variable set, does read ZENV variables from that place. * proto/mailrm.sh.in, proto/newaliases.in, proto/newfqdnaliases.in, proto/post-install.sh.in, transports/expirer/manual-expirer.in, transports/reroute/manual-rerouter.in, utils/policy-builder.sh.in, utils/rotate-logs.sh.in, utils/zmailer.init.sh.in: Eugene Crosser noted, that not all scripts inherit / are otherwise configurable to use alternate (than configured-in) value of ZCONFIG. 2003-05-02 Matti Aarnio * smtpserver/contentpolicy.c, smtpserver/readpolicy.c, include/policy.h, smtpserver/policytest.c, smtpserver/policytest.h, doc/guides/smtp-policy, proto/db/smtp-policy.src: Added 'filtering +' phrase to smtpserver's IP ADDRESS tests. * utils/perl/mailq/Makefile.PL, utils/perl/mailq/mailq.pm, utils/perl/mailq/MANIFEST, utils/perl/mailq/README, utils/perl/mailq/Changes, utils/perl/Makefile.in, utils/Makefile.in, configure.in: Made old perl-test tool into proper OO Perl code. Requires perl 5.8. (I think). * scheduler/agenda.c, scheduler/msgerror.c, scheduler/threads.c, scheduler/scheduler.c, scheduler/prototypes.h: Debugging the long living thing where "mailq -Q" mode reports negative counts for stored messages. Apparently fixed it! ( Didn't :-( ) 2003-05-01 Matti Aarnio * smtpserver/smtpserver.c, smtpserver/smtpserver.h, man/smtpserver.8, proto/smtpserver.conf.in: tar-pit things from James MacKinnon. Actually diff against 2.99.55, which has been adapted to current CVS. * proto/zmailer.sh.in: Quote typo introduced when adding --zconfig option.. Oops. 2003-04-30 Matti Aarnio * Makefile.in: Version 2.99.56-pre5 * README.UPGRADING: The router internal logic change IS script incompatible! Carefull out there with your aliases.cf, and fqdnaliases.cf ! * proto/cf/aliases.cf, proto/cf/fqdnalias.cf, router/functions.c: - When a list-expansions fails to produce recipient list, do produce an error message only, when the message being routed is _not_ an error message in itself. - In scripts, do understand that the 'listexpand' can yield a NULL list, and branch at it conditionally to do different codepath behaviour. * scheduler/update.c: Change in debug outputs to use more usefull spoolid reference. 2003-04-28 Matti Aarnio * transports/smtp/smtp.c, transports/libta/diagnostic.c: Allow multiple calls to diagnostic() function, and make sure that only first of them does actual DIAGNOSTIC reporting, all the rest do just logging to given FILE*. * transports/smtp/smtp.c: More code re-arrangements regarding connection retries. * transports/smtp/smtp.c, transports/smtp/smtp.h: A bit experimental re-arrangements of EZMLM related processing of messages in smtp-level. Moved smtp connection formation from before deliver() to within it, so that it can (when needing) re-open the connection. * utils/mxverify/mxverify-cgi.c: Auto-detect, if can use IPv6 addresses in testing. * scheduler/mailq.c, man/mailq.1: Added '-Z' option to mailq, alike at scheduler, et.al. 2003-04-17 Matti Aarnio * transports/libta/ctlopen.c, transports/libta/mimeheaders.c: Cleaning up the for-clause code a bit. Removing unnecessary malloc()/free() things, which are dropping core at one test system. (In the end it was mallocated buffer end overscribling that dropped those cores at mysterious places, but as this was the only altered place...) * man/zmailer.1, proto/zmailer.sh.in, router/router.c, man/router.8, scheduler/scheduler.c, man/scheduler.8, smtpserver/smtpserver.c, man/smtpserver.8, transports/sm/sm.c, transports/smtp/smtp.c, transports/errormail/errormail.c, transports/hold/hold.c, transports/mailbox/mailbox.c, transports/reroute/reroute.c, proto/newdb.in, proto/newdbprocessor.in: - Zmailer, Router, Scheduler, Smtpserver: explicite options to define non-compiled-in path to ZCONFIG ZENV file. - Transport Agents: if ZCONFIG _environment_ variable is set, read ZCONFIG from that file. Inherits that from scheduler... - Some tools: if ZCONFIG _environment_ variable is set, read ZCONFIG from that file. Inherits that from zmailer(1). * scheduler/scheduler.c scheduler/agenda.c, scheduler/mq2.c, scheduler/msgerror.c, scheduler/prototypes.h, proto/zmailer.sh.in, scheduler/transport.c, scheduler/update.c: Automated (within 30 seconds of previous) log-file re-opening * man/manual-rerouter.8, man/reroute.8, man/Makefile.in: Man-pages for the reroute transport-agent. * transports/hold/hold.c: Long time bug in produced envelope metadata entry. * transports/reroute/reroute.c, transports/reroute/Makefile.in, transports/reroute/manual-rerouter.in: Experimental version of reroute -- e.g. command some channel/host target queue back to the router. * transports/libta/mimeheader.c: When outputting ORCPT value into "for clause", do decode the XTEXT encoded data into plain format. 2003-04-16 Matti Aarnio * SiteConfig.in, transports/libta/mimeheader.c: More beauty tuning of the for-clause processing subroutine. Added also a way to disable the addition of the for-clause via a ZENV variable. 2003-04-15 Matti Aarnio * smtpserver/mxverify.c: Another error spotting by Eugene Crosser -- for a while CVS versions have not done any sort of DNS-BL lookups correctly. (Since 2002-09-01.) * transports/smtp/smtp.c: "RSET" processing didn't -- various MAIL FROM related state variables were reset before the optional "RSET" protocol exchange was done, and thus in some cases PIPELINEd processing did lead to undesirable behaviour. This bug was introduced on 2003-02-14 in order to fix scattered around RSET code paths so that there will always be an RSET, when there is need for it. Urgh. It appeared when a message was tried to send in the connection right after something needing that RSET; thanks to reports about "there seem to be multiple messages to these troublesome destinations" for finally finding this. 2003-04-14 Matti Aarnio * configure.in: Quite old configuration thing bites us with newer autoconf produced sed-scripts. Namely @ROUTEUSER_IN_ABNORMAL_UNIX@ value in running $MAILSHARE/zmailer.conf file will cause sed to spin in place. * include/ta.h, transports/libta/ctlopen.c, transports/libta/fwriteheader.c, transports/libta/mimeheaders.c, transports/libta/swriteheader.c: Tuning a bit more of "from-clause" processing so that it sticks better within 80 columns. * libc/myhostname.c: From Eugene Crosser: find out my local hostname only once (e.g. scheduler et.al. long-running processes.) * transports/errormail/errormail.c: - If the source address is 'channel error', send the message back to domainless "postmaster" in our own machine. - If the message is to be discarded ("NOTIFY=NEVER" at all recipients, or some such), indicate that at the syslogged string. * libc/mail.c, libc/sfmail.c, SiteConfig.in: Add INPUTDIRS= ZENV variable processing, also INPUTDIRHASH= and INPUTNOTIFY= ! Presently just testing! * router/db.c, router/rfc822.c: Been doing a bit of debugging of one odd system built a while ago, with reports "this isn't working.." (definitely it wasn't, wrong configs running...) 2003-04-11 Matti Aarnio * SiteConfig.in, configure.in: Updated default configuration values a bit, rearranged SiteConfig.in parts to be more "groupped by function". 2003-04-07 Matti Aarnio * lib/sleepycatdb.c: Eugene Crosser noted, that this file fails to compile, when no SleepyCat DB is available in the compilation environemnt. Oops! Adding couple conditionals to appease the compiler... 2003-04-06 Matti Aarnio * lib/sleepycatdb.c: Adding debugging things, trying to understand couple CDB-mode related problem things. Possibly this is about a bug in system thread library, which interacts badly with SleepyCat DB 4.0.14 ? 2003-04-04 Matti Aarnio * transports/smtp/smtp.c: Urgh, two chain-gang movement pointer mistakes, and a long running process went into busy spinning... * transports/libta/mimeheaders.c: Additional book-keeper things for knowing when the thing can fold, and when it can't. * router/rfc822hdrs.c: Tuning (part of) folding header printing code. 2003-04-03 Matti Aarnio * transports/sm/sm.c: Oops, bad parameter for header_received_for_clause() function. Used one was always NULL, which is kind of bad idea... * transports/smtp/smtp.c: Finished complete RFC 1123 part 5.2.2 CNAME rewriting both in MAIL FROM and RCPT TO. Also cleaned up code somewhat in relevant parts. * transports/smtp/smtp.c, transports/smtp/getmxrr.c, transports/smtp/smtp.h: Third round, completing processing also in cases, where target domain is not the same as to what is being used for connection forming. * transports/smtp/smtp.c, transports/smtp/getmxrr.c, transports/smtp/smtp.h: - getmxrr() -- prepared for further refinements of CNAME cache related lookups. - add_cname_cache(), cname_lookup() -- two interfaces to RFC 821 related RCPT TO rewriting for CNAME aliased domain names. This DOES NOT touch on visible headers, nor ORCPT data, only RCPT TO domain. Still a bit incomplete... Actually that reference is: RFC 1123 part 5.2.2 2003-04-02 Matti Aarnio * man/smtpserver.8: Forgotten document of several months old PARAM tls-random-source /sock/path parameter. * include/ta.h, transports/libta/mimeheaders.c: New header_received_for_clause() function to add a FOR clause into top-most RECEIVED: header. * transports/smtp/smtp.c, transports/mailbox/mailbox.c, transports/sm/sm.c, transports/errormail/errormail.c: Use header_received_for_clause() function. * transports/smtp/smtp.c, transports/smtp/getmxrr.c, transports/smtp/smtp.h: Follow strictures of RFC 821 about rewriting RCPT TO envelope domain, if target domain did resolve via CNAME. Actually that reference is: RFC 1123 part 5.2.2 * transports/smtp/smtp.c: If target system has DSN capability, but our sent address does not (for some reason) have DSN data, generate them from default values... * router/daemonsub.c: When the child process closes, as its final act, do: exit(0) * smtpserver/cfgread.c: A bit of cleanup of the parser: wrap all OpenSSL related options inside appropriate IFDEF test, AND from now on, begin reporting in smtpserver's verbose log, if configuration contains PARAM entry which isn't known/doesn't have enough parameters. (Helps debugging...) 2003-03-13 Matti Aarnio * scheduler/update.c, scheduler/statuslog-ids-to-spoolids.pl, man/scheduler.8: Change in statistics-log file format to have easier correlation with syslogged data. Also a tool to convert the statistics-log file to new format. 2003-03-06 Matti Aarnio * smtpserver/rfc821scn.c: Turned off the code assigning semantic meaning to the local-part contained % and ! characters. Relaying policy control will look at them, but only when the address domain is _local_, and the address is on RCPT TO. * man/smtpserver.8: Cleaned up a bit the OpenSSL/TLS related documentation. Added missing 'h' option document. * smtpserver/smtptls.c: Changed couple internal warning texts into a lot less alarmistic "could not load .. from file, will use built-in default" (of DH pamaters.) * transports/libta/mime2headers.c: Limit the length of produced "MIME-2" header tokens. This is mostly a beauty-wart, but might in some cases be usefull thing. 2003-03-05 Matti Aarnio * include/mailer.h, router/rfc822.c, router/rfc822hdrs.c, router/rtsyslog.c, lib/taspoolid.c, router/prototypes.h, smtpserver/smtpserver.c: - Generate file taspoolid at message envelope parse time into 'struct envelope', and pass it around instead of generating it at multiple times. - In Received: header, former id is now changed to: id spoolid (without angle-brackets) - Message-id generation uses now taspoolid value as one component of its formulation. - explicitely constified some constant data. * router/functions.c: Comment away dead declarations. * proto/smtpserver.conf.in: Comment away the default PARAM help text where "Yoyodyne Propulsion" is mentioned as system owner... People don't edit those... 2003-02-28 Matti Aarnio * proto/cf/standard.cf, router/libdb/header.c: Remove headers which we produce into e.g. messages stored in mailbox / processes run in pipes, etc. * router/rfc822hdrs.c: When producing our Received: header, don't anymore put the spool-id inside angle-brackets. 2003-02-20 Matti Aarnio * smtpserver/smtptls.c, transports/smtp/smtptls.c: Updated one function prototype to match version 0.9.7. * configure.in: In RedHat (8.0 / Rawhide ?) system the OpenSSL 0.9.7 is compiled with Kerberos inside it. Compilation, and linkage will need additional CFLAGS and LIBS to succeed. To provide those, a system specific pkg-config tool is used, where available. * configure.in: Updated things to match rules of autoconf 2.57 (e.g. LIBOBJS is no longer allowed direct assignment) * scheduler/mailq.c, smtpserver/smtpserver.c: Typo fixed from Eugene Crosser. * TODO: Updated / rearrangements / removing completed items 2003-02-14 Matti Aarnio * Makefile.in: Version: 2.99.56-pre4 * WILL DO ALSO A RELEASE TARBALL OF THIS, * * BUT DON'T TRUST IT YET! ***************** * transports/smtp/smtp.c, transports/smtp/smtp.h: Centralized RSET sending into one place in MAIL/RCPT/DATA loop. Using state variable do_rset in the SmtpState context to indicate POSSIBLE need for doing it. Definite non-need is when DATA/BDAT has completed successfully, and also right before the first MAIL. * router/libsh/bsdhash.c, router/libdb/bsdbtree.c: Correct clearing of DBM private data storage pointer. * scheduler/mailq.c: Fix "mailq -s" access failure printing to use correct directory string pointer. 2003-02-06 Matti Aarnio * compat/sendmail/sendmail.c, libsh/builtins.c, libsh/execute.c, router/libdb/bsdbtree.c, router/libdb/bsdhash.c, router/libdb/ndbm.c, transports/libta/ctlopen.c: Do #include -- that is definitely needed for correctly compiling code at glibc 2.3.1+, but is correct for other cases, too. * transports/smtp/smtp.c: Convert couple LOG_ERR level syslog()s to LOG_NOTICE. They don't quite deserve being blasted to people's consoles.. * transports/smtp/smtp.c, transports/smtp/smtp.h, transports/smtp/smtptls.c: Additional "writeclosed" flag to inform that the socket shall not be written to, no matter what. * libsh/execute.c, libsh/zmsh.c: Added option 'J' into ZMSH interpreter to trace a subset of 'I' traced things. * libsh/expand.c, libsh/interpret.c: Disable eating away backquotes from expanded variable values. EFFECTS TO BE TESTED MORE THOROUGH! * proto/cf/aliases.cf, proto/cf/rrouter.cf, proto/cf/server.cf: Moved $(dequote ...) calls into separate variable assignments to analyze things for backquote (non-)processing. * libsh/path.c: Don't include , it isn't needed. * router/daemonsub.c: If message has been (tried to) process earlier, and the "lock pid" is ours, do process the message, otherwise see at first if the system has a process with that pid in it, then skip that item.. (FIXME: this is not perfect...) * libsh/regex.h: Moving one #endif upwards a bit to make sure there is always some definition for __restrict et.al. 2002-10-07 Matti Aarnio * smtpserver/smtpcmds.c, smtpserver/smtpserver.c, smtpserver/smtptls.c: Oops.. Wrap HAVE_OPENSSL around data entries which are available only when the macro is defined... 2002-10-04 Matti Aarnio * smtpserver/smtptls.c: Make CA file/directory issuance optional for smtpserver. * router/daemonsub.c: E.Crosser reported that on some platform (apparently at least Solaris), child process death causes file- handle leakage in process. (E.g. missing close() calls somewhere.) While at it, I cleaned a bit of the code.. * lib/loginit.c: Do lessen the number of syscalls done by this routine. (At Linux with glibc 2.2.x). 2002-09-23 Matti Aarnio * compat/sendmail/sendmail.c: Improved the verbose-trace printing code. Turned out the lattest glibc stdio code didn't resume reading from fd, when it had reached EOF once, or some such equally weird thing... Also noticed temporary file usage insecurity whereby somebody could persuade us to append texts into some existing file in systems where mkstemp() call does not exist.. (Not overwrite, nor read what invocator's privileges don't allow to be read...) * transports/smtp/smtp.c, transports/smtp/smtp.h, transports/smtp/smtptls.c: - If TLS mode is mandated, but not available, yield a TEMPFAIL at connection.. Hopefully sysadmin will detect it before a full-blown failure occurs, and traffic is lost. - Also included template code for handling server certificate verification failure at SMTP connection level. (final integration is a TODO thing,) - A bit more of Postfix TLS code into SMTP client system. (Still the session-cache in TODO..) 2002-09-21 Matti Aarnio * smtpserver/smtpserver.h, smtpserver/smtpserver.c, smtpserver/smtpcmds.c, smtpserver/cfgread.c, smtpserver/smtptls.c: Radical revisioning; adapting current Postfix TLS code into ZMailer environment. TODO: session-cache... * transports/smtp/smtp.h, transports/smtp/smtp.c, transports/smtp/smtptls.c: Radical revisioning; adapting current Postfix TLS code into ZMailer environment. TODO: session-cache... 2002-09-19 Matti Aarnio * transports/smtp/smtptls.c, transports/smtp/smtp.h, man/smtp.8, proto/smtp-tls.conf.in: The OpenSSL code doesn't (anymore) like to work without proper random pre-seeding... 2002-09-10 Matti Aarnio * scheduler/scheduler.c: Default the 'expiry2' to be zero, if it can't be set to be anything more appropriate. 2002-09-06 Matti Aarnio * SiteConfig.in: Changed the default value for SCHEDULERDIRHASH to be 2 It is important thing for large queue sites, while at small systems it uses deep hashing unnecessarily.. * configure.in: Mark for presense of header * man/scheduler.8, scheduler/scheduler.c: The -E option for the scheduler can now have two parameters to better control the pre-scan of the input queue. Also altered default values. * scheduler/readconfig.c, scheduler/scheduler.h, man/scheduler.8, scheduler/agenda.c, scheduler/conf.c, scheduler/scheduler.c, scheduler/prototypes.h, scheduler/threads.c: Another form of expiry -- when the queues become truly huge, ZMailer's scheduler has not been able to expire messages because it must at first do a delivery attempt at least once. A new "expire2=" time limit _beyond_primary_ _expiry_. * smtpserver/mxverify.c: The bad T_TXT data detection and cutback, nominally the same thing as with e.g. sendmail. Also clean up the received string to have only printable characters in ISO-8859-* character-sets. All that this could have done in ZMailer is to pick garbage data from beyond the (received) buffer, not overwrite anything. * contrib/clean-mean-contentfilter-1.00.tar.gz: Package from Eugene Crosser: "if you think it's worth, maybe you'll include the attached thing into contrib/. It's running here for more than a week under serious load. It's not very smart and not very portable but it's light (I did not feel any noticable growth of load on the servers)." 2002-09-03 Matti Aarnio * smtpserver/smtpserver.c, smtpserver/zpwmatch-pipe.c: Running the legacy application authentication against SASL via an external program.. Turns out the old pipe-matcher code fails (by looping with ECHILD error) when we have active child exit reaper... Oops. * smtpserver/smtpauth-sasl.c: The external program mentioned above. This program is NOT to be considered as official part of ZMailer MTA, the goal is to provide an official version in the smtpauth.c at some point... I just needed a QUICK and working thing. * smtpserver/smtpcmds.c: From user feedback (helpdesk manager report to developers, wow...): The "rejection due to MX rule" report needs improved text. Altered wording, and added client IP address printout into the text. * router/prototypes.h, router/rfc822.c, lib/rfc822scan.c: Modifications to header parsing. - An all-white-space line shall be CONTINUATION, not end-of-headers (ancient BITNET card-images are other story, but lets not worry about them...) - Altered the makeLetter() function parameters removing a parameter that has not been used for several years. - WorkInProgress: internal changes into makeLetter()... * configure.in, smtpserver/smtpauth.c: Initial configuration hooks for enabling CMU Cyrus Sasl-2.x The code itself is NOT functional as of yet. 2002-09-02 Matti Aarnio * router/rfc822.c: Did something wrong, while scrubbed dead code.. Detection of header/body (in RFC822 sense) separation fixed. (Well, not entirely.. there are things which probably SHOULD stay in headers...) * transports/libta/diagnostic.c: A beauty-wart in syslog()ed data, when truncking embedded CR containing diagnostic message, skip over the last CR also... * transports/smtp/smtp.c: Support -F connection to UNIX:/... sockets. This MUST be defined at the command line! No routing controlled version allowed! (security, you see..) (This is subject to system having AF_UNIX support in the first place, but that should be quite common...) * README.UPGRADING: Change in what sm outputs for envelope meta-data to be in line with other transport-agents... (mailbox, in particular.) * transports/sm/sm.c: Testing reveals... NULL pointer referrals :-( * libc/Makefile.in, libresolv/Makefile.in, router/Makefile.in, router/libdb/Makefile.in, scheduler/Makefile.in, smtpserver/Makefile.in, transports/hold/Makefile.in, transports/libta/Makefile.in, transports/mailbox/Makefile.in, transports/reroute/Makefile.in, utils/makedb/Makefile.in, utils/vacation/Makefile.in: Modified usage of @GENINCL@ autoconfig macro so that it will always be the LAST inclusion define for the each compilation. This way if the (as has happened to yours truly) GENINCL ( /usr/local/include ) has a file with same name as something important ZMailer internal (mail.h from UW-IMAP C-client), it won't override internal file.. * proto/cf/fqdnalias.cf: When doing fqdnroute processing on a "plus-full" address, pass proper parameters to the database lookup. * include/zmsignal.h, router/daemonsub.c, router/db.c, router/router.c, router/rfc822.c: Observed certain kind of difficulties at abruptly closed database accesses -- SleepyCat DB 3.x CDS goes into bad lock state, unless properly cleaned after with atexit(). * transports/sm/sm.c: Do similar X-***: header processings as the 'mailbox' does for ORCPT, EnvelopeID, and Envelope-To. Streamline a bit separatorsequence processing. 2002-09-01 Matti Aarnio * lib/rfc822scan.c, router/functions.c, router/rfc822.c: Removed code (for printaliases function) that is no longer used, observed situations, when the malformed headers can be split by ZMailer. * smtpserver/mxverify.c: When doing varioud "RBL" lookups, always catenate a "." into end of the lookup domain in order to suppress unnecessary "search" work by the resolver libraries. Doing: _res.options &= ~RES_DNSRCH; might do it too... * smtpserver/smtpcmds.c: Noticed that in unused code there is potential for serious trouble with router call. '#if 0' wrappers in.. Also, modify SIZE= value parameter parser to accept strictly valid value in range 1 thru 1 000 000 000. * proto/db/smtp-policy.relay: Add a comment covering user-specific destination addresses. 2002-08-23 Matti Aarnio * smtpserver/smtpdata.c: Reworked the ta-spoolid generation and presentation in the DATA and BDAT phases after the contentpolicy() call (and even before it..) * transports/libta/mime2headers.c: Arnt Gulbrandsen reported a GI (of GIGO) case producing invalidly encoded header MIME token. namely the double-quote is not legal char unencoded char inside the Q-format token. * configure.in: Added missing utils/smtp-contentfilter into generated files list. * utils/makedb/dblook.c: Use db_strerror() to report error strings properly. Usage of strerror() does not work at DB3/DB4 .. * man/smtpserver.c, smtpserver/contentpolicy.c, utils/smtp-contentfilter.in, README.UPGRADING: A change in the smtpserver <-> contentfilter interface protocol! Some users have written content-filters based on the sample framework, and those will need matching changes (rather simple!) in the filter code! See README.UPGRADING ! 2002-08-22 Matti Aarnio * proto/post-install.sh.in: Changed behaviour rules a bit. Now the prototype of whatever there are prototypes of will be kept in its original place, and only copies will be made. * lib/sleepycatdb.c: Spotted a bug in database opening error handling. 2002-08-19 Matti Aarnio * transports/smtp/smtp.c: Changed logtag() returned tag-string format. New one will tell time (in secs) since start of the program, and incremental sequence number. * proto/post-install.sh: Having been asked ... printing more verbose progress reporting while doinga POSTOFFICE subdirectories. * smtpserver/smtpauth.c, smtpserver/smtpcmds.c, smtpserver/smtpserver.c, smtpserver/smtpserver.h: Format change in the verbose protocol log tags. 2002-08-12 Matti Aarnio * transports/smtp/smtp.c: In some cases the smtp_sync() is called when outbound write socket has already been killed. * proto/newdbprocessor.in: Updated internal comment carried documentation, and sample of dbases.conf in the script. * proto/db/dbases.conf: Documented the existence of "-r" as newdb_compile_option. 2002-07-11 Matti Aarnio * sfio2002/* Entire subtree -- SFIO version 2002.1 Basic injection, no adaptations yet. * include/libz.h, lib/selfaddrs.c, lib/zgetifaddress.c, router/libdb/selfmatch.c, smtpserver/mxverify.c, smtpserver/policytest.c, smtpserver/smtpserver.h, transports/smtp/smtp.c, transports/smtp/smtp.h: Moving commonly used 'Usockaddr' type into libz.h, and updating interfaces playing with addresses to use that type for the parameters. * proto/smtpserver.conf.in, smtpserver/cfgread.c: Eugene Crosser noticed that binding to named interface does not quite succeed.. Address storage goes into slightly wrong place -> binding fails. 2002-07-05 Matti Aarnio * smtpserver/smtpcmds.c: A practically unexploitable potential buffer overflow. Exploit requires presense of 'h' flag in EHLO greeting match patterns (non documented, non-default style flag), and that the host address reversal text is long in the extreme. Getting the server to crash is less difficult, than to get it to execute anything.. Repeat: A 'h' must be present in style-flags at the end of the smtpserver.conf file EHLO/HELO-patterns, AND the hostname reversal PTR text must be longer than around 100 bytes. Then, maybe, it will be possible to feed working buffer overflow exploit.. * Makefile.in: Version 2.99.56-pre3 (CVS tagged 2-99-56-pre3) * transports/smtp/smtp.c: An attempt to create a smart way (at least at Linux and at some BSD's) to de-nagle tcp streams in high-performance manner. Low-performance de-nagle is: sleep(1) ... * configure.in, acconfig.h, aclocal.m4: Updates to remove couple of deprecated features, adding third parameters (often also second) to AC_DEFINE() macroes. * proto/post-install.sh.in, proto/cf/fqdnalias.cf, proto/cd/rrouter.cf, proto/db/dbases.conf, proto/db/fqdnroutes, proto/Makefile.in: - cleaned up a bit the routes_spec() function parameter variables -- less confusion with the new name, I hope.. - new routing mechanism: fqdnroutes -- user@domain -> chan!data.. * lib/sleepycatdb.c, include/sleepycatdb.c, router/libdb/bsdbtree.c, router/libdb/bsdhash.c More rework on SleepyCat DB driver routines to better use the ZSleepyPrivate structure to carry relevant reference data. * router/libdb/gdbm.c, configure.in: Similar code work to SleepyCat DB for GDBM. Testing for library having gdbm_fdesc() function. * router/db.c, router/function.c, router/prototypes.h, router/libdb/search.h: - Renamed the low-level dblookup driver routine from 'db()' to 'dblookup()' after having problems of finding the caller... - Reworked the %-substitution code radically; now %[1-9] will be a key of same index at every time. The %0 MAY change in case some driver routine (e.g. 'pathalias') is defined at the 'relation' statement. The relation MUST have '-%' option for the %[0-9] substitutions to be active! 2002-06-25 Matti Aarnio * transports/smtp/smtp.c: Separating DATA ending dot into separately done write(2), which hopefully leads into separately pushed out frame, and finally packet separation at the DATA-phase ending dot sending.. World is full of braindead firewalls... * include/sleepycatdb.h, lib/sleepycatdb.c, router/libdb/bsdbtree.c, router/libdb/bsdhash.c: More work at the SleepyCat DB Concurrent Data Store support. * router/db.c: Clean a bit of the database call mechanism passed parameter block by doing initial memset(,0,) on it.. * libc/mail.c, libc/sfmail.c: Change (slightly) of how the initial temporary file name is formed so that there are no leading white spaces at a host which has shorter than 7 chars hostname.. 2002-06-20 Matti Aarnio * include/sleepycatdb.h, lib/sleepycatdb.c, proto/newdbprocessor.in: More of SleepyCat DB support stuff. * smtpserver/smtpauth.c: Stole code (yet incomplete, and not integrated) from sendmail 8.12.* for using SASL code in SMTP server authentication. 2002-06-18 Matti Aarnio * configure.in, acconfig.h: Full support for finding SleepyCat DB 4.0, and to inform the system about it. * lib/Makefile.in, lib/sleepycatdb.c, include/sleepycatdb.h: Librarize support for SleepyCat DB by allowing more versatile parametrisation, e.g. DBENV things, etc. This to make it possible to work with "Concurrent Data Store" kind databases. E.g. single-writer, multiple readers to allow online modifying of the databases. * router/db.c, router/libdb/search.h: For each online database, supply a private variable usable by the database method itself for whatever needs it has (e.g. storing own structure with private data, like DB structure handle.) * router/libdb/bsdbtree.c, router/libdb/bsdhash.c: Usage of librarized SleepyCat DB interface, and radical surgery to discard a bunch of splay-trees that are no longer necessary as the database method handle carries "dbprivate" pointer. * smtpserver/policytest.c, utils/makedb/dblook.c, utils/makedb/makedb.c, utils/vacation/vacation.c: Support SleepyCat DB 4.0; use (where can) librarized support interface. * transports/smtp/smtp.c: Fix the call parameters to "select_sleep()" so that there is explicite knowledge of when the select is for writing, and when it is for reading. * utils/smtp-contentfilter.in: Fix originally inverted test on meta-environment scanner stopping condition part. 2002-05-28 Matti Aarnio * smtpserver/smtpserver.c: Eugene Crosser reported smtpserver spinning in place, and that is indeed possible if the deprecated ssmtp service is in use. Moving relevant initialization code upwards slightly fixes the problem. (Later Eugene admitted that there could have been a pilot error -- omitting "make clean" before "make".) 2002-05-20 Matti Aarnio * smtpserver/smtpserver.c: In some testing cases the "progname" variable is not set with any sane (non-NULL) value. * utils/policy-builder.sh.in: Output the "smtp-policy.relay*" file entries only, when there is some data in the line, e.g. no comments, nor blank lines.. * smtpserver/contentpolicy.c: A bit more of debug output into local log.. * transports/smtp/smtp.c: Sometimes the SS->smtpfp will be NULL, and at such times sferror() on stream yields SEGV.. Additional code armouring. * utils/smtp-contentfilter.in: A bit cleaner use of perl's implied default variables.. 2002-05-16 Matti Aarnio * scheduler/transport.c: The code didn't consume the scheduler-notify input socket at the mux() code, although it was looked for by the select() in the code. 2002-04-20 Matti Aarnio * transports/smtp/smtp.c: In the obscureish EZMLM mode we LACKED mandatory trailing ">" character at MAIL FROM envelope address. A major goofup.. :-/ (Especially as I use the beast myself sometimes..) * transports/mailbox/mboxpath.c: At the usage() do document also the -X[X] option. 2002-04-18 Matti Aarnio * transports/mailbox/mailbox.c: Finally a bit of understanding about how to manage persistent errno status on sfio streams. * transports/smtp/smtp.c, transports/smtp/smtp.h, transports/smtp/smtptls.c: Persistent error storage things, but as of yet the detail analysis/reporting at calling layers is vague. * libc/mail.c, libc/sfmail.c, router/rfc822.c: Add missing #include "zmsignal.h" lines... 2002-04-17 Matti Aarnio * include/ta.h, transports/libta/diagnostic.c, transports/errormail/errormail.c, transports/smtp/smtp.c, transports/expirer/expirer.c, transports/hold/hold.c, transports/libta/ctlopen.c, transports/mailbox/mailbox.c, transports/smtp/smtp.h: - Add 'FILE *verboselog' parameter to diagnostic() API call, and get a cascade of things.. That is often the last, undocumented, and most interesting thing anyway.. - Play with MAILBOX to detect (and neatly report) when write to a file fails due to EDQUOT ... Perhaps fail it immediately ? (TODO item..) 2002-04-12 Matti Aarnio * transports/smtp/getmxrr.c: Extended the test-harness a bit by telling what are the two standard tests for it, and their expected results. 2002-04-05 Matti Aarnio * transports/smtp/smtp.h, transports/smtp/smtp.c: Debugging oddball 'yield OK when DEFER is correct' problems.. .. and fixed the blunder in the call of smtpopen(). * transports/libta/ctlopen.c: The RP data block is not initialized completely in all cases, which causes occasional bugs. Doing memset() for it at first, then filling in non-zero values. This resulted from RP structure expansion a month ago. 2002-03-08 Matti Aarnio * transports/reroute/reroute.c, transports/reroute/Makefile.in, transports/Makefile.in: Added a new Transport Agent -- "reroute", e.g. send given messages back to router. This is intended for quite special (and intimate) interaction with the Scheduler. * transports/mailbox/mailbox.c, man/mailbox.8: Add a few subprocess execution environment variables which carry more of message meta-data, than previously. * transports/libta/ctlopen.c: Pass the INFROM= DSN metadata into rp->infrom field. * transports/hold/hold.c: Detect "channel error" type messages properly. Pass more of DSN Metadata to outgoing message envelope. Output "env-end\n" meta-header after the envelope headers. * transports/expirer/expirer.c: Update the comments a bit. * router/rfc822.c, libc/mail.c, libc/sfmail.c: Take precautions (just in case) "the next-stage notify" socket closes, and causes a SIGPIPE to be sent to relevant process. * proto/scheduler.conf.in: Remove "-8" from a few prototype command lines. 2002-02-22 Matti Aarnio * smtpserver/policytest.h, utils/makedb/makedb.c, utils/makedb/dblook.c: System with only SleepyCat DB 3.x in it (not 2.x, nor 1.x) didn't support it properly. Oops.. * configure.in, libc/Makefile.in, man/Makefile.in: Found out, why --libdir=, --includedir=, and --mandir= options didn't have effect. D'uh.. 2002-02-20 Matti Aarnio * router/rfc822.c: Remove few earlier added ":" characters from places where those went into report form filenames. My blunder :-( 2002-02-18 Matti Aarnio * utils/makedb/makedb.c: Remove else -brances causing problems in some cases. I got definite reports about this earlier, but had solid neutronium in my skull... 2002-02-15 Matti Aarnio * libc/mail.c, libc/sfmail.c, router/rfc822.c: Oops, MSG_NOSIGNAL seems to be Linux specific thing, at least Solaris 2.6 doesn't support it for sendto() system call. 2002-02-11 Matti Aarnio * scheduler/transport.c: Match the mailq socket tcp-wrapper behaviour to that of what the man-page says -- "tcp-wrapper is not used for MAILQ-v2 type sockets". * SiteConfig.in: By default publish SCHEDULERDIRHASH=1 value. Publish also SCHEDULERNOTIFY= and ROUTERNOTIFY= variables, which are used by the notifier codes. * include/libc.h, transports/libta/nonblocking.c: Moved the non-blocking code to be macro in libc.h. At some places a #include is needed before #include "libc.h". * include/ta.h, scheduler/mq2.c, scheduler/transport.c, transports/libta/warning.c: The preceding fd-blocking macro thing affected these files. (Other usage places already did suitable #includes) * libc/mail.c, libc/sfmail.c, router/daemonsub.c: If ROUTERNOTIFY is defined in zmailer.conf, then mail_close() (not .._alternative() !) will presume there might exist an AF_UNIX socket with that name into which a notification message about a new job can be sent. * router/rfc822.c, scheduler/transport.c: The Scheduler Notify socket got changed to use SCHEDULERNOTIFY= ZENV variable, and be AF_UNIX socket. 2002-02-07 Matti Aarnio * scheduler/scheduler.c, scheduler/conf.c, scheduler/prototypes.h, scheduler/transport.c: Implemented a 'notifysocket', which is an UDP socket (mailq/udp) receiving simple "NEW X/Y/123-456" text lines, and verifying the inputs a bit.. Perhaps should have this configurable, so that the system could have multiple schedulers around... * scheduler/scheduler.c: Reworked 2001-10-27 Stephen scheduler queue rescan timer thing into less magic.. 2002-02-05 Matti Aarnio * router/db.c, proto/cf/p-routes.cf, router/libdb/search.h: Support _multiple_ instances of fallback keys, e.g.: $(routesdb "$domain" -: ".:$(get $A type)" -: .) which will look up fallback keys: .:sender .:recipient depending on address type, and . if neither of those exist.. 2002-02-01 Matti Aarnio * proto/cf/fqdnaliases.cf: Instead of: returns (bitbucket ...) do it properly: returns (((bitbucket ...))) * scheduler/update.c: At u_error2() log the name of the event fully. * router/libdb/bind.c: h_errhost[] change into source material length safe construction. * router/libdb/dbm.c: Mark this file with a "#warning" for being suspicuously old-fashioned system with various problems... 2002-01-31 Matti Aarnio * proto/newdb.in: Match the test in router/libdb/ndbm.c about which component file to move _last_. This has been a bug since 2000-05-18 ... 2002-01-31 Alexey Antipov * router/libdb/bind.c: h_errhost[] buffer overflow in some oddball cases. 2001-12-15 Matti Aarnio * man/sm.8: Revising the man-page. 2001-12-15 Nicolas Baumgarten * transports/sm/sm.c: Added 'O' flag to configuration to output ORCPT data into X-Orcpt: header. 2001-12-07 Matti Aarnio * transports/mailbox/mailbox.c, transports/libta/swriteheaders.c: When message writing into a file fails with e.g. EDQUOT (quota exceeded), abort writing completely, discard sfio buffers, and truncate away the incompletely received message. This problem appeared at Solaris 8, when using quotas. * transports/smtp/getmxrr.c: Cleanup some compiler warnings. 2001-11-07 Matti Aarnio * transports/smtp/smtp.c, transports/smtp/smtptls.c, man/smtp.8: Trying to figure out what to do with problematic case on a Solaris 2.6 system speaking to some Win-NT SMTP server application which looses all but the first two TCP frames it receives. The Solaris goes to retransmit backoff from which it never really recovers, and the write connection times out. Implemented thus a '-w 1000' option to define the value of the socket option SO_SNDBUF. 2001-10-27 Stephen * scheduler/scheduler.c: Observed problems with scheduler's input queue scanning. Sent in a patch (which didn't quite do it..) 2001-10-25 Matti Aarnio * transports/smtp/smtp.c, transports/smtp/smtptls.c: - edge comparisons in time handling code cause short (1 sec) busy-loopings every now and then. - When MAIL FROM results a 5xx status with immediate closure of the socket following that report, yield proper rejection, not EX_TEMPFAIL. * transports/smtp/smtptls.c, transports/smtp/smtp.h, transports/smtp/smtp.c, transports/libta/ctlopen.c, transports/libta/tasyslog.c, include/ta.h: Move the "taspoolid()" invokation into ctlopen() processing from within tasyslog() call. At smtp, copy also the "->taspoolid" pointer field into the SmtpState structure. 2001-10-23 Matti Aarnio * scheduler/msgerror.c, scheduler/scheduler.c, scheduler/scheduler.h, scheduler/update.c: Added couple zsyslog() calls to conditionally syslog when scheduler has completed processing of a message. * router/rfc822.c: Added a bit of double-colons on squirel() and reject() calls when the thing is FATAL. (Its me again being dense and forgetting documentation elsewere, but the ":error-on-error" tag looked so neat...) * router/libdb/bsdbtree.c, router/libdb/bsdhash.c, smtpserver/policytest.c, utils/makedb/dblook.c, utils/makedb/makedb.c: Reworked SleepyCat DB 1.x/2.x/3.x conditionals. 2001-10-21 Matti Aarnio * scheduler/mq2.c: Tracked wrong resource accounting at mailq2 code; it used to increment the number of (tracked) childs in a place where such increment is harmfull, and did lead to sort of "inflation" of children to such an extent that there were phantoms in all allowed resource slots.. Thus the 'mailq -Q' thing begun to fail in most annoying way. Also allowed for 'challenge' subfield to be filled with a NULL pointer by the strdup(), which a bit later could lead to NULL referral, and thus SigSEGV.. 2001-10-21 Michael Loftis * scheduler/scheduler.c: In queue injection spotted a test of time which is reversed! 2001-10-16 Matti Aarnio * scheduler/mq2.c: Correct the way how orderly close ("QUIT") of mailq is handled -- the socket needs to be discarded properly... * libsh/builtins.c, router/functions.c, router/daemonsub.c: Fixup the various return codes, specifically use EX_USAGE for the usage report. 2001-10-10 Matti Aarnio * proto/cf/aliases.cf, proto/cf/aliases-new.cf, proto/cf/canon.cf, proto/cf/crossbar.cf, proto/cf/fqdnaliases.cf, proto/cf/server.cf, proto/cf/fussy-aliases.cf, proto/cf/i-smtp.cf, proto/cf/i-uucp.cf, proto/cf/p-bitnet.cf, proto/cf/rrouter.cf, proto/cf/standard.cf: Noticed strange bounces which turned out to be due to calls of functions which have optional options beginning with '-' character. These fixes handle cases where localpart and/or domain name begins with a minus. A full throughout analysis of all functions inside the router, and related documentation in the ZMailer Manual is future project. 2001-10-09 Matti Aarnio * man/smtpserver.8, smtpserver/smtpserver.c, smtpserver/smtpserver.h, smtpserver/smtpcmds.c, smtpserver/cfgparse.c, proto/smtpserver: PARAM force-rcpt-notify-never Whatever the incoming DSN NOTIFY= value is, we force it always to be NOTIFY=NEVER. This is for those who won't like to let others even to find out that the message made into the system, but still want to support incoming DSN. (That is, not to disable DSN!) 2001-10-04 Matti Aarnio * proto/cf/p-smtp.cf, proto/cf/standard.cf, proto/cf/trusted.cf: Add a bit of 'provide nnn' things. * scheduler/mq2.c, scheduler/mq2auth.c, scheduler/prototypes.h, scheduler/transport.c, scheduler/update.c: Track the number of open mailq2 sockets. Track also memory usage, and abort the connection of MQ2 socket if the buffer size needed for the input line exceeds 16k. 2001-10-02 Matti Aarnio * scheduler/readconfig.c: Support arbitrary large sets of: AAAA/* BBBB/* CCCC/* DDDD/* params channel/host selector pairs preceding common parameter sets. Original was just AAAA+BBBB, adding third did fail the first dataset... * proto/cf/crossbar.cf, proto/cf/fqdnalias.cf, proto/cf/p-uucp.cf, proto/cf/rrouter.cf, proto/cf/standard.cf: A bit more of revisions of 'returns' vs. 'return' primitives. Mostly adding comment: "error code return" * libsh/builtins.c: Damn! The 'returns' primitive WAS NOT returning _string_ in all cases. 2001-10-01 Matti Aarnio * configure.in: SH-script bug in SleepyCat DB 3.x tests. * proto/newdbprocessor.in, proto/cf/UTAI.cf, proto/cf/UTGPU.cf, proto/cf/aliases-new.cf, proto/cf/aliases.cf, proto/cf/canon.cf, proto/cf/crossbar.cf, proto/cf/fqdnalias.cf, proto/cf/p-routes.cf, proto/cf/p-iproutes.cf, proto/cf/p-smtp.cf, proto/cf/p-usenet.cf, proto/cf/p-uucp.cf, proto/cf/rrouter.cf, proto/cf/standard.cf: Reports about problems when aliases map into purely decimal digit containing local name prompted rewriting parts of the scripts (and newdbprocessor tool!) with old(*) 'returns' return in place of the original 'return' return. The difference is that 'returns' will _always_ yield a string, while 'return' may yield an integer value if the username/whatever is made of purely decimal digits... ( *: "old" as "long existing" ) * Makefile.in: CVS tagged version 2.99.56-pre1 * include/libc.h, proto/zmailer.sh.in, man/zmailer.1: Update the beast at Solaris 2.6, and learn some new things... 2001-09-30 Matti Aarnio * configure.in, acconfig.h: Autoconfig detection of SleepyCat DB 3.x * router/db.c, router/libdb/bsdbtree.c, router/libdb/bsdhash, smtpserver/policytest.c, utils/makedb/dblook.c, utils/makedb/makedb.c, utils/vacation/vacation.c: Use the SleepyCat DB 3.x call API; even prefer it! (Except for vacation: Prefer NDBM, then GDBM, BSD DB...) * README.UPGRADING: Note the usage of SleepyCat DB 3.x ! * transports/mailbox/mailbox.c: Got a spotting of code which isn't needed.. * utils/rotate-logs.sh.in: Let the $PATH to handle command binary locating. That is, 'mv' and 'touch' are trusted to $PATH... * proto/zmailer.sh.in: Fix typos at '-q' option processing! 2001-09-27 Matti Aarnio * packaging/rpm/*: Updated version of RPM specs file, et.al. from Xosé Vázquez 2001-09-26 Matti Aarnio * smtpserver/policytest.c: For HELO/EHLO parameters, do a bit less work, ignore case of '[1.2.3.4]' type parameters completely. * proto/scheduler.conf.in: Set default expiration to be 5 days (from old 3 days). RFC 2821 does suggest 4-5 days. 2001-09-22 Matti Aarnio * acconfig.h, config.h.in, configure.in, include/hostenv.h, lib/nobody.c, lib/trusted.c, router/db.c, router/functions.c, router/rfc822.c, router/libdb/nbm.c, scheduler/scheduler.c, sfio/src/lib/sfio/sfio.h, sfio/src/lib/sfio/Sfio_dc/sfdcdio.c, sfio/src/lib/sfio/Sfio_dc/sfdcdos.c, smtpserver/policytest.c, sfio/src/lib/sfio/Sfio_dc/sfdcfilter.c,smtpserver/policytest.h, sfio/src/lib/sfio/Sfio_dc/sfdcseekable.c, smtpserver/smtpcmds.c, sfio/src/lib/sfio/Sfio_dc/sfdcsubstream.c, transports/hold/hold.c, sfio/src/lib/sfio/Sfio_dc/sfdctee.c, transports/expirer/expirer.c, sfio/src/lib/sfio/Sfio_dc/sfdcunion.c, transports/fuzzyalias/db.c, transports/mailbox/mailbox.c, transports/smtp/getmxrr.c, transports/sm/sm.c, transports/smtp/smtp.c, utils/makedb/dblook.c, utils/makedb/makedb.c, utils/vacation/vacation.c: A very large OpenBSD patch cluster; including: - Revision of database facility autoconfig - several printf warnings - sfio #define collision avoidance - SETEUID() and friends * proto/zmailer.sh.in: A -q option for completely quiet operation (part of OpenBSD patch cluster) * utils/mxverify/mxverify-cgi.c, utils/mxverify/getaddrinfo.c: Some radical cleanups in the TEXT/HTML presentation * transports/smtp/smtp.c: Certain diagnostics control bits fixed so that RSETs won't overwrite last real command performed on connection and having done something which e.g. caused a reject * smtpserver/cfgread.c, smtpserver/smtpdata.c, smtpserver/smtpserver.c, smtpserver/smtpserver.h: "PARAM outlook-tls-bug" -- Outlook Express seems to kick on TLS when it is told to use TLS, and the destination port is not 25, but _anything_else_ ! 2001-07-31 Matti Aarnio * smtpserver/smtpdata.c: Had omitted couple fflush() calls in DATA and BDAT to commit accumulated buffered data into file so that external contentfilter() program can analyze it. 2001-07-29 Matti Aarnio * utils/autoanswer.pl.in: Fixing buggy tests; originally about "$nosend", but there were some other too.. * include/mail.h.in: Forgot this from the inject below... 2001-07-05 Matti Aarnio * include/libc.h, include/mail.h.in, libc/mail.c, libc/sfmail.c, scheduler/msgerror.c, smtpserver/smtpdata.c, transports/errormail/errormail.c, transports/smtp/getmxrr.c: Changed (properly) the _*mail_close_() inode parameter to be 'long'. 2001-07-04 Matti Aarnio * transports/errormail/errormail.c: Recognize source channel 'error', and produce the destination address accordingly -- ugly thing to do an "error-in-error" message that way, but.. * transports/mailbox/mailbox.c: Recognize (and wipe out, and re-insert at delivery) incoming headers "Original-Recipient:" (for RFC 2298 section 2.3), and "Envelope-ID:". 2001-06-15 Matti Aarnio * transport/smtp/smtp.c: In synchronous SMTP connections, handle the result processing correctly so that permanent reject with MAIL FROM will permanently reject the message.. 2001-06-11 Matti Aarnio * utils/mxverify/mxverify-cgi.c: Print out the user supplied input (arg strings) properly protected so that even if the input happens to be HTML text in disguise, it shows up as "a string". 2001-06-04 Matti Aarnio * scheduler/transport.c, scheduler/msgerror.c: Oh no, one-off string length after cfgdirpath() function behaviour change. Got around to write \0 to child, which promptly misbehaved... * scheduler/transport.c, scheduler/conf.c: Under some mysterious conditions (Linux kernel 2.4.5-pre4. ?) the scheduler does not get select on read on dead childrens stdout ?? * smtpserver/smtpserver.h, smtpserver/smtpserver.c, smtpserver/smtptls.c, smtpserver/smtpdata.c: Got rid of all alarm() calls in the server code, and use select() instead everywhere. Along with NON-BLOCKING sockets. The previous code has apparently allowed accumulation of servers, which never die out... 2001-06-03 Matti Aarnio * scheduler/mailq.c: Oops time; closing FILE * object twice is BAD THING to do in some cases. Must not do so. Early closing of the underlying read socket is ok. * proto/smtpserver.conf.in: In smtp-auth related PARAMs, do emphasise their security implications! 2001-05-30 Matti Aarnio * scheduler/scheduler.c, scheduler/msgerror.c, scheduler/qprint.c, scheduler/threads.c, scheduler/transport.c, scheduler/update.c: I suspect that in certain cases the scheduler should do "resync" on message file instead of just logging a complaint into stderr like it now does at the end of findvertex(). Adding some debugging data so that I can verify the thing. 2001-05-25 Matti Aarnio * README.UPGRADING: Semi-forgotten note about the results of 2001-05-15's libta/dianostic.c + scheduler/update.c + smtp TA/scheduler interaction protocol detail. You can have NEWER scheduler, than smtp-ta, because this is backwards channel detail. * lib/cfgets.c: Darryl Miles found strange looking problem with certain kind of "malformed" input for scheduler configuration file. Turns out that when this function encounters EOF, and it isn't at the beginning of the line, the line buffer never gets zero-terminated causing possibly strange side-effects when thus revealed previous content of the buffer may reappear. Additional coding to always zero-terminate the allocated buffer. The code is being used both as classical fgets(), but also by returning exact number of characters collected. I am doubtfull that the previous state was in any way capable to produce system security threat. 2001-05-22 Matti Aarnio * proto/cf/fqdnaliases.cf, proto/cf/process.cf, proto/cf/aliases.cf: Some experiments on doing message-wide 'didexpand' analysis to see if complicated duplicate recipient removal code works. 2001-05-21 Matti Aarnio * smtpserver/smtpetrn.c: Tell to the remote scheduler [source.ip.address]@mynodename when issueing ETRN thru MQ2 mechanism. Helps to hunt down cases where lots of ETRN requests are appearing into the scheduler for domains that we don't serve, and/or other such strange things... * configure.in, libc/gai_strerror.c: Problems at compling "replacement-ipv6-libc" with RH 7.1 libc. 2001-05-16 Matti Aarnio * ssl/Makefile.in, ssl/ssl.c, ssl/ssl:sst.[ch], ssl/ssl_sst.[ch]: Changed references from ssl:sst.[ch] files to ssl_sst.[ch] 2001-05-15 Matti Aarnio * include/ta.h, scheduler/update.c, transports/smtp/smtp.c, transports/libta/diagnostic.c, transports/smtp/getmxrr.c: Define and use EX_DEFERALL exit code, which introduces a new, somewhat differently behaving handling of temporary kind of failures -- intended for smtp transport-agent for DNS lookup failures of some domains, and connection failures. To large extent this behaves as if "retryat +60" were defined, but this will also mark ALL subsequent messages in the thread (via "nextitem" chain) to 1) have been attempted, and 2) with the same message as we initially have. (It does not, however, expire any messages - at least directly.) * scheduler/mailq.c: The 'mailq -v' did show the ta-work-specification-file size, not message size in 'bytes %ld' printout.. :-( Also: Have options: "-c channel -h host" which TOGETHER will let querying of some particularly interesting destination channel/host pair among all that the system may have active. Requires MAILQ-V2 mode, and '-v' option use (not '-Q'!). 2001-05-10 Matti Aarnio * include/libc.h: Copy the 'GETNAMEINFOFLAGTYPE' test code into this file too from libc/getnameinfo.c. Bloody Solaris.. 2001-05-08 Matti Aarnio * Makefile.in: Version 2.99.55-patch1 * transports/smtp/smtp.h, transports/smtp/smtp.c: Reworking (simplifying) DATA phase processing, and quick-retry handling of socket write/read failure induced closes. There is limit of three attempts on any given host for the delivery before giving up entirely. (Passing the failure to the scheduler..) 2001-05-04 Matti Aarnio * proto/cf/p-routes.cf: Wrong function called to pick the flag data.. * transports/smtp/smtp.c: Fixup into pipelined error diagnostics accumulation methodology. * proto/newdbprocessor.in: PERL 5.6.0+ specific syntax altered back to earlier syntax, which even the lattest perl does support. * configure.in: --with-zconfig-noload option to disable preloading of possibly existing ZCONFIG file contents into configuration variables. 2001-04-04 Matti Aarnio * transports/smtp/getmxrr.c: If destination has one MX, and lookup for its A yields EAI_AGAIN, remember to report SS->mxcount = 0 ... 2001-04-03 Matti Aarnio * proto/forms/delivery: Changed the URL reference to a "decode key" at http://www.zmailer.org/delivery-report-decoding.html * acconfig.h, configure.in: When supporting local _getaddrinfo_() function, have a HAVE_ variable for it too! * include/mail.h.in, router/rfc822.c: Rearranged the _CF_*** defines into the order at which the router uses them -- while editing documentation, and needing coherent view at the things. * scheduler/threads.c: When running parallel TAs at same thread, remember to traverse pointer chain when showing process list.. (verbose debugging to stderr..) * libc/getaddrinfo.c: Present at the debug vlog file the g->gaih() return code in octal (it is a bitset, after all) * smtpserver/mxverify.c, transports/smtp/getmxrr.c: Use the HAVE__GETADDRINFO_ flag for local debug version.. Also done changes for retrieving AF_INET and AF_INET6 addresses separately, as there seems to be some problems when trying to use ai_family == 0 ("any goes") form. 2001-03-08 Matti Aarnio * include/mailer.h, router/rfc822.c, router/libdb/header.c, transports/errormail/errormail.c, scheduler/msgerror.c, proto/cf/p-routes.cf, router/functions.c: New envelope header: "errormsg" telling that a message is internally generated error message; router script has new function: $(iserrmessage) yielding this flag. * include/libc.h, include/libsh.h, include/sysprotos.h, libc/getopt.c, libsh/builtins.c, libsh/interpret.c, libsh/main.c, libsh/zmsh.c, router/router.c, router/db.c, router/functions.c: Changed all Router/ZMSH internal instances of getopt() to be zgetopt() with additional capability -- optionset can now contain doublecolon as an option, IF the ':' is the FIRST character at the option specifier string... * proto/newdbprocessor.in, proto/db/dbases.conf, router/db.c, proto/cf/p-routes.cf, proto/cf/rrouter.cf, proto/db/routes, router/libdb/search.h: Implemented DBlookup routine to be able to pick some options at runtime, '-:' (with string arg) supplies default lookup key in case the normal iteration fails. Cases where the "zmailer newdb" should generate the lookup wrapper function with user parameter passing. Also support 'incore' database by compiling from source file to "db add" statements in the ``relationname.zmsh'' script. * router/rfc822.c, router/functions.c, router/prototypes.h: Streamlined the recognition of the message being coming from an Error Channel ( not same as "errmessage" !) * proto/cf/aliases.cf: The FOO+TAG needs to cut off at the first '+' char, not by the last (or whatever the REGEXP code happens to use) * libsh/regex.c libsh/regex.h: Updated this pair to a very recent GLIBC contained version. * transports/smtp/smtp.c: At response wait timeout, do have a GOAL TIME of when to timeout, not returning to it over and over again.. * configure.in, aclocal.m4, config.h.in: Using experimental autoconf 2.49d mandated couple of changes... 2001-03-02 Matti Aarnio * README.UPGRADING: Added a note about the ALTERED default locations of zmailer.conf and mail.conf files. * man/smtpserver.8, proto/smtpserver.conf.in: Fix up the tcp-wrapper related documentation, and the default stuff in the smtpserver.conf template. * scheduler/threads.c: Print the gauges of message counts in the queue with SIGNED values, thus making underflows (negatives) more obvious. 2001-03-01 Matti Aarnio * scheduler/mailq.c: Embarassing uninitialized array elements used for checking if loop ends - Tomasz Bojakowski reported errors. 2001-02-27 Matti Aarnio * transports/smtp/smtp.c: At smtp_sync() when running in PIPELINING, if the MAIL FROM has produced 400 or 500 series response, do fault the RCPT TO responses in every case. (With EX_UNAVAILABLE for 500-series, not EX_TEMPFAIL.) * smtpserver/smtpserver.h, smtpserver/smtpserver.c: Track Z_read() return errno code more properly to the 'Session closed w/o QUIT' log report. Case where this was needed was actually some firewall product which f**d OUTBOUND SMTP session of raporter... 2001-02-26 Matti Aarnio * utils/listexpand.c, transports/smtp/smtp.c, include/ta.h, transports/libta/ctlopen.c: Implemented "EZMLM-alike" listexpander which does the envelope MAIL FROM address content rewrite at tne outgoing SMTP session. This is done with 'DSN' dataentry EZMLM= which is presumed to have valid stuff to be appended to the localpart of the address envelope: '+recipient=40address.dom', for example. When compared with more classical approach, this does message routing in system-friendly way, and only changes the output stage slightly. (At classical case the message is expanded for each recipient, say 3000 copies of 60 kB each.. This mechanism is way faster.) 2001-02-24 Matti Aarnio * smtpserver/smtpcmds.c: Make RCPT TO synchronous router reporting to be a copy of MAIL FROM one -- report things correctly, that is. 2001-02-23 Matti Aarnio * router/rfc822.c, router/rtsyslog.c: Added the time to process the message to the router syslog report. Both 'delay' and 'xdelay' just like at the TA's. 'delay' is time from message injection to routing completion, 'xdelay' is the time spent at the routing. * utils/listexpand.c, utils/Makefile.in: Added '+ezlml' option, changed the low limit at the -bundlesize -- and set its default to 6000! Also re-learned that 'skip821address()' can only be used with *valid* addresses, rfc821_path2() is for possibly invalid input scanning! 2001-02-21 Matti Aarnio * Makefile.in: Version 2.99.55 (the real release!) * router/functions.c: At $(listaddresses ..) and $(listexpand ..) functions, removed the '"%1"@..' type substitutions with '-l' option. See the comment about '-%' at 'relation' function below. 2001-02-20 Matti Aarnio * scheduler/agenda.c, scheduler/scheduler.c, scheduler/mq2.c, smtpserver/smtpetrn.c: Report the ETRN information with contactee's IP address, and timestamp when writing it into the scheduler's log. * proto/newdbprocessor.in, proto/db/dbases.conf, man/router.8, proto/cf/fqdnalias.cf, router/db.c, router/functions.c: Add '-%' option to the 'relation' routine to control, when looked up data contained '%0' thru '%9' may get substituted. * proto/Makefile.in: Remove at 'distclean' files which are autoconf generated. 2001-02-19 Matti Aarnio * transports/smtp/getmxrr.c: Format printing of '%s' with NULL argument -> NULL- referral -> ... * scheduler/mailq.c: When doing some slow 'mailq -vv', and using MAILQv2, do query all data out of the scheduler before poking into the files (which may be a lot). Limit the time of inter- action with the scheduler to the minumum. 2001-02-16 Matti Aarnio * scheduler/readconfig.c: One-off bug in ${ZENV} substitution code. * configure.in: Derive ZCONFIG (formely defaulted to /etc/zmailer.conf) from --prefix ! The same way as --prefix derives $MAILSHARE. Officially zmailer.conf's new location *is* $MAILSHARE. 2001-02-15 Matti Aarnio * proto/cf/fqdnaliases.cf: Moved the local-part (user) extraction up a bit so that it can be given to the listaddresses as expansion parameter. * router/functions.c: At $(listaddresses ..) and $(listexpand ..) functions, support '"%1"@..' type substitutions with '-l' option. I am now considering this *wrong* approach, and am doing more sendmailish approach to %N substitutions. * router/functions.c, router/db.c, router/libsh/search.h: Support up to 8-9 additional substitution options at the database lookups. At the moment is lacking the actual %N substitution code! WORK IN PROGRESS! * scheduler/mailq.c: a bit of comments & one missing newline on error printouts. 2001-02-14 Matti Aarnio * router/router.c: Cut down logging/log printing a bit -- duplicated data actually. 2001-02-13 Matti Aarnio * proto/scheduler.conf.in, scheduler/readconfig.c: Allow ${ZENV} substitutions in PARAM lines. ("PARAMauthfile" and "PARAMmailqsock", to be exact.) * smtpserver/policytest.c: Do Test-Rcpt-DNS-RBL also based on FULL userid lookup. Allows some users to want the lookup, others not. * proto/cf/SMTP.cf.in, proto/cf/SMTP+UUCP.cf.in: Search for the 'mail.conf' from $MAILSHARE/mail.conf, and not from /etc/. 2001-02-08 Matti Aarnio * router/functions.c: At $(filepriv ...) function, when the call supplies file userid, do presume valid REGULAR file. This is true for several databases which produce privilege controlling results. * proto/cf/canon.cf, proto/cf/rrouter.cf: More issues at DECNET-IV type address analysis. Now we support a bit of being DECNET-IV gateway, although we propably need to rewrite headers major way to be able to route to and from DECNET-IV domains. * smtpserver/Makefile.in: Do refer to wantconn.o file in the OBJS macro, it needs to be compiled in case it is wanted to be used... 2001-02-01 Matti Aarnio * man/mailbox.8, transports/mailbox/mailbox.c: Add two more environment variables to pre-existing ones at the mailbox driven subprocesses: - MSGSPOOLID= - MESSAGEID= These were created when I had a need, of course.. Pipe driven application should be able to syslog(3) meaningfully what was the message disposition. (Majordomo in my case ;) ) 2001-01-31 Matti Aarnio * configure.in, acconfig.h: - --with-routeruser-is-abnormal option - detect possible pre-existing ROUTERUSER_IN_ABNORMAL_UNIX value in zmailer.conf file - in case of wanting to use tcp-wrappers, do analyze if it needs adding -lnsl library in addition to -lwrap -- some unfortunate systems (RedHat 6.2-7.0) compile the bloody wrapper library with one NIS call, and the ZMailer does not use NIS libraries for anything (except if router wants..) * smtpserver/smtpserver.h, smtpserver/smtpserver.c, smtpserver/smtpdata.c: Arm the alarm(2) just before call to read(), and re-arm it right after return from it. This *may* cut down the number of alarm(2) calls a bit. * smtpserver/smtpserver.h, smtpserver/smtpserver.c, smtpserver/wantconn.c, scheduler/transport.c, scheduler/wantconn.c: #ifdef USE_TCPWRAPPER in addition to #ifdef HAVE_TCPD_H 2001-01-30 Matti Aarnio * libresolv/gethnamaddr.c: Silence the code in several occasions. Getting SIG data is *not* a problem, it *does* happen. * smtpserver/smtpserver.c: Clean "RSET" response from "250 2.0.0 2.0.0 Reset processed.." to one with: "250 2.0.0 Reset processed.." 2001-01-25 Matti Aarnio * transports/smtp/smtp.c: Change all instances of getaddrinfo() parameters of service name "smtp" to port-number "0". Give also some detail on what the getaddrinfo() diagnostics gave. * router/functions.c: - In $(listaddresses ..) and $(listexpand ..) functions, explicitely support receiving lines with '#'- starting them (comments, you see). Also when found no addresses, explicitely do mention A) what EUID the expansion was running, and B) that user should check the file/directory protections of the original file. - Also, the $(filepriv ...) function will from now on allow wider opening privileges in some cases -- if residence directory does not have X-bits in group and/or world, then related bits in the file are not checked -- i.e. if the directory is not world- accessible, it is irrelevant if file has w-bit in world set. * proto/cf/SMTP+UUCP.cf.in, proto/cf/SMTP.cf.in: Couple of fixes, and code rearrangements into the sample boiler plate. * transports/smtp/getmxrr.c: Play a bit more with rmsgappend() calls here. Specifically report if getaddrinfo() calls report some diagnostics! 2001-01-24 Matti Aarnio * lib/selfaddrs.c, transports/smtp/smtp.c: If resolver supports option flag RES_USE_INET6, but system doesn't support IPv6 socket, clear that flag. * libresolv/gethnamaddr.c, libresolv/netdb.h: When using --with-bundled-libresolv silence things a bit as one particular #define may be set before calling this. Also, do explicitely at first lookup /etc/hosts file for host, only then do resolver lookups. 2001-01-22 Matti Aarnio * include/libz.h, include/mailer.h, compat/sendmail/sendmail.c, lib/nobody.c, lib/pwdgrp.c, lib/trusted.c, router/functions.c, router/router.c, scheduler/readconfig.c, smtpserver/zpwmatch.c, transports/hold/hold.c, transports/mailbox/fmtmbox.c, transports/mailbox/mailbox.c, transports/mailbox/mboxpath.c, transports/smtp/smtp.c, transports/smtp/smtp.h, utils/vacation/Makefile.in, utils/vacation/vacation.c: Use zgetpwnam() everywhere where there used to be getpwnam(), supply also our very own struct Zpasswd for the data referral. Library lib/pwdgrp.c is the location of the wrapper function, which intention is to clarify error handling of various systems -- getpwnam() returning NULL may mean many things, and interpreting errno is -- challenging. * libsh/builtin.c, libsh/interpret.c, libsh/variables.c, proto/cf/SMTP+UUCP.cf.in, proto/cf/SMTP.cf.in, proto/cf/aliases-new.cf, proto/cf/aliases.cf, proto/cf/consist.cf, proto/cf/fuzzy-aliases.cf, proto/cf/process.cf, proto/cf/rrouter.cf, proto/cf/server.cf, proto/cf/standard.cf, README.UPGRADING: Change internal variable expantion to never do IFS tokenisation at expanded strings. Supply $(ifssplit ...) function in case it is needed for something, and just token-lists don't make it. Write some definite words on README.UPGRADING ! 2001-01-22 Eugene Crosser * transports/libta/mimeheaders.c: Sometimes quoted string (to de stripped) is really not, and we must be carefull to do it only with a string of minimum length of 3. 2001-01-22 Matti Aarnio * transports/smtp/smtp.c: When the reply collection goes to timeout ( smtp_sync() ), CLOSE the smtp-connection. Also do rmsgappend() a report, that reply collection went into timeout.. 2001-01-16 Matti Aarnio * transports/hold/hold.c: Do same processing here as at router/functions.c. * router/functions.c: Condition "no such user" can be signaled with getpwnam() yielding NULL and errno == 0. Not only errno == ENOENT. * lib/dbpwdgrp.c, lib/nobody.c, lib/pwdgrp.c, lib/trusted.c, router/daemonsub.c, router/functions.c, router/router.c, transports/mailbox/mailbox.c: Essentially in front of *every* getpwnam(), do place explicite clear of errno variable. Apparently at Solaris 2.6 it isn't cleared/set quite always when doing error paths... 2001-01-15 Matti Aarnio * transports/mailbox/mailbox.c: Typo -- mmfd_mode which should be mmdf_mode ... * transports/libta/diagnostic.c: #include -- apparently cc at Solaris 2.6 didn't like it without :-/ * smtpserver/smtpserver.h, smtpserver/cfgread.c: More Sol26/SunSoft CC things. * lib/rfc822scan.c: Recognize "HOST::USER"@decnet-IV.gw syntax, and don't ever unquote it. We don't support being the DECNET-IV gateway ourselves. * proto/zmailer.sh.in: If dbases.conf is not in use, and we execute 'zmailer newdb', do 'newdb' the $MAILSHARE/db/mboxmap also. 2000-12-31 Matti Aarnio * transports/mailbox/mailbox.c: At normal success path, do proper fsync(2) return status analysis - if the filesystem where the write goes to is ever truly asynchronous, the only way to know if it runs out of diskspace, or whatever is after the return of fsync(2). * libc/mail.c, libc/sfmail.c, scheduler/msgerror.c, router/rfc822.c, transports/libta/diagnostic.c: More of carefull fsync(2) calls. * transports/mailbox/mailbox.c, smtpserver/smtpcmds.c, transports/libta/diagnostic.c: Also carefull wrappers for ftruncate(2) calls. 2000-12-30 Matti Aarnio * transports/libta/mimeheaders.c: Survive bad control characters in input. * proto/cf/aliases.cf: Sender address rewrite function to yield full "$user$plustail$domain" - incoming quad arrives without domain part. * router/rfc822.c When in 'trace sequencer', do debug print call parameters to $(crossbar ...) before the call. * scheduler/readconfig.c, scheduler/transport.c: setpriority() syscall autoconfig things 2000-12-14 Matti Aarnio * smtpserver/smtpdata.c: "BDAT 0 LAST" is valid per explicite example at RFC 3030. It is also valid per syntax, but not quite so by mere engineering thought... 2000-12-04 Matti Aarnio * transports/libta/mimeheaders.c: ElectricFence is your friend! Writing over the end of the allocated array by on element :-/ 2000-12-02 Matti Aarnio * transports/smtp/smtp.c: Don't soften reception of RCPT TO if MAIL FROM gave 500 series code -- there was some reason for this... * transports/libta/tasyslog.c: Report correctly if we have detected MULTIPART(QP). * lib/selfaddrs.c: Something odd at Solaris -- none of supported functions got called ??? -- realloc() bug :-( * All over the place; Changeing getzenv()'s return value to 'const char *' caused all kinds of nuisance warnings to sprout... * All over the place; (Makefile.in): More "make -j N" work. 2000-12-01 Matti Aarnio * lib/linebuffer.c, lib/selfaddrs.c, router/daemonsub.c, scheduler/scheduler.c: Changed some code doing if (!ptr) ptr = malloc(..); else ptr = realloc(ptr,..); into ptr = realloc(ptr,...) after wanting to simplyfy one part of code and reading man-page.. * Makefile.in, libident/Makefile.in, libs/Makefile.in, libsh/Makefile.in, router/Makefile.in, smtpserver/Makefile.in, router/libdb/Makefile.in, sfio/Makefile.in, sfio/src/lib/sfio/makefile.in, sfio/src/lib/sfio/Stdio_b/Makefile.in, sfio/src/lib/sfio/Stdio_b/sfstdhdr.sh: Tried to "make -j 20" and learned things about parallellism.. Things really aren't correct yet, and at least sfio may end up being compiled some N-4 times in parallel... * transports/smtp/smtp.c: Modify some loops - removed unnecessary state free()er code from within the smtp_sync() -- with a theory that the pipecmds[] array goes boinkers somewhen... 2000-11-30 Matti Aarnio * transports/libta/mimeheaders.c, transports/mailbox/mailbox.c, transports/smtp/smtp.c, include/ta.h: Content-Type Parser had reversed pointers -- oops Changed parsers (multiple) parameters from (char **) to (char *) * transports/libta/mimeheaders.c: RFC822 ATOM skipping didn't.. (single-char atoms) * transports/libta/swriteheaders.c, transports/libta/fwriteheaders.c: Reusing variable caused wrong content at the pointer, and rare SIGSEGVs lurking ... 2000-11-24 Matti Aarnio * transports/libta/mimeheaders.c: Simplify! More of the simplification of code. * transports/libta/ctlopen.c: add a bit to an allocation -- not that it really should matter, but ... * transports/smtp/smtp.c: sprinkled smtp_flush() around places while hunting one nasty memory corruption. * transports/libta/swriteheaders.c, transports/libta/fwriteheaders.c: Some leftover counter adjustments in places where it now harmed... * transports/smtp/smtp.c: Debugging most peculiar problems - propably at swriteheader.c Showing at the verbose log the WRITTEN headers. * transports/smtp/smtp.c: A bit different handling on pipelined response collection. If MAIL FROM yields an error, report it, never mind what socket read did later.. * transports/libta/mimeheaders.c: Simplify! Don't try to be smart when playing with malloced buffers -- it doesn't pay, quite contrary... Found a number of cases where a buffer got reallocated, but a pointer into it wasn't adjusted into the new area. * transports/libta/swriteheaders.c, transports/libta/fwriteheaders.c: Changed the code from "scan the input to determine future space needs" to "(re)expand as more space is needed" to make sure there is no discrepancy in between the two. 2000-11-23 Matti Aarnio * transports/smtp/smtp.c: Few hours of going over the smtp_sync()'s smtp_nbread() call, and its surroundings. Inverted its logic, and finally got simple, clean, understandable code. ("I smell Wumpus!") * scheduler/mailq.c: Wondered that "mailq -vv" didn't show diagnostic messages along other data. Now it does. Plus other small tuneups of the result. * transports/smtp/smtp.c: A bit more of care on referencing 'logfile' variable. Not using uninitialized one. * transports/mailbox/mailbox.c: A trivialish return-receipt-to processing fixup at the mailbox so that the user-supplied string is unfolded into single line with spaces in places of folds (newlines). (Proper fix is FULL RFC 822 scanner to pick the recipient address, but perhaps not...) * transports/smtp/smtp.c: Blocking read waiting was done with busy-loop, and if the connection *stays* in ESTABLISHED state, but never timeouts --> busy spin goes on, and on... * scheduler/msgerror.c: Don't try to be clever with cfp->mid field reuse at the slurp()ed instance -- use incoming 'cfpi->mid'. 2000-11-21 Matti Aarnio * transports/smtp/smtp.c: Meaningless variable name changes (and some revoked changes which got committed into CVS - and now removed.) * transports/libta/fwriteheaders.c, transports/libta/swriteheaders.c: One subtle bug in BDAT buffering mode left in line-ending newline handling --> surprising crashes seen. * transports/libta/mimeheaders.c: Reserve a bit more space for coalesced mime-headers (of "totlen" type). * transports/smtp/appendlet.c: Outch! Messages without Content-Type: header did SIGSEGV! * lib/selfaddrs.c: there is no field called ifr_ifindex, there is: ifr_index -- returned to previous: memset(ptr, 0, size) -- because Linux glibc 2.1+ has ifr_ifindex, solaris has ifr_index (barf...) 2000-11-19 Matti Aarnio * man/smtpserver.8, smtpserver/policytest.c: Lowercasify the lookup key for alias item lookups, and fill in the doc/guides/smtp-policy header into the manual page of the smtpserver. 2000-11-17 Matti Aarnio * scheduler/threads.c, scheduler/mailq.c: Some experiments of making the mailq-v2 mode mailq to show same things as the old version was able with 'verbose' mode turned on. * **/Makefile.in: Changes on what we refer at LIBDEB= definitions. * transports/libta/mimeheaders.c: - Missing initialization of one buffer -> result is memory scrible and utter chaos.. - Wondering of X-Warning: writer does something fishy, as the Received: header generated by the site where that writer is used does not appear in the message.. * proto/scheduler.conf.in, proto/db/routes: Define channel smtp-log which (obviously) logs the SMTP protocol into log-file. See "-l" option at smtp(8) man-page. 2000-11-16 Matti Aarnio * configure.in: Some reworking around NDBM tests -- we *MAY* have problems when is actually from BSD DB 1.85. Specifically in OpenBSD 2.7, several other BSDs as well. * transports/smtp/getmxrr.c, scheduler/mailq.c, libc/myhostname.c: BSD getaddrinfo() call returns ai_canonname member as NULL in very common cases! Uargh :-( * lib/Makefile.in: At 'make selfaddrs-test', do use autoconfigured @LIBRESOLV@ library reference. 2000-11-14 Matti Aarnio * lib/selfaddrs.c, lib/zgetifaddress.c: Fix the support for FreeBSD 4.2*/OpenBSD 2.7* featured getifaddrs() function. * Makefile.in: Mark version "2.99.55-pre1cvs" (which means nothing much) * include/ta.h, transports/libta/fwriteheaders.c, transports/libta/swriteheaders.c, transports/libta/markoff.c, transports/libta/mime2headers.c, transports/libta/stringlib.c, transports/libta/mimeheaders.c, transports/smtp/appendlet.c, transports/smtp/smtp.c, transports/smtp/smtp.h: Initial steps towards doing proper format downgrade during (SMTP) transport. - Modified header handling to process folded lines as single string, not as a sequence of strings. - Wrote proper RFC822/MIME token scanner into mimeheaders.c which is able to correctly skip over places where (folded)LWSP is allowed, including possible (comments) in them. 2000-11-12 Matti Aarnio * CVS tagged this as Version zmailer-2-99-54-patch1 Made also CVS based diff against 2.99.54. * lib/selfaddrs.c: Learned something about the SIOCGIFFLAGS/SIOCGLIFFLAGS ioctl parameters, and what is necessary for initializing its call. * transports/smtp/smtp.c: Some oddball moments don't fail properly after response timeout during certain critical periods -- mainly of DATA/RSET. Doing smtpclose() where we used to RSET right after DATA has been sent, and response to it has been waited for way too long.. 2000-11-02 Matti Aarnio * smtpserver/smtpserver.h, smtpserver/smtpserver.c: Turn the listening sockets into NON-BLOCKING mode, and once connection is accept(2)ed, turn the communication socket into blocking mode. * transports/smtp/getmxrr.c: Randomish bounces seen at vger.kernel.org for domains which have non-functional DNS servers (likely due to network failure, or some such.) Symptom seems to indicate NXDOMAIN DNS query reply status, but the darn thing happens *rarely*... ... I am not sure this change is a smart thing to do, ... and later on I removed this change. It looks like getting NXDOMAIN is trouble at whatever system happens to be the DNS server, and we are better off when we don't trouble ourselves too much with circumventations at the resolver code for it. 2000-10-31 Matti Aarnio * transports/smtp/smtp.c: Move the notarystatsave() function call up so that it will always be called for all collected replies. Formelly there were cases where it didn't get called, which caused somewhat weird looking results when the TA program switched to alternate destination domain. * libc/getaddrinfo.c: Disable entirely the AF_LOCAL support. The code in question used tmpnam() which created some complaints from the recent linker/libc at RH 6.2/7.0 hybride system. * configure.in, proto/smtpserver.conf(.in): Turns out that "make install" when working in "adjacent directory" setup doesn't pick correctly the smtpserver.conf for the installation. Returning to generated smtpserver.conf file, although strictly speaking that is not needed. * proto/db/dbases.conf, utils/policy-builder.sh.in, proto/newdbprocessor.in: Spotted a few mistakes at the perl scripts of newdbprocessor. With a hindsight -- me stupid... Spotted also missing "-d pathalias" at "routesdb" specification. And some awk cleanups at policy-builder as well, of course. * packaging/solaris/README: Some additional comments on usage of this tool subset. 2000-10-24 Matti Aarnio * proto/cf/rrouter.cf, proto/cf/canon.cf: Support magic trick: @[CHANNELNAME]:user@domain in such a manner as to produce (((CHANNELNAME domain user@domain ATTRIBUTES))) routing result 2000-10-23 Matti Aarnio * transports/smtp/smtp.c, utils/mxerify/mxverify-cgi.c: For IPv4 addresses, verify that our connect(2) destination is of valid IPv4 network (first octect is in range of 1..223), at mxverify exclude loopback. * transports/smtp/Makefile.in: Add into "make install" installation of $MAILBIN/getmxrr-test so that it will be available at all installations... * utils/mxverify/mxverify-cgi.c: Small text changes. * transports/smtp/smtp.c: Noticed a few places where we need to reset the "SS->rcptstates" variable. 2000-10-20 Matti Aarnio * proto/Makefile.in, packaging/Makefile.in: 'make install' will install also 'doc/guides/*' files into $MAILSHARE/guides/ -- there seems to be call for those files from some users... In other environs, like Linux distributions, there could be some sort of /usr/share/doc/zmailer-NNN/guides/ for them. * doc/guides/smtp-policy, utils/policy-builder.sh.in: Defined ".manual" versions for ".mx", and ".relay" of smtp-policy driver files. This enables more wide-spread automated data fill-in, plus local ".manual overrides". 2000-10-17 Matti Aarnio * transports/libta/ctlopen.c: Return NULL, not 'void' in certain silence case. * transports/libta/diagnostic.c: Try not to do so verbose diagnostic report over syslog as via TA->scheduler channel. * smtpserver/smtpdata.c: Parse possible "contentpolicy" analysis report, and present it (in several cases) to the user. * smtpserver/smtphelp.c: Small fixes for help presentation, and parameter recognition. 2000-10-16 Matti Aarnio * smtpserver/smtpdata.c, smtpserver/smtphelp.c, smtpserver/smtptls.c, smtpserver/smtpserver.c: Rethought of things, killed the Z_printf() abomination. Also fixed inadverent NULL referral which Eugene Crosser found as well. (It blocked my incoming email :-( ) * man/smtpserver.8, smtpserver/contentfilter.c, smtpserver/smtpdata.c: Improved documentation of the CONTENTFILTER INTERFACE ( = wrote it ), and modified a bit of the interface itself. * configure.in, proto/smtpserver.conf, smtpserver/cfgread.c, man/smtpserver.8, proto/Makefile.in, Allow $ZENV expansions in smtpserver's configuration. Thus allow simpler moving of the thing to other location without needing to edit anything but zmailer.conf file.. Also removed smtpserver.conf from autoconfigured protofiles. * scheduler/readconfig.c, smtpserver/cfgread.c, transports/sm/sm.c, transports/mailbox/mailbox.c, transports/smtp/smtptls.c: Changed all instances of macro-in-macro style CPP things of 'SKIPWHILE(testfunction,varname)' to be appropriate 'SKIPSPACE(varname)' and 'SKIPTEXT(varname)'. (RedHat 7.0 gcc version does not work correctly with those..) * smtpserver/policytest.c, smtpserver/rfc821scn.c, smtpserver/smtpcmds.c, smtpserver/smtpdata.c, smtpserver/smtpetrn.c, smtpserver/smtpserver.c, smtpserver/smtpserver.h, smtpserver/smtphelp.c, smtpserver/mxverify.c: Aimed for some coherence at debug mode printouts/flushes in some cases. Mainly the problem is that at the policytest.c, and at the mxverify.c the debug output goes to stdout via printf(), while elsewere it goes via type() to SFIO attached to the connection. TO BE ACCURATE, AS THINGS ARE NOW IS NOT SUITABLE FOR DEBUGGING E.G. SSL WRAPPED SESSIONS! * smtpserver/smtpserver.h, smtpserver/smtpserver.c, smtpserver/cfgread.c, smtpserver/smtptls.c: - Have new runtime flag (-t) telling that this is a session running immediately under TLS (port tcp/465, "ssmtp"), which feature predates STARTTLS facility. - Have parallel listen socket for two different services; namely plain 'smtp', and 'ssl/smtp') * scheduler/mq2auth.c, scheduler/msgerror.c, include/libz.h, scheduler/readconfig.c, lib/cfgets.c, smtpserver/cfgread.c, transports/errormail/errormail.c, transports/mailbox/mailbox.c, transports/sm/sm.c, transports/smtp/appendlet.c: Renamed that function to be csfgets() -- after I myself got confused about its calling convention... (That is: FILE * or Sfio_t * thing..) 2000-10-13 Matti Aarnio * sfio/src/lib/sfio/makefile.in, sfio/src/lib/sfio/Stdio_b/Makefile.in: Noticed an omission in cross-directory compilations. Sometimes 'iffe' would fail when directory 'FEATURES' didn't exist at the running position. Oops. 2000-10-12 Matti Aarnio * utils/makedb/dbdump.c: New option: -policydump which decodes the policy binary set into a text. * smtpserver/smtpserver.h, smtpserver/smtpserver.c, man/smtpserver.8, smtpserver/cfgread.c, smtpserver/smtphelp.c, smtpserver/smtpcmds.c, smtpserver/smtpdata.c: Implemented experimental LMTP (RFC 2033) support, this is *not* for real LMTP use, merely for debugging LMTP clients. * transports/smtp/smtp.c, transports/smtp/smtp.h, proto/scheduler.conf.in: Completed the LMTP (-M) mode, tested it quite a bit with and without remote speaking PIPELINING. Can support even remotes announcing STARTTLS capability in the LMTP port. Removed duplicate remote response analysis code from synchronous (non pipelined) code path, now using only the smtp_sync() to collect replies. Supplied sample "smtp-lmtp/*" configuration for use. * proto/zmailer.sh.in: Do cd to $POSTOFFICE when starting smtpserver. Dropping cores to system root is not polite.. * man/sendmail.8, man/zmsh.1: Changed to * man/smtp.8: Mention RFC 2033 and LMTP at the RFC list in the end. * man/scheduler.8, scheduler/mq2auth.c, proto/scheduler.auth.in: Allow multiple instances of same userid by scanning the address ACLs at each UID. Thus nobody@localhost can get different level of access from nobody@elsewere. 2000-10-10 Ambrose Li * utils/vacation/vacation.c: Removed some duplicate code, and detect autoreplies whose source address is: <> 2000-10-10 Matti Aarnio * lib/Makefile.in: "make install" did put zmailer.h into $prefix/$prefix/include/, which is "somewhat" wrong... * man/smtp.8, transports/smtp/smtp.c, transports/smtp/smtp.h: Implemented LMTP (RCF 2033) at the smtp TA program. To enable LMTP, use option '-M'. * scheduler/mq2.c: When reporting "unknown" command, do tell things differently: "Unknown command, or refused by access control;" ... * smtpserver/mxverify.c: Fix the IPv6 nybble-reverse code to do proper in6 type reverser, although an alternate (binary labels) is defined these days... * transports/libta/ctlopen.c: See if the message to be opened is an ETRN request -- if yes, just return with failure... (manual-expirer triggers this.) * transports/libta/mimeheaders.c: While parsing for Content-Type: parameters, allow LWSP in same lines. Full specification really calls for RFC 822 tokens along with CWSP and commend handling. * transports/mailbox/mailbox.c: Separates "From MAILER-DAEMON " from "Return-Path: <>". Produces "X-Envelope-To: <...>" header. * ChangeLog, INSTALL, Overview, README.TCP-WRAPPER, README.UPGRADING, compat/sendmail/sendmail.c, doc/guides/configure, doc/guides/smtp-etrn, include/zmsignal.h, lib/hashtest.c, man/smtpserver.8, proto/smtpserver.conf.in, proto/cf/aliases-new.cf, proto/cf/aliases.cf, proto/cf/rrouter.cf, proto/cf/standard.cf, router/prototypes.h, router/rfc822walk.c, scheduler/msgerror.c, scheduler/scheduler.c, smtpserver/rfc821scn.c, transports/sm/sm.c, smtpserver/smtprouter.c, transports/libta/diagnostic.c, transports/mailbox/mailbox.c, transports/smtp/smtp.c, utils/smtpserver-log-parser.pl.in, utils/pop-proxy/pop-proxy.c: A great speller-bee of "propab*" -> "probab*". 2000-09-26 Matti Aarnio * man/sm.8, transports/sm/sm.c: Updated (with a radical hand) the manual page to list what sendmail 8.11 does with flag characters, and what the ``sm'' does with same characters (and some more). 2000-09-25 Matti Aarnio * transports/hold/hold.c: This was left out from the commit 4 days ago. * libresolv/sys/bitypes.h, sfio/src/lib/sfio/sfhdr.h, ssl/ssl.c: Experimental changes to support compilation at AIX 4.1. * proto/db/dbases.conf: The 'thishost' relations used to carry '-d pathalias' driver option on original systems, now it was missing.. 2000-09-22 Matti Aarnio * proto/Makefile.in: An adjacent directory install failed when system does not have suitable binary 'install' program, and 'install.sh' is needed. Rearranged things. 2000-09-21 Matti Aarnio * scheduler/mq2auth.c, scheduler/scheduler.c: File leakage at authuser(), which caused opendir() to fail, and its resulting NULL was used unprotected -> SEGV. (Ran out of free file descriptors.) * router/functions.c, transports/hold/hold.c, proto/cf/aliases.cf: "hold" handling for the $(homedirectory ...) function. 2000-09-20 Matti Aarnio * proto/cf/rrouter.cf, proto/db/routes: Generalizing on various smtp* mechanisms. * proto/scheduler.conf.in: Channel prototypes for smtp77/*, smtp77x/*, and others. Some reordering, and log option rearrangements. * transports/smtp/smtp.c, man/smtp.8: ESMTP disabler feature. ( -77 option ) * packaging/solaris/Makefile.in, packaging/solaris/README, More work at package manufacturing, now it likely works and produces suitable Solaris PKG files. * Makefile.in: Copy current compilation time valid config.status and ChangeLog files into $MAILSHARE directory. Those give a very good idea of HOW the package was made. * packaging/solaris/*: Reworked a bit with Solaris pkgmk wizard.. * proto/cf/aliases.cf: A nuisance (unlikely a security problem) with *-group expansion. One pair of double-quotes missing. * scheduler/mq2.c: mq2_discard() function did completely bogus chain traverse. Oops.. 2000-09-19 Matti Aarnio * Makefile.in: Version 2.99.54 (retagged here) * lib/getaddrinfo.c, lib/getzenv.c, smtpserver/mxverify.c, smtpserver/smtpcmds.c: Pointer arithmetic with (void*) !! Urgh.. * smtpserver/Makefile.in: Oops.. QC failure, wrong dependency order -> compilation failure * router/functions.c, proto/cf/aliases.cf, proto/cf/fqdnalias.cf: Suspended the $(zapDSNnotify ...) processing development. The issues I am trying solve there are rather -- complicated. * router/daemonsub.c: Do test for "an inode in processing" while looking for tasks in the router subdirs -- thus won't reinsert a task already in active handling back into the queue. * proto/newdbprocessor.in: Test for "^bind", "ldap", "ordered" and "unordered" db types while regenerating the dbases.conf relations. 2000-09-18 Matti Aarnio * scheduler/transport.c: A verbose debug printing of the 2000-09-14 Matti Aarnio * smtpserver/smtpetrn.c: Teached IPv6 support to the ETRN-Cluster code. * scheduler/transport.c, scheduler/mq2auth.c: If system allows, create the "mailq" port in IPv6 space. Know how to turn IPv4-Mapped-in-IPv6 address into IPv4 one, and thus be able to test it neatly. 2000-09-08 Matti Aarnio * transports/smtp/getmxrr.c: Even when we have seen EAI_AGAIN results, be a happy camper in the end if we get ANY MX entries which have valid address records. 2000-09-07 Matti Aarnio * router/libdb/bsdbtree.c router/libdb/gdbm.c: At iterative DB walkthroughs there were some problems at these two functions. Now they should a) find data, b) terminate. 2000-09-04 Matti Aarnio * configure.in: Change linkage order of -lnsl vs. -lxnet. If -lxnet leads, SEGV is about guaranteed. :-( (That is utterly weird problem...) * libc/myhostname.c: If system does not carry getaddrinfo(3) function in its standard repertoire (or we are told not to use possibly existing system instance due to e.g. bugs), then use "old-fashioned" gethostbyname(3) function to resolve canonic hostname. * utils/makedb/dblook.c: The "-dump" mode has been broken for a while. Turned out that auto-variables for key & data need (apparently) be zeroed before even the R_FIRST position pickup succeeds. D'uh ! * lib/selfaddrs.c: Solaris resolver function don't apparently like (at all) if the host parameter to e.g. gethostbyname() is 0-size string. * smtpserver/cfgread.c, man/smtpserver.8: Recognize "PARAM bindaddr iface:...." syntax to define named interface on which SOME address you want to bind. (Prefers IPv6 if such exists, otherwise IPv4.) * smtpserver/smtpdata.c: During DATA and BDAT processing, if the feeder failed to send first 1024 bytes of content, our data transfer watchdog timer was not initialized. I was wondering why one Solaris machine kept collecting hanging smtpserver processes... * libc/mailvars.c, libc/mail.c, libc/sfmail.c: - Rearranged couple of variables to separate file so that mail.c object won't be included when not needed.. (at the scheduler, and at the mailbox) - reworked file renaming to always use ino based decimal number string at the destination directory. - reconverted mail.c to SFIO based sfmail.c 2000-08-30 Matti Aarnio * compat/sendmail/sendmail.c: - Recognize "sendmail -bd" as an order to run "zmailer" script. - Input and encode INFROM= data into DSN dataset. * proto/cf/aliases.cf, proto/cf/crossbar.cf, proto/cf/fqdnalias.cf, proto/cf/standard.cf, router/functions.c, router/prototypes.h, router/rfc822.c, router/shliaise.c: - New experiental code to handle DSN data handling bugs in alias expansion-like situations. This is NOT complete thing! - Debug code to analyze the effects of the SCHEDULERDIRHASH ZENV variable. * transports/libta/ctlopen.c: Recognize (but skip) INFROM= DSN metadata. * utils/Makefile.in, utils/mxverify/getaddrinfo.c: Copied the getaddrinfo.c from libc/ to this utility. Made "make clean" and "make distclean" work from toplevel so that they will also clean the utils/mxverify/ subdir. * Makefile.in: At the end of "make install", correct the "post-install" suggestion to correct program name. * scheduler/mailq.c: Do set the PORTNUM to the destination socket addresses before trying to connect -- connecting to port 0 is not usually a successfull thing... * libc/getaddrinfo.c: At Solaris the uname(2) call return NON NEGATIVE value on success, not necessarily ZERO as in Linux... 2000-08-29 Matti Aarnio * scheduler/mq2auth.c: Copy addresses into separate range buffers for masking and comparison. (Generalized the function). * scheduler/msgerror.c: Construct the report message somewhat differently * scheduler/udpate.c: Do certain diagnostics only when (verbose) flag is set. * transports/smtp/getmxrr.c, transports/smtp/smtp.c: Remove more old debug statements, make sure that lack of memory from getaddrinfo() is handled properly (e.g. exit(EX_OSERR) -- not best, but..) 2000-08-23 Matti Aarnio * smtpserver/mailq.c: Use getaddrinfo() to pick the peer address. * include/listutils.h: Incorrect ANSI-C pre-processor string concatenation directive use. * libc/myhostname.c: Pick any of my possible addresses, don't specify port. (We are looking for CANONIC NAME, after all..) * smtpserver/smtpdata.c: Pipelined BDAT after failed MAIL FROM segfaulted on one unprotected SS->mfp referral. Oops. 2000-08-21 Matti Aarnio * libc/getaddrinfo.c: Complete IPv4/IPv6 functionality with proper timeout error handling, and malloc()ed memory garbage collection. This supplies the function for --with-ipv6-replacement-libc cfg option. (Necessary at all systems with GNU LIBC versions up to 2.2. -- up to 2.3.2, very least) 2000-08-19 Matti Aarnio * proto/newdbprocessor.in: Process the owner/group/mode triplet too. 2000-08-18 Ambrose Li * configure.in, acconfig.h, scheduler/transport.c, transports/smtp/getmxrr.c: Linux 2.2.* with libc5 fixes. 2000-08-17 Matti Aarnio * smtpserver/policytest.c, router/libdb/bsdhash.c, router/libdb/bsdbtree.c, utils/makedb/makedb.c, utils/vacation/vacation.c: RedHat 6.0 BSD DB surprised somewhat... Fixes for the configuration senses. * configure.in, libs/Makefile.in: library building order correction, and autoconfiguration straightening so that lack of dbm databases is reported corrently. (The NDBM even with its faults is sort of ok, but the DBM - its predecessor - is not...) 2000-08-02 Matti Aarnio * include/libz.h, scheduler/scheduler.c, scheduler/update.c, lib/esyslib.c, include/ta.h: To track strange looking eunlink() problem reported by Eugene my first approach is to add identity tags for eunlink() calls so I can have some additional clue as to *which* of four possible places the problems are occurring at. (Fift place is impossible, fortunately.) 2000-08-02 Eugene Crosser * transports/smtp/getmxrr.c: Yet another wrong index variable usage -- one which affects MX-list purge of self-identity matching of MX servers. Symptom: "No MX, no address" [mea: That may have been part of the problem, getaddrinfo() yielding EAI_NONAME in places of EAI_AGAIN may have been another.. -- at least for me.] 2000-08-02 Matti Aarnio * libc/getaddrinfo.c, libc/getnameinfo.c: Picked lattest versions out from GLIBC 2.1.92 source, and cleaned it up to be simple single-threaded thing without locks, and without IPv6 support. STILL PROBABLY LEAKS MEMORY! 2000-08-01 Matti Aarnio * transports/smtp/smtp.c: Some debug printout revisions. Somewhat embarassing to have a SEGV at debug code when work-code without it works.. * configure.in: Modifications on how NDBM vs. GDBM vs. BSD DB libraries are linked to ZMailer. NDBM must be explicitely linked as '-lc' if it is at libc, *and* that linkage reference must be before any other as only then will there be change to get real NDBM, and not some fake layers which all those others do provide! 2000-07-31 Matti Aarnio * scheduler/scheduler.h, scheduler/scheduler.c, scheduler/msgerror.c: Recognize our internal NOTIFY=TRACE flag. * transports/smtp/smtp.c: When 'NOTIFY=TRACE' flag is present, don't send it out as it is our INTERNAL thing. Thus the change from string based output to flag-based output routine. * include/ta.h, include/mailer.h, include/libz.h: Moved STREQ()/STREQN()/CISTREQ()/CISTREQN() macroes from mailer.h (and ta.h) to libz.h. * utils/md5sum.c: Completed usage() function, and some comments. * compat/sendmail/sendmail.c: Added acceptance for our internal -N TRACE flag, which is for eased debugging of RFC 2852 features. 2000-07-30 Matti Aarnio * libident/identuser.c: A buglet in read loop. The 'break' affects only the innermost loop construct, and it needed to go out the outer too.. Nuisance thing, I would say. (And usage of indent is questionable anyway..) * smtpserver/smtpserver.h, smtpserver/smtpcmds.c, smtpserver/cfgread.c, smtpserver/smtpserver.c, include/ta.h, transports/libta/ctlopen.c: Initial hooks for RFC 2852. * transports/libta/diagnostic.c: Turn on, temporarily, O_APPEND flag on the file. * scheduler/msgerror.c: Missed variable inits, and malformed format statement. * transports/smtp/getmxrr.c: dnscache by djb produces replies which don't have NS or AD sections, and when scanning replies, we had one-off comparison error which bailed out in case the ANSWER section reached the end of the reply packet. * man/scheduler.8: Updating on data formats and communication protocols. * transports/libta/lockaddr.c: Be carefull as *not* to overwrite part of diagnostic data with spaces. This has never been a bug, except that trying to understand some aspects of the system while updating documents revealed that we do unnecessary scrubbing there. 2000-07-29 Matti Aarnio * transports/smtp/Makefile.in, transports/smtp/smtp.h, transports/smtp/smtp.c, transports/smtp/getmxrr.c: Moved getmxrr() function to separate file, and wrapped it with a test shell in attempt to understand *why* it seems to fail at some inputs at some machines. * scheduler/msgerror.c: Hooks for "expanded" action status. * transports/smtp/smtp.c: Streamlined error handling when returning from getmxrr(). 2000-07-27 Matti Aarnio * transports/smtp/smtp.c: - At getmxrr() a missed state variable initialization (to zero) caused errorneous eternal loops. OOPS! - smtp_sync() didn't handle RCPTSTATE_500 at all.. (oops). - calls to stashmyaddresses() don't need to carry determined self-host-identity. 2000-07-24 Matti Aarnio * scheduler/msgerror.c: A simple spelling error in normally invisible text sent towards the user inside the delivery status reports. 2000-07-18 Matti Aarnio * transports/smtp/smtp.c: Possibly ssfgets() fell on over-by-one allocation/storage error so that the buffer tail got stomped over, and some systems didn't like it --> they crashed. * configure.in, router/db.c, router/libdb/bsdbtree.c, router/libdb/bsdhash.c, smtpserver/policytest.h, smtpserver/policytest.c, utils/makedb/makedb.c, utils/makedb/dblook.c, utils/vacation/vacation.c, README.UPGRADING: Reworked things to support another kind of set of SleepyCat DB libraries. Not yet covering version 3.x. * lib/rfc822scan.c: A bit more characters into "no need to quote these here" set -- allow address <@[1.2.3.4]:foo@bar> and 2000-07-17 Matti Aarnio * libs/Makefile.in, Makefile.in: Merge libsfio.a generation into this file also, so that there is no absolute need to run its make as the first thing among subdirs. Also made top-level "make libs" target which recompiles and/or just reconstructs (with ar) all libraries. * transports/smtp/smtp.c: Oops.. Solaris 2.6 --with-bundled-libresolv does not know about BIND 8 resolver's AD and CD bits. Don't use them. * Makefile.in: Version 2.99.53 * smtpserver/smtpcmds.c: At MAIL FROM policy reporting we did always yield 400-series reply even when actual code directed us to yield 500-series.. 2000-07-14 Grace He * router/dateparse.c: Fix for "Date: 00-06-29 03:50:11" induced crashes 2000-07-14 Matti Aarnio * lib/selfaddrs.c: Rearrange SIOCGLIFCONF / SIOCGIFCONF code a bit, will prefer SIOCGLIFCONF over SIOCGIFCONF if both are available. * lib/selfaddrs.c, lib/zgetifaddress.c, include/libz.h, configure.in, utils/makedb/Makefile.in: Solaris 8 test derived fixes. 2000-07-12 John D Groenveld * man/Makefile.in: If "${MANDIR}/man3" directory does not exist, create it. 2000-07-11 Matti Aarnio * transports/smtp/smtp.c: Sometimes printing to logfp was done unconditionally -- when the logfp variable carries NULL :-/ 2000-06-22 Matti Aarnio * transports/smtp/smtp.c, smtpserver/mxverify.c: Further commentary from bind-workers told me that unless the AA-bit (authoritative answer) is set, the received ADDATA isn't possibly good enough to trust - some of A/AAAAs may have expired. Thus we shall use that reply section for A/AAAA only if AA-bit is set. 2000-06-21 Matti Aarnio * smtpserver/mxverify.c, transports/smtp/smtp.c: Made smtpserver MXVERIFY to use DNS ADDITIONAL DATA, and reorganized A record collection so that if some MX server does not have A, that is asked for it explicitely, and if ipv6 is in use, if AAAA data does not exist for a server, it is asked for. In case ipv6 is not in use, this collapses into plain A type data retrieval. * lib/selfaddrs.c: Removed a test which is never true. * transports/smtp/smtp.c: Some more tuning of getmxrr() pathology processing. 2000-06-20 Matti Aarnio * transports/smtp/smtp.c: Use zgetifaddress() for '-L iface:eth0' binding. * configure.in, include/libz.h, lib/Makefile.in, lib/selfaddrs.c, lib/zgetifaddress.c: Autoconfigure BSDI derived getifaddrs() call, and have sensible IOCTL set for both interface list, and for interface address lookup code. * scheduler/mailq.c: One-off in array realloc() biting us at Solaris.. * transports/smtp/smtp.c, nam/smtp.8: Coded '-L' option processing to bind to named local interface. * lib/selfaddrs.c: Experimental SIOCGLIFCONF code for IPv6 speaking some-BSD systems -- to pick IPv6 interface addresses from the system. * transports/smtp/smtp.c, transports/smtp/smtp.h: When doing getmxrr() for MX data, will use also the DNS supplied ADDITIONAL DATA segment materiel for MX machine addresses. 2000-06-19 Matti Aarnio * scheduler/update.c: Organically relating to smtpopen() -> "retryat +60" code, if NEXT thing to feed has its expiry tag in past (and non-zero), then *stay* in CFSTATE_STUFFING so that possible next thing will be quickly diagnosed out of the queue at the end of its expiry time.. ALSO fixed double-counting of attempts on therads and vertices! * transports/smtp/smtp.c: If smtpopen() yields EX_TEMPFAIL, always do diagnostics with 60 seconds timeout flag.. ( -> "retryat +60" ) so that the entire message queue will not be retried there just for no good reason.. * smtpserver/smtptls.c: Following some discussion at OpenSSL list, I noticed that DH type keys are not supported by our current code. Now we should match that of OpenSSL: apps/s_server.c sample code again. * transports/smtp/smtp.c: Allow ssfgets() to enlarge the input string in runtime, which enables us later to make weird things like multiple hosts feeding in same connection... (perhaps) * transports/smtp/smtp.c: Continued the rewrite by handling weird looking errors from some sites which mistakenly limit the number of recipients per smtp connection (session)! Will do close of the socket after each case where RCPT TO yields 400 series code. (That will also knock on some real temporary failures, like DNS timeouts when recipient does verify the through-going address for e.g. MX analysis.) * transports/smtp/smtp.h, transports/smtp/smtp.c: Rewrote (in a sane way) the pipelined smtp_sync() status code formulation so that we won't reject recipients just because their system yields 400 code for all RCPTs which they get, and then said to DATA that 'no recipients given yet'.. 2000-06-14 Matti Aarnio * proto/cf/fqdnalias.cf: Don't fqdn-alias if it looks like RFC-822 source route address. * lib/rfc822scan.c: Allow RFC-822 source route addresses to go thru without needing quotes -- allow ',' and ':' to be included into addresses without quotes. 2000-06-13 Matti Aarnio * proto/cf/fqdnalias.cf: A missed thing in privilege processing in case db lookup succeeds. * scheduler/scheduler.h, scheduler/mq2.c, scheduler/mailq.c, scheduler/threads.c: Implemented (in C) initial part of 'mailq' and 'mailq -v' commands for MAILQ-V2 protocol. * scheduler/scheduler.c, man/scheduler.8, scheduler/mq2.c, scheduler/mailq.pl: Playing with MAILQ-V2 code to get (experimental) version of detailed reporting a'la old mailq-v1 code. * scheduler/mailq.pl: Now this PERL mailq supports functions: mailq mailq -Q mailq -QQ mailq -QQQ and has also option '-U' for logging with non-default username/password pair. 2000-06-07 Matti Aarnio * transports/libta/ctlopen.c: If ctlopen() contained open() of message body file fails, send a warning() up to the scheduler log WITH errno of that open() call.. * proto/db/smtp-policy.src: _RBL* alias labels are turned to LOWER CASE -- those keys are lowercased during encoding, but data attributes are not modified while compiling the database, nor when looking up tags.. * smtpserver/smtpserver.h, smtpserver/smtpdata.c, smtpserver/cfgread.c, smtpserver/contentpolicy.c, utils/smtp-contentfilter.in: CONTENT policy interface, and sample filter implementation.. * scheduler/scheduler.c, scheduler/conf.c, scheduler/msgerror.c, man/scheduler.8: Modify the '-n' option semantics to control how DSN RET option value is set in absense of originator supplied value. 2000-06-02 Matti Aarnio * scheduler/mailq.c: Count files in POSTOFFICE/router/ directory also when it is subdir-hashed.. * router/daemonsub.c: Tried actually to use ROUTERDIRHASH=1 mode, and learned of bitter errors at the subchild feeder code... Dirname was written twice to the workfilename. * router/libdb/header.c: Allow just about *any* character to appear at headers database key fields.. old limits were too strict anyway. * proto/post-install.sh.in: More minor tweaks. 2000-05-28 Matti Aarnio * proto/Makefilein, proto/post-install.sh.in: Small fixes that disjoint installation test shows up.. 2000-05-26 Matti Aarnio * transports/smtp/smtp.c, transports/smtp/smtp.h: Cosmetics-like things about diagnostics tracking. * proto/newdb.in, proto/newdbprocessor.in, proto/post-install.sh.in: These are all PERL scripts now -- that last one should be renamed, thinking about it... 2000-05-25 Matti Aarnio * transports/smtp/smtptls.c: Proper network timeout monitored non-blocking wrapper for SSL_connect() code. * transports/smtp/smtp.h, transports/smtp/smtp.c: Reworked diagnostics codes a bit to collect only relevant SMTP transaction bits per each recipient. (Although still not *quite* correct for problems diagnosed at DATA/dot time, which need all recipients, not just the last one...) 2000-05-24 Matti Aarnio * proto/newdb.in: Rewritten into PERL(v5) script. * smtpserver/rfc821scn.c, smtpserver/smtpcmds.c: Eugene Crosser wanted to have: , our scanner demanded an alphanum starter character for the sub-fields of the local-part. Now allowing also the underscore as starter there -- for a username.. (smtpcmds.c has some diagnostics report beauty warts.) 2000-05-18 Matti Aarnio * proto/newdb.in, router/libdb/ndbm.c, router/libdb/dbm.c: At NewDBM (ndbm) the dbm_dirfno() does not always yield success at the dbm_open(). It *may* yield some scrap data for a while -- especially if the db is *very* small. The dbm_pagfno() yields always ok filehandle number, and thus we use it -- and change file moving order to .DIR .PAG to match which is tested later. 2000-05-10 Matti Aarnio * proto/Makefile.in, proto/post-install.sh.in: Moved POSTOFFICE directory subdir generation into the post-install.sh script. Execution of that script before system becomes ready for running is now MANDATORY! (At least at fresh installations.) 2000-05-09 Matti Aarnio * router/rfc822.c: Existence of ZENV variable ``SCHEDULERDIRHASH'' drives movement of files from router to the scheduler to go straight to hashed subdirs at the scheduler. * router/daemonsub.c: A lot more work at the daemon subsystem. Now it runs a dedicated set of router processes for the primary router directory, plus one for each of ``ROUTERDIRS'' variable defined auxiliary directories. * libc/mail.c, router/daemonsub.c, man/router.8, SiteConfig.in: If ZENV variable ``ROUTERDIRHASH'' exists, and has value ``1'', all message submissions will be using direct moving nto ``A'' thru ``Z'' hash-subdirectories -- if they don't exist, things FAIL! * scheduler/mq2auth.c, scheduler/transport.c: Mailq-V2 IP-ACL system works! * scheduler/scheduler.c, scheduler/transports.c: Call syncweb() also inside the mux() routine so that new jobs will get assimilated faster. 2000-05-08 Matti Aarnio * libc/mail.c, router/rfc822.c: Simplified message injection code a bit; especially tried to lessen the number of directory operations. (Even though comments in the source say that rename(2) isn't safe, we are using it...) * INSTALL: Refer to HTML renderation of ZMailer manual's install part.. * scheduler/msgerror.c: Drop leading \r from report data. No weird looking blank line in resulting report. * transports/smtp/smtp.c: Don't diagnose TLS status unconditionally, it is annoying thing to have accumulating into scheduler's log.. * smtpserver/smtpserver.c: When --with-whoson is used at configuration, and smtpserver is run without '-w' option, policy test is falling on default of "yes this is always ok relay source" behaviour -- which in general is wrong :-( 2000-05-04 Matti Aarnio * lib/selfaddrs.c: Yet Another small surprise at Linux systems about self-address extraction. :-/ 2000-04-20 Matti Aarnio * smtpserver/smtpcmds.c, smtpserver/smtpserver.c: Fix "from_box" setting to match current treatment of user input string (e.g. not stomping things with '\0' chars..) Synced MaxErrorRecipients default value with documentation. * smtpserver/cfgread.c: Documentation (proto/smtpserver.conf.in) says that PARAM name is: RcptLimitCount, while it was recognized only as RcptLimitCnt ... Oops. 2000-04-18 Matti Aarnio * router/db.c: A mistake in cache LRU chain handling caused eternal loop to form sometimes. Oops. 2000-04-14 Matti Aarnio * transports/libta/ctlopen.c: An odd corner case of error handling allowed access to uninitialized elements of an array -> junk data blew up.. * scheduler/threads.c, scheduler/update.c: Trying to understand why 'deferred' diagnostics causes practically immediate reissuance of a task.. Understanding dawns, thread rescheduling was implemented in fumes, so crazy it was (but then, I don't drink nor smoke...) * router/functions.c, lib/rfc822scan.c: Router function "$(dequote ...)" was previously castrated a bit too much. This isn't like sendmail-8's dequote class, but far more brutal beast.. * proto/Makefile.in: Fix db subdir installation to work on a system without GNU/BSD install (using locally supplied script). * compat/rmail/Makefile.in: Allow "mkdir -p ..." to fail on Solaris. (Anywhere actually) * libc/whathost.c: Improved error handling by adding posixy redo of things in case stat()s yield an error.. (Solaris 2.6 is complaining somewhat in these routines..) * transports/smtp/smtptls.c: Moved things around so that it works also when compiled WITHOUT (Open)SSL. 2000-04-13 Matti Aarnio * transports/smtp/smtp.c: Moved one instance of tls_conf_file inside HAVE_OPENSSL conditional. Oops.. * transports/smtp/smtp.c, transports/smtp/smtptls.c, transports/smtp/smtp.h: Try to combat certain TLS related problems mentioned before. (An hour later: Cured them apparently..) * smtpserver/smtptls.c, smtpserver/smtpserver.c: Under TLS do always flush replies - to simulate one other MTA running under TLS wrapper. (This was removed a bit later.) 2000-04-12 Matti Aarnio * router/db.c: For Matthias Urlich; a method to declare at RELATION entry if NEGATIVE CACHE is used at the particular relation (pre- requires that cache is also used..) Driven with: -N option. 2000-04-11 Matti Aarnio * scheduler/mq2.c, scheduler/mq2auth.c, scheduler/prototypes.h, scheduler/scheduler.h, scheduler/scheduler.c, scheduler/transport.c: Hooks for verifying MAILQv2 contactee IP address. * router/db.c: Mathias Ulrichs suggest negative cache for DB lookup related caches. It is simple to do, but not so simple to explain to be usefull... 2000-04-08 Matti Aarnio * scheduler/msgerror.c, scheduler/prototypes.h, scheduler/threads.c, scheduler/transport.c: Tracked possible reasons for processing crashes when maxthr>1. * transports/smtp/smtp.c: Changed the commentary log accumulation method, now it will clear only at "MAIL FROM..." and from then on collect all until next "MAIL FROM...". Should make easier to understand some odd error reports. * scheduler/scheduler.c, scheduler/threads.c, scheduler/update.c: Added some debugging related memset()s just before freeing some pointer containing memory blocks. Also modified unvertex() slightly to forward nextfeed in case current vertex is (at all) in any sort of processing. 2000-04-07 Matti Aarnio * scheduler/update.c: Log only really important diagnostics about error conditions in the system, not errors in email receiving. * scheduler/threads.c: One instance of vertex free() has been processed improperly. If the 'nextfeed' is still pointing to the vertex which is being deleted, it must be advanced.. Although getting this means that the scheduler has just fed the same job twice to TAs.. 2000-04-06 Matti Aarnio * transports/smtp/smtptls.c: Disable TLS read_ahead setting. That mode doesn't work quite well, and can definitely groak against Stalker Software's CommuniGate Pro smtpserver when running in PIPELINING mode. 2000-04-02 Matti Aarnio * proto/cf/aliases.cf, proto/cf/fqdnalias.cf: - More work on rewritten expansion loop prevention measures. - Added UserDB sender address rewriter into fqdnalias.cf too. * lib/rfc822scan.c, router/functions.c, proto/cf/aliases.cf: Reworked alias (et.al.) expansion loop preventation machinery. Now a message aliased as: user1: user2, user1 will be delivered to both, unless some other mapping diverts. NOT YET A COMPLETE THING! 2000-03-30 Matti Aarnio * proto/zmailer.sh.in, proto/newdb.in, proto/db/dbases.conf: Now 'zmailer newdb' uses the $MAILVAR/db/dbases.conf file, if it exists. (But the converted needs to be rewritten, and security issues need more thought...) 2000-03-29 Matti Aarnio * scheduler/threads.c: thread_reschedule() crashed on thr->thvertices being NULL ?? * router/rfc822hdrs.c: Sometimes missed sending out at least one space after the header name tag. * scheduler/transport.c, scheduler/update.c, scheduler/mq2auth.c, transports/smtp/smtptls.c: Various bad diagnostics seen (from smtp) which in turn crashed scheduler :-/ Oops.. Some new comments into mq2auth.c about possible future directions. * proto/cf/crossbar.cf, proto/db/userdb: Some more work on this, now allowing rather peculiar ZMailerish keying at least for ":mailname" keys.. * router/db.c: Reworked cache code to be faster (I hope) by means of having a crc32 hash value against which the matching is done. The goal is to support in order of 10k entries in the cache without too large of an impact on performance. ( The hash collisions are resolved by means of doing actual key strcmp() ) The cache at each DB relation interface is still an LRU with linear (urgh..) search time. (A splay-tree was also considered, but decided not to use.) 2000-03-28 Matti Aarnio * proto/cf/aliases.cf, proto/cf/crossbar.cf, proto/cf/standard.cf, proto/db/dbases.conf, proto/db/kill-headers, proto/db/userdb: Created a way (always on when DB is defined, and rewrite (at crossbar.cf) happens via 'internet') to do ``username'' mapping to given FQDN address, e.g. ``user.name@here.domain'' Reworked kill-headers initialization. * include/mailer.h, proto/Makefile.in, proto/cf/standard.cf, proto/db/kill-headers, router/rfc822.c, router/libdb/header.c: Runtime configuration mechanism for killing named headers. We have some uses for removing e.g. gateway generated "X-Login-ID:" type of things which no recipient has any business of seeing. * Makefile.in: Version 2.99.53-pre1 2000-03-27 Matti Aarnio * transports/smtp/smtp.c: When the outbound SMTP fp is closed, don't truncate the reported address -- mainly to stop "." appearing as the last outbound state, when "data" really is it.. * scheduler/MAILQ-V2, scheduler/mailq.pl, scheduler/mq2.c, scheduler/prototypes.h, scheduler/threads.c: Playing with ideas regarding MAILQ-V2 protocol. 2000-03-24 Matti Aarnio * configure.in, transports/mailbox/dotlock.c: Autosense 'HAVE_GETHOSTID', and if it isn't, try to create it in couple different ways (For earlier Solaris, and HPUX systems, very least.) 2000-03-23 Matti Aarnio * INSTALL, configure.in, Makefile.in, sfio/Makefile.in, sfio/src/lib/sfio/makefile.in, sfio/src/lib/sfio/{Sfio_{dc,f},Stdio_{b,s}}: Made SFIO compilation environment to work also when running in "subdirectory" configuration (e.g. using ../configure ...) Also cleaned up the SFIO configuration system a bit. * doc/guides/smtp-policy: Added some words about the policy testing by using -T option. 2000-03-21 Matti Aarnio * man/smtpserver.8, smtpserver/smtpserver.h, smtpserver/smtpserver.c: Implemented -T option for a local interactive policy testing without having connectivity all around the world. See the man-page. 2000-03-18 Matti Aarnio * router/daemonsub.c, router/conf.c: Spotted a self-healing misbehaviour in the child process feeder, which skipped messages from queue quite *often*. Cured that by waiting in proper place for successfull delivery. Also changed some default values for "stability" processing. * router/rfc822hdrs.c, router/rfc822.c, lib/token.c: Aborted (and deactivated) foldable tokens processing on headers we don't know about right now. Perhaps need to have some other approach on the problem... 2000-03-17 Matti Aarnio * smtpserver/smtpcmds.c, smtpserver/smtprouter.c: Instead of using printf format: %*s we need to use %.*s * transports/libta/swriteheaders.c: Eugene Crosser spotted CHUNKING mode problem with message header TAB expansion. Outch! * router/router.c: logit() Changed stdout written log printf() format so that quite large lists of addresses are allowed. 2000-03-15 Matti Aarnio * configure.in, transports/mailbox/mailbox.c: "--without-maillock" configuration option from Eugene Crosser. We don't seem to be quite as happy chaps with Solaris maillock() as we would like to be in all cases :-/ (E.g. if we use any other than /var/mail/ directory for mailboxes, Solaris maillock() will still put locks into /var/mail/ directory :-? Hmm..) 2000-03-14 Matti Aarnio * scheduler/msgerror.c: Reworked report suppression; if all RCPTs have NOTIFY=NEVER, no report shall be made. Some cleanup of report generation. * transports/mailbox/mailbox.c: Small cleanup of RFC 1894 action status reports. * transports/smtp/smtp.c: - Do (at least try) shunt new scheduler TA driver state- machine only when connection fails, not when a message delivery yields some 400 series status code. - PIPELINING mode in smtp_sync() did react suboptimally (wrong) to DATA yielding 500 series code when MAIL FROM or any/all RCPT TO commands yield 400 series status. - DRUMS/SMTPUPD tells that if MX records are available, but none are usable (preference pruning rule), this must be reported as an error. We allow '-W' mode to override this so that when '-W' is present, this case is not an error, IF destination domain has A + WKS/smtp permitting delivery there. * smtpserver/smtpcmds.c: DRUMS/SMTPUPD tells that policy based rejections of recipients shall use 550 (and 450) codes. 2000-03-09 Matti Aarnio * transports/libta/diagnostic.c, include/ta.h, transports/smtp/smtp.c: Some diagnostics related rework due to absolutely insanely behaving SMTP reception. Diagnostic output in EX_TEMPFAIL or EX_IOERR are not per default (actually never) stored directly to the file by the TA program, and anyway doing diagnostics in such case is marginally usefull -- retries may happen, for example! Now things should allow recipient address to be retries properly even with mode locking working ok. 2000-03-05 Matti Aarnio * smtpserver/smtpcmds.c: More systematic approach at MAIL FROM and RCPT TO related reporting where related addresses are wanted to be included into the diagnostic messages. All that just for completely braindead approach prevalent in M$ client software... 2000-03-03 Matti Aarnio * scheduler/msgerror.c: Basic footwork for DELAYED reporter. (Incomplete yet.) * scheduler/scheduler.c: Reverted earlier change which counted already processed recipients into MTA-MIB variables. * doc/Makefile.in, man/Makefile.in, libresolv/Makefile.in, scheduler/Makefile.in: Fixed "make depend" to work. * scheduler/Makefile.in, scheduler/servauth.h: DELETED servauth.h -- something entirely forgotten, and same stuff reimplemented elsewere. * man/smtp.8, man/smtpserver.8: Updates/fixes to text. * scheduler/scheduler.c, scheduler/scheduler.h, scheduler/conf.c, scheduler/agenda.c, scheduler/msgerror.c, scheduler/readconfig.c, scheduler/prototypes.h, man/scheduler.8: - Slightly simplified doagenda() - Created accumulated report collector/reporter subservice. 2000-03-01 Matti Aarnio * proto/forms/delivery: Added an URL reference to a "decode key" at http://www.zmailer.org/reports.html * scheduler/scheduler.h, scheduler/update.c, scheduler/threads.c, scheduler/prototypes.h, scheduler/readconfig.c, scheduler/msgerror.c, man/scheduler.8: - Reworked how IDLE handles idle process recycling; always the channel must match, host may or may not need to match. - Reworked parts of error reporting in preparation for creating delay reports. (And earlier FAILED/DELIVERED/RELAYED reports at the same time, of course..) * proto/cf/fqdnalias.cf: Spotted some FIXME things.. * transports/smtp/smtp.c: When socket write failure causes shutdown of write direction socket, do turn all back reported 500 series errors into softer 400 series codes. Won't unnecessarily reject messages. 2000-02-29 Matti Aarnio * smtpserver/smtpetrn.c, man/smtpserver.8, man/scheduler.8: Completed the etrn-cluster code. Seems to work, even. Updated man-pages a bit. * smtpserver/smtpserver.c, smtpserver/smtpserver.h, smtpserver/cfgread.c, smtpserver/smtpcmds.c: - PARAM sum-sizeoption-value -- modify how MAIL FROM:<> SIZE=nn option value is processed when compared against various limits in the system; if the option is present, effectively the system required free space is message size multiplied by the recipient count. - PARAM max-unknown-commands nn -- how many pure junk commands can be accepted before our side hungs the connection. Default limit: 10 2000-02-28 Matti Aarnio * smtpserver/smtpserver.c, smtpserver/smtpserver.h, smtpserver/cfgread.c, smtpserver/smtpetrn.c, proto/smtpserver.conf.in, doc/guides/etrn-cluster: - A bug in '-l SYSLOG' which bite me at Solaris.. Oops.. - some reworking of ETRN-CLUSTER code, but no completition yet. 2000-02-25 Matti Aarnio * transports/smtp/smtp.c: If DATA-dot-phase ending dot yields EX_TEMPFAIL, don't do diagnostics right after smtpwrite("."). It leads to bad retry results :-/ Reworked something else; simplified code for header output -- which will be "obviously correct". * smtpserver/smtpserver.c: Make '-l SYSLOG' option to work in parallel with possible '-l /path/to/file' logging option. 2000-02-18 Matti Aarnio * configure.in: Moved --with-zconfig= code around a bit, apparently it was quite ineffective, and has been that way for quite a while :-/ (More so the documentation spoke of some other option..) 2000-02-17 Eugene Crosser * smtpserver/smtpdata.c, configure.in, acconfig.h: (Auto)configure USE_TRANSLATION into smtpserver. [mea] minimized Eugene's patch inside smtpserver code by rearrangeing 'USE_TRANSLATION' usage in smtpdata.c (e.g.: Eugene wrote it, Matti possibly botched it..) 2000-02-15 Matti Aarnio * scheduler/agenda.c, scheduler/mq2.c, scheduler/prototypes.h, scheduler/scheduler.c, scheduler/scheduler.h, scheduler/threads.c, scheduler/transport.c, man/scheduler.8, scheduler/MAILQv2: - Streamlined MAILQv2 processing -- it should now be as low latency service as possible, even with long-lasting operations blocking at things like input queue scanner. - Implemented a 30 second apopteosis for MAILQv2 sockets. - Implemented MAILQv2 action ETRN -- exactly alike the SMTP- server one. 2000-02-12 Matti Aarnio * lib/selfaddrs.c: In *addition* to interface picked addresses, collect also z-environment defined SELFADDRESSES= listed addresses. (For the support of an ETRN-CLUSTER.) * transports/smtp/smtp.c: A read timeout at response wait after write timeout is now limited to 300 seconds. * smtpserver/Makefile.in, smtpserver/smtpserver.h, smtpserver/cfgread.c, smtpserver/smtpserver.c, smtpserver/smtpcmds.c, smtpserver/smtpetrn.c, SiteConfig.in, proto/smtpserver.conf.in, doc/guides/etrn-cluster: Some initial work for an ETRN-CLUSTER. 2000-02-08 Matti Aarnio * man/smtpserver.8, smtpserver/smtpserver.h, smtpserver/smtpserver.c, smtpserver/smtpdata.c, smtpserver/smtpcmds.c, smtpserver/smtpauth.c: Created smtpserver runtime command-line option "-l SYSLOG", which is a magic thing to send logfile material through syslog() facility. Beware there is *huge* amount of dataflow there! Some updates to smtpserver.8 man-page. 2000-02-04 Matti Aarnio * smtpserver/smtpserver.c, lib/taspoolid.c: Changed (again) the logtag prefix for smtpserver's "-l" parameter defined logfile. Considering enabling it to be also diverted via syslog() to where-ever that goes. (But server process can have only one syslog() session active with one facility..) 2000-02-03 Matti Aarnio * transports/smtp/smtp.c, transports/smtp/smtptls.c: Grumble... Still seeing wrong kind of timeouts.. 2000-02-02 Matti Aarnio * transports/smtp/smtp.c: An oddball log error - misreported outwritten message header size into smtp client's log file. * README.UPGRADING, include/mail.h.in, scheduler/scheduler.c, transports/libta/markoff.c: Prepared for adding _CF_MIMESTRUCT multiline entity into the TA-specsfile. It will contain preparsed result of what the message structure is - and be able to prevent e.g. decoding QP when the end result exceeds linelength limits.. (And do proper body structure processings where appropriate) * transports/smtp/smtp.c: Lengthened the -Ttcpw= default to 5 minutes (RFC 1123: 3 minutes) * router/parsedate.c: Armour lilian() function against bad MON or DAY inputs. (Especially zero month is deadly..) However this code is currently completely dead -- not used, so... * scheduler/mq2.c, transports/mailbox/mailbox.c, transports/sm/sm.c, transports/smtp/smtp.h, transports/smtp/smtptls.c, transports/smtp/smtp.c, transports/smtp/appendlet.c: Turns out the sfsetfd() has a undesirable side-effect, which might be turnable off, but... Namely it does sfsync() on the stream before setting the fd -- we don't want that. * libsh/io.c, router/functions.c, router/prototypes.h, router/rtsyslog.c, smtpserver/smtprouter.c, transports/libta/diagnostic.c: Cleaned up several GCC 2.96 reported/complained things. (Couple of which stopped compilation..) * transports/smtp/smtp.c: More of hunting of strange errors at protocol transmission. * smtpserver/smtpserver.c: When reporting '-- pipeline input exists', do that immediately after collecting previous input line, not after SYNCing out previous reply. 2000-02-01 Matti Aarnio * transports/smtp/appendlet.c, transports/smtp/smtp.h, transports/smtp/smtptls.c, transports/smtp/smtp.c: More of the same work as yesterday evening/today early morning.. * transports/smtp/smtp.c, transports/smtp/smtptls.c, transports/libta/swriteheaders.c: Some code rearrangements to better(?) handle stream write errors. 2000-01-31 Matti Aarnio * utils/mxverify/mxverify-cgi.c: Move "test with user-suppled address" before the "test with postmaster" case. In case the Postmaster doesn't exist (which is an error in itself, though..), the user supplied address was tested very least. 2000-01-25 Matti Aarnio * transports/smtp/smtp.c, transports/smtp/appendlet.c: Small error test rearrangements so that if sferror() is set at the output stream, no writing will happen. 2000-01-24 Matti Aarnio * transports/libta/swriteheaders.c, transports/libta/fwriteheaders.c: Do proper TAB-expand in entire header strings, if so directed. * utils/mxverify/* Made it as stand-alone as possible; it still uses ZMailer's configuration system, but does not need rest of the system for compilation -- all parts it needs are localized copies. * transports/smtp/smtptls.c: smtp_sfwrite() didn't break out at write-timeout quite as I had intended.. * transports/smtp/smtp.c, transports/smtp/smtptls.c: More debug stuff for hunting strance message breaks.. 2000-01-23 Matti Aarnio * lib/token.c, router/rfc822hdrs.c: Turn all header start spaces to optimized TABs plus spaces. * smtpserver/smtpcmds.c: Now (!) people tell me that the most common M$ email clients are absolutely stupid, and won't be able to correlate outgoing SMTP verbs to responses they receive. They just tell to user the error which Z returns, and as it doesn't repeat the address in fault (why should it?), user has no clue... (So the one-line reports will from now on also mention the received address...) * utils/mxverify/Makefile.in, utils/mxverify/mxverify-cgi.c, utils/mxverify/mxverify-cgi.html, utils/Makefile.in, utils/mxverify-cgi.c, utils/mxverify-cgi.html Moved the thing to subdirectory. Now it can be compiled without compiling the *entire* other ZMailer -- just: ./configure --prefix=/usr cd utils/mxverify ; make 2000-01-20 Matti Aarnio * utils/listexpand.c: Added parametrization on how large bundles to make. Default is the same old 200.. * transports/smtp/smtp.c: Turned out that Linux glibc 2.1 getaddrinfo() has some strange bugs when it is left to its devices to scan for a generic address suite -- it doesn't scan only IPv4 and IPv6, but also some others, and if it gets error anywhere, it seems to barf.. Will now scan for IPv4 and IPv6 separately :-/ So much for the great transparent API ... * utils/mxverify-cgi.c, utils/mxverify-cgi.html: Detected the above problem while playing with this utility, and developed that change here.. Reworked quite some things after initial deployment, added that page into CVS for supplementary info. 2000-01-19 Matti Aarnio * include/ta.h, transports/libta/diagnostic.c, transports/smtp/smtp.c, transports/libta/tasyslog.c: Report (at smtp) also the conversion mode used at the reported session. * utils/mxverify-cgi.c, utils/Makefile.in: Writing a utility to ease destination domain verification by testing *all* MXes of given domain. 2000-01-17 Matti Aarnio * smtpserver/smtpserver.h, smtpserver/smtpserver.c: Accept SMTP verbs which have numbers in them. Treat verb "550" by responding SILENTLY to it -- e.g. don't reply anything when get it. * smtpserver/mxverify.c: Treat RBL-type lookups as ERRORS only when the DNS yields A record under mask: 127.0.0.0/24. There exists a zone: .ok.orbs.org which has 127.127.0.0/24 and THOSE are verified to be ok. If using ORBS, use first the OK zone, only then the RELAYS zone. ( -> On larger blocked zone, an individual OK entry can accept email. ) 2000-01-14 Matti Aarnio * scheduler/msgerror.c, include/ta.h, transports/libta/diagnostic.c, transports/smtp/smtp.c: Extend the notary report with "X-ZTAID:" which is diagnostic about which TA process did generate the error report. Allows one to hunt things from the syslog.. * smtpserver/smtpserver.c: Doing longjmp() from within a signal handler isn't kosher -> sigsetjmp()/siglongjmp() and friends.. (Not that that signal handler should ever be called in normal operation of the smtpserver..) * scheduler/readconfig.c: Pulled back (disabled) code related to selector clause wild-card testing in clause's "channel" part. * proto/cf/standard.cf: Fixed recent thoughts mentioned below - coded test end statement ("fi") was missing. 2000-01-07 Matti Aarnio * scheduler/threads.c: Sometimes adding new tasks into a long-running thread did cause the "unfed" count to go to *negative* value. Now counting tasks there always when they arrive to the thread, not only when all previous ones have already been fed.. (caring about it only when a process is active on the thread.) * proto/cf/aliases.cf, proto/cf/fqdnalias.cf, proto/cf/i-iprouter.cf, proto/cf/i-router.cf, proto/cf/standard.cf, proto/db/dbases.conf: Some collected thoughts about how to do multiple instances of several popular databases (like aliases) in varying DB formats. * router/libdb/bind.c, proto/cf/i-smtp.cf: Created sub-mode 'mxlocal' for the BIND database. If local system is lowest MX for given domain, then it yields string (query domain) and zero return code, else NIL and non-zero return code. * Makefile.in, bin/usa-itar-stripper.sh: 'make dist-usa' creates now - DIRNAME-usa.tar package (vs. DIRNAME.tar of 'make dist') - crypto-DIRNAME.diff file * transports/smtp/smtp.h, transports/smtp/smtp.c: Seen strange junk commands fed out to the remote systems, trying to determine what is going on.. 2000-01-05 Matti Aarnio * smtpserver/smtpcmds.c: Change several policy situation '550' codes to '553'; temporary failure indicators shall stay as '440'. 2000-01-04 Matti Aarnio * smtpserver/smtpserver.h, smtpserver/smtpserver.c: Killed 'HAVE_TCPD_H' code from these files; using smtpserver's policy database is better -- way better.. * transports/smtp/smtp.c, include/libz.h, lib/parseintv.c, lib/Makefile.in, scheduler/readconfig.c: - Moved parse_interval() into library (from scheduler). - Using it at smtp-ta for timeout parameter parsing * transports/smtp/smtp.c: If 'DATA' yields '500' series, don't just do RSET + retry at the same stream. (EX_IOERR) 2000-01-03 Matti Aarnio * router/libdb/headers.c: Darn.. Can't just disable "date" from known headers, however changeing their semantics to 'nilHeaderSemantics' should do it. * scheduler/threads.c: abort() replaced with correct code -- "the thing never to happen *did* happen"... 2000-01-02 Matti Aarnio * router/rfc822.c, lib/skip821address.c, transports/libta/ctlopen.c: Have seen one mighty strange local non-delivery, and learned ways to use quotes in addresses which are not always safe from router to scheduler, and to TAs.. * man/smtp.8, proto/smtp-tls.conf{,.in}, doc/guides/openssl, configure.in, proto/Makefile.in, proto/scheduler.conf.in, proto/post-install.sh.in: - Updated 'smtp.8' man-page - Updated OpenSSL GUIDE - Updated the prototype of smtp-tls.conf (and renamed it for autoconf reasons) - Install the smtp-tls.conf prototype file (for post-install) - Updated post-install.sh - Added example usage clause to scheduler.conf.in * transports/smtp/smtp.c, transports/smtp/smtp.h, transports/smtp/smtptls.c, transports/smtp/appendlet.c: Reworked the timeout parametrizations somewhat (per what a colleque of mine had asked for) * router/libdb/header.c, router/rfc822hdrs.c: Remove the "Date" headers from specially recognized set. Will end PARSING of the headers completely - it is an UA problem, after all. --> WHOOPS! Must fix also router/rfc822hdrs.c ! * scheduler/scheduler.c: Properly analyze how many FDs each bidirectional TA connection needs. One or two. Calculate the maximum number of TA childs we can run. * scheduler/threads.c, scheduler/readconfig.c, scheduler/scheduler.h, scheduler/transport.c: Modified the process state algorithms a bit. - Process will stay in STUFFING until it gets #hungry in that state without that the thread has any 'nextfeed' available. - Only when there are no more things to feed in STUFFING- state, process moves to FINISHING-state ("retryat" processing being sole exception.) - A new vertex to a thread while there are active processes, but 'nextfeed == NULL' will be placed into 'nextfeed'. - A new clause parameter ``WakeupRestartOnly'' works by allowing a newly created thread to be processed immediately, but if existing thread is in ``Retry Delay'' state, this new message will not auto-start the thread right away. * proto/scheduler.conf.in: Explained a lot (more) about the scheduler usage of various selector clause parameters in the initial comment lines. * transports/smtp/smtp.c, transports/smtp/smtptls.c, transports/smtp/smtp.h, doc/guide/openssl: Debugged the client-mode STARTTLS, it *works*! 1999-12-31 Matti Aarnio * Makefile.in: Version 2.99.52-patch1 * transports/mailbox/mailbox.c: Spotted lurking SFIO usage bug which caused missing safe buffer zone newlines at the ends of the mailbox stored messages. At least my MUTT seem to be a happy camper even with this problem present, but perhaps not all are so good.. * transports/libta/swriteheaders.c: Missed reserving space for TAB-expansions in CHUNKING mode operation of SMTP. Only against other ZMailers, and a few other rare systems.. 1999-12-30 Matti Aarnio * Makefile.in: Version 2.99.52 1999-12-30 Matti Aarnio * contrib/zmailer-spec/*: SPECS files for RPM generation from XosÉ Vázquez * Makefile.in: Version 2.99.52-pre5 * scheduler/scheduler.c: At RESYNC processing do lock verification SILENTLY, and do break the lock... * transports/mailbox/mailbox.c: When SFIO was introduced here, write to pipes wasn't anymore working quite as it should -- namely EPIPE is something which shall just be *ignored*. Wrote SFIO write discipline to handle this case. * utils/Makefile.in, utils/smtpserver-log-parser.pl: Autoconfigure the perl for that script, and install it during normal system install (and install-bin). * transports/smtp/smtp.c: - Some remote servers misbehave by issueing an error to MAIL FROM just because receiver doesn't do proper state clearance at end of DATA phase, but claims that "MAIL FROM already given" :-( Issue an RSET, and try feeding the thing again. - Subtle dangling pointers seen related to reconnect processing, fixed them now. * scheduler/threads.c: Print "mailq -Q" always as is threads are running in multi-TA mode (instead of the default of maxthr=1) * scheduler/update.c: An oddball ETRN caused SEGV - not quite sure this fix was right in the long run... * libmalloc/botch.c, libmalloc/defs.h, libmalloc/dumpheap.c, libmalloc/malloc.c, include/zmalloc.h, libmalloc/_strdup.c, libmalloc/align.h, libmalloc/globals.c, libmalloc/emalloc.c, libmalloc/malloc.h, libmalloc/strdup.c memory debugging tools -- glibc malloc hooks are better! (See smtp.c "_hook" things..) * scheduler/mq2.c: Use SFIO instead of STDIO for some debug reports, and do clean them too! * include/ta.h, transports/libta/ctlopen.c, transports/libta/diagnostic.c, transports/libta/mimeheaders.c, transports/libta/fwriteheaders.c, transports/smtp/smtp.c, transports/libta/swriteheaders.c, transports/libta/routermxes.c: Tracked memory freeup order failure! D'uh it was difficult! ( ctlclose() did free(d.contents) too early... ) 1999-12-29 Matti Aarnio * scheduler/scheduler.h, scheduler/prototypes.h, scheduler/mq2.c, scheduler/threads.c, scheduler/transport.c, scheduler/readconfig.c, scheduler/qprint.c, scheduler/update.c, proto/scheduler.conf.in, scheduler/agenda.c: Implemented changes needed for running multiple TAs at threads. 1999-12-28 Matti Aarnio * include/netdb.h, libc/getnameinfo.c: Fixes from Eugene Crosser for Cobalt box. 1999-12-23 Matti Aarnio * transports/hold/Makefile.in, transports/mailbox/Makefile.in, transports/sm/Makefile.in, transports/smtp/Makefile.in: Modify library linkage orders; in case @LIBMALLOC@ is used, link it *last*. That should avoid some problems I had recently when running memory-debug TAs.. * scheduler/transport.c: One more oddball corner case - missed to do *thread* rescheduling at the thread-switch.. * scheduler/scheduler.c, scheduler/scheduler.h: After some fixed amount of resyncs on a file, throw it away alltogether for later input queue scanner to find. 1999-12-22 Matti Aarnio * lib/selfaddrs.c: A programming error has lurked here ages -- if system has more than a few (6-8) IP addresses, previous code failed to pick *any*. Cured things so that it will now pick all. 1999-12-21 Matti Aarnio * scheduler/scheduler.h, scheduler/transport.c: Report at reclaim() to scheduler main error log if subprogram has returned with non-zero exit status. * scheduler/agenda.c, scheduler/threads.c, scheduler/prototypes.h: Moved QUEUE_ONLY flag observation inside the thread_start(). * scheduler/scheduler.c: Timeserver waking up on select() every circa 0.3 seconds. (Goal is to have *some* reliability at getting a bit more accurate 1 second tick..) * scheduler/threads.c: Fixed 'thr->attempts' accounting, threw away a pile of dead (#if 0 ... #endif) code. 1999-12-20 Matti Aarnio * scheduler/transport.c: Inverted flush_child() return code - to work like specified.. * scheduler/update.c, scheduler/scheduler.h, scheduler/threads.c, scheduler/transport.c, scheduler/conf.c, scheduler/prototypes.h: More tracking of referrals thru freed memory blocks, and various core logic cleanups. * libc/getaddrinfo.c: Include wrapper test fix 1999-12-19 Matti Aarnio * scheduler/scheduler.c: Wondered why (at Solaris) the system doesn't run with timeserver process. Possible fix implemented. (SFIO usage mistake.) * scheduler/threads.c: Spotted the final lurking thread-auto-switch bug -- "ho" and "ch" pointers at proc object were not updated at the rewritten pick_next_thread() function. * transports/smtp/smtp.c: Spotted aplenty of suboptimal diagnostic() calls, and guarded them with if (rp->lockoffset) tests. After a diagnostic is done, that variable is zero. * transports/smtp/smtp.c: Consider DNS originating NONAME/NODATA errors to be fatal. (until we again spot something which tells me otherwise..) * smtpserver/policytest.c, smtpserver/mxverify.c: If at MAIL FROM or RCPT TO the address domain is not found from the DNS, and code is EAI_NONAME or EAI_NODATA, we yield 500-series result. * include/netdb6.h, libc/getaddrinfo.c: Synced getaddrinfo() API constants with that of Linux glibc 2.1 proper. 1999-12-16 Matti Aarnio * scheduler/prototypes.h, scheduler/threads.c, scheduler/transport.c, scheduler/update.c: Reworked vertex processing some more -- the child feed function (and *only* that) will call pick_next_vertex(). 1999-12-15 Matti Aarnio * scheduler/scheduler.c: In sync_cfps() vertex index limits one-off :-( 1999-12-14 Matti Aarnio * scheduler/scheduler.h, scheduler/prototypes.h, scheduler/update.c, scheduler/transport.c, scheduler/threads.c: Reworked the logic of TA child process feeding and job picking. This is initial edition - not yet confirmed to *work*. Killed a *lot* of hairy code -- thus must have done something right... (Added about 440 lines, removed about 550 lines.) 1999-12-14 Matti Aarnio * configure.in: '--disable-pam' option. * scheduler/*: Pulled back most of logic changes introduced during last 10 or so days, but kept (of course) all SFIO things. * smtpserver/smtpserver.c, smtpserver/smtpcmds.c: changed most (all?) 553 and 453 return codes to 550 and 450 respectively. * smtpserver/zpwmatch.c: Don't call (at all) pam_acct_mgmt() function. 1999-12-12 Matti Aarnio * scheduler/msgerror.c: Put original (or rewritten) message headers to the end of the acompanying letter - makes things easier for M$ users to ask for advice (and me to give it..) * scheduler/scheduler.c: Found curious problems at the resyncing code; sometimes resulting VTX->NGROUP count was zero, and things didn't quite work with that.. Such vertices are now thrown away. * scheduler/scheduler.c: A major walk-thru of MIB variables for Recipient counts. Are they now in control ?? * smtpserver/smtpserver.c, smtpserver/smtpcmds.c: Found out that Lotus SMTPMTA software goes ballistic if it sees response code it doesn't expect -- its writers are stupid and expect everybody to always produce only codes listed at RFC 821 -- without doing the hundred- series "approximative" processing more common in all public source based system.. (beginning with sendmail..) 1999-12-09 Matti Aarnio * transports/libta/ctlopen.c, router/rfc822.c, scheduler/scheduler.c, scheduler/threads.c: Immediately obvious bugfixes, once the system is under load.. 1999-12-08 Matti Aarnio * transports/smtp/smtp.c, transports/mailbox/mailbox.c: Noticed case where still do MIME downgrading in place where we really DO NOT KNOW HOW TO! Added code to once and for all disable conversions unless the message in question is TEXT/PLAIN. * include/mail.h.in, router/rfc822.c, scheduler/scheduler.c, transports/libta/ctlopen.c, include/ta.h: Introduced _CF_FORMAT tag to the scheduler/TA spec file. From now on adding new gimmic formats is easier - I hope. * include/libta/warning.c: Make sure that writing of the messages will happen in blocking mode socket, and do fflush() afterwards! * transports/smtp/smtp.h, transports/smtp/smtp.c, transports/smtp/appendlet.c, include/libc.h: ZBUFSIZ macro to replace possibly originated BUFSIZ macro. Easier to be consistent and predictable.. Added couple missing extern...() function declarations. * lib/esyslib.c: At Solaris the unlink(2) may yield EINTR or EBUSY, do a wrapper to retry the operation when that happens. * router/rfc822.c: At Solaris the unlink(2) may yield EINTR or EBUSY, do a wrapper to retry the operation when that happens. (This is different from eunlink() which complains..) * smtpserver/smtpserver.c, smtpserver/policytest.c, smtpserver/smtpserver.h: Track successfull configuration, and refuse to accept email if no rudimentary configuration exists. 1999-12-04 Matti Aarnio * scheduler: - Some more SFIO fixes which I missed last night - Possibly fixed the 'mailq -Q' reported stored recipient counts. (cosmetics) * lib/symbol.c, libc/getzenv.c: Compiler pleasing.. * ALL OVER THE PLACE: - Replaced *all* stdio usage within the Scheduler with SFIO -- now it can go beyond 255 open file descriptors at Solaris systems... (FILE->_file is 'u_char' there, while Sfio_t->file is a 'short' - huge difference!) - More work on SMTP TA code to eradict last vestiges of alarm(2) based timeout breaks, now SFIO stream discipline routines will handle the timeout gracefully - Some changes to few very commonly used headers, and resulting small call changes all over.. 1999-12-02 Matti Aarnio * proto/smtp-tls.conf, transports/smtp/smtp.c, transports/smtp/smtp.h, transports/smtp/smtptls.c: - Completed initial write-thru of TLS client mode for the SMTP TA, however *no* testing has been done yet! - Both client and server got (from Postfix) session cache, but that code is *untested* * smtpserver/smtptls.c, proto/smtpserver.conf.in, smtpserver/cfgread.c, smtpserver/smtpserver.h, smtpserver/smtpserver.c, smtpserver/smtpcmds.c: - Updates to smtpserver for evolved original code from Postfix. - Both client and server got (from Postfix) session cache, but that code is *untested* * proto/Makefile.in: With "make dirs" creates $POSTOFFICE/TLSsrvrcache/ and $POSTOFFICE/TLSclntcache/ directories with protection of 700 .. (For the SSL session cache storage.) 1999-12-01 Matti Aarnio * utils/policy-builder.sh.in: Add some comments to the generated smtp-policy.dat file about which source section (file) is beginning.. Add some comments to the generated smtp-policy.dat file about which source section (file) is beginning.. * Makefile.in, configure.in, sfio/Makefile.in, sfio/README, sfio/src/lib/sfio/makefile{,.in}, include/ta.h: Autoconfigure key parts of the SFIO library make system to fit ZMailer environment. Use SFIO stuff at Transport Agents. * transports/libta/writeheaders.c, transports/libta/Makefile.in, transports/errormail/errormail.c, transports/hold/hold.c, transports/sm/sm.c, transports/mailbox/mailbox.c, transports/libta/fwriteheaders.c: Renamed former writeheaders() to be fwriteheaders(). * transports/libta/swriteheaders.c, transports/libra/Makefile.in, transports/smtp/Makefile.in, transports/smtp/smtp.c: Created modified version of writeheaders(), namely swriteheaders() which instead of 'FILE *' has 'Sfio_t *' as its 'fp' parameter. Using that at smtp.c * transports/smtp/Makefile.in, transports/smtp/smtp.c, transports/smtp/appendlet.c, transports/smtp/smtp.h: Split up the big monolithic smtp.c a bit: 608 16378 transports/smtp/appendlet.c 4182 112512 transports/smtp/smtp.c 410 12057 transports/smtp/smtp.h Further smarts at the transport-time conversions will go into the appendlet.c alone.. 1999-11-30 Matti Aarnio * router/Makefile.in, router/daemonsub.c, router/functions.c, router/router.c: Moved daemon management functions into own file, and rewrote them somewhat when Linux 2.3.30pre3 caused a hangup.. (And cleaned some dead code from router.c ) * scheduler/mailq.c: Do line-start dot-duplication a'la SMTP... * utils/smtpserver-log-parser.pl: Synced this tool with my lattest online edition, fairly usable.. 1999-11-17 Matti Aarnio * router/functions.c: More system rework, now signal handlers don't muck with stdio stuff, and associated trouble (which *might* be problem at some systems). Report child-process runtime statistics when the child dies. Use select() to detect when children have things to report, if available. A dead router child is reclaimed properly, and new ones are started as the need arises. * scheduler/scheduler.c, scheduler/transport.c, scheduler/update.c: Fix message-specific interactive report logging to use vfp_open() routine (which had a feature/security problem in itself). Rearranged several logging points so that 'scheduler processing' messages are done correctly. Wrote 'Feeding to child..' report into the transport.c. At update.c rewrote logstat(), and dead code now using vfp_open() to write individual recipient statuses; problem being that 'vp->cfp->contents' is NULL at the call time.. * scheduler/transport.c: mailq interface specification had a case where data can be omitted, and one pointer is thus NULL -- referring via which is an open invitation to trouble (SIGSEGV usually..) 1999-11-16 Matti Aarnio * transports/smtp/smtp.c: Noticed that (again) smtp doesn't always timeout in 3 days (or whatever the queue timeout is), turned out EX_TEMPFAIL diagnostics don't always carry timeout value (60 seconds). * router/Makefile.in, router/functions.c, router/router.c, proto/cf/standard.cf: Revised the way how several things are done: - files are processed under control of a work-gang leader - logging isn't done anymore via router logscript "dribble" redirection, but via a work-gang leader 1999-11-15 Matti Aarnio * Makefile.in: version 2.99.52-pre4 * configure.in: Autodetect when PAM-support (for SMTP authentication) needs also "-ldl" library, and when not. * configure.in: Fix on fix on fix of autodetecting MD5SUM program location. * proto/Makefile.in, Makefile.in, proto/post-install.sh.in: An attempt at writing a post-install script, which conditionally moves files from respective "proto" subdirs to their proper places. * scheduler/mailq.pl: A simple demo core for PERL version of the MAILQv2 query utility.. * include/md5-global.h: Found the reason why our MD5 instance did produce wrong results at my Alpha... It is very important to truncate intermediate calculations to 32-bits during the processing! * lib/selfaddrs.c, libident/identuser.c, scheduler/mailq.c, scheduler/pipes.c, scheduler/transport.c, smtpserver/smtpserver.c, transports/mailbox/mailbox.c, transports/smtp/mprobe.c: Instead of AF_xxxx we call socket parameter PF_xxx (not: Address Family, but Protocol Family ...) * man/scheduler.8, man/mailq.1, scheduler/mailq.c, scheduler/threads.c: Updated man-pages a bit (quite a bit!), implemented 'mailq -QQQ' in MAILQv2 mode. (E.g. 'SHOW SNMP') 1999-11-12 Matti Aarnio * configure.in: Autodetect header file * proto/scheduler.conf.in, scheduler/mailq.c, scheduler/mq2auth.c, scheduler/prototypes.h, scheduler/readconfig.c, scheduler/scheduler.c, scheduler/transport.c: * scheduler.conf file can now have: PARAMmailqpath = "UNIX:/path/to/pf_unix/mailq/socket" PARAMmailqpath = "TCP:174" * mailq -p /path/to/pf_unix/mailq/port * mailq -p mailq (check /etc/services for tcp/mailq) * mailq -p 174 (default case) Those are for running multiple schedulers in the system, and allowing each a different mailq port. (Out of PF_UNIX, and PF_INET protocol families..) 1999-11-11 Matti Aarnio * scheduler/mq2.c, scheduler/scheduler.c, scheduler/prototypes.h, scheduler/threads.c, scheduler/readconfig.c, proto/scheduler.conf.in, scheduler/mq2auth.c, proto/scheduler.auth: Modified somewhat commands, and how they are selected. Now "scheduler.auth" file defines which operations are allowed (a bit-mask). * proto/Makefile.in: Wrong file referenced; must be scheduler.auth ! ( not smtpserver.auth! ) * lib/selfaddrs.c: When HAVE_SA_LEN is defined, we used wrong definition for sa_len field within the union. * transports/libta/tasyslog.c, router/rtsyslog.c: An experiment on alternate TRANSPORT syslog format: if ZENV variable SYSLOGFLG has 'T', the classical ", " separated fields are printed. If the flag is 't', field separator is TAB-character. Similarly for ROUTER there are 'R' and 'r' keycodes. (Turns out that either syslogd, or syslog() routine at libc converts all control characters to ^I (in TAB case) pairs.. D'uh...) 1999-11-09 Matti Aarnio * utils/makedb/makedb.c: When using NDBM format, if the STORE returns error with errno=ENOSPC, we report that KEY.LEN+DATA.LEN exceed 1kB which NDBM allows... 1999-11-04 Matti Aarnio * scheduler/readconfig.c, scheduler/threads.c, scheduler/conf.c, scheduler/mailq.c, scheduler/mq2.c, scheduler/mq2auth.c, scheduler/prototypes.h: 'mailq -Q' and 'mailq -QQ' work when scheduler runs with '-M2' now :) Authentication wants PARAMauthfile = "/path/to/scheduler.auth" into the boilerplate defining section of the scheduler.conf, and that file -- well, see the example. WORK IN PROGRESS! * scheduler/mq2.c, scheduler/mq2auth.c: More of the theme... * configure.in, utils/Makefile.in, utils/md5sum.c: Initial attempt at writing (from scratch)a MD5SUM program, but for some reason the result is *not* the same as GNU textutils one gives :-/ Perhaps it will not matter in the end, though.. * scheduler/Makefile.in, scheduler/transport.c, scheduler/mq2.c, scheduler/scheduler.h, scheduler/prototypes.h: Some more work on MAILQ-V2; now needs fleshing out only the real protocol interpreter at mq2.c: mq2interpret(), and corresponding things in the system.. * utils/policy-builder.sh.in: Some script tweaks; removed externally sourced 'bad source addresses' list pickups. 1999-11-02 Matti Aarnio * include/md5.h, libc/md5c.c: Move call-interface renaming macro into more appropriate place.. * proto/cf/server.cf: Rectify report codes for EXPN and VRFY. * smtpserver/smtpserver.c: Don't do ident query when in interactive mode. * smtpserver/smtpserver.c, smtpserver/smtpserver.h: Call s_setup() with appropriate fd's for input, and output. Sometimes we are *not* running with bidirectional socket as the point of contact.. * scheduler/scheduler.h, scheduler/scheduler.c, scheduler/mailq.c, scheduler/prototypes.h, libc/md5c.c, scheduler/MAILQ-V2: Initial steps to write an interactive command channel via the MAILQ port. 1999-10-30 Matti Aarnio * smtpserver/Makefile.in, smtpserver/smtpserver.c, smtpserver/wantconn.c: Reversed the effective kill of the tcp-wrapper after the ./configure got a way not to link it in per default.. * configure.in: "--with-openssl" (or its more explicite cousins) must now be used to have OpenSSL enabled at the system "--with-tcp-wrappers" (or its more explicite cousins) must now be used to have TCP-Wrapper enabled at the system * man/smtpserver.8, proto/smtpserver.conf.in, smtpserver/cfgread.c, smtpserver/smtpserver.h, smtpserver/smtpserver.c, smtpserver/smtpcmds.c: A "tarpit" feature implementation per an idea from Arjan van de Ven , plus Mark Visser . (Who reported of what postfix/vmail can do.) 1999-10-26 Matti Aarnio * configure.in: Smarter (?) default picking for SENDMAILPATH and MAILBOX directory path for those systems which have file with system specific directory paths... * libc/Makefile.in: Fixed $(prefix) usages in 'make install' targets. * smtserver/Makefile.in, smtpserver/smtpserver.c, smtpserver/wantconn.c: Essentially killed tcp-wrapper from being used for ZMailer's smtpserver. SMTPSERVER's policy-database has superior technology for the same thing. 1999-10-20 Matti Aarnio * man/Makefile.in: Make the destination directories if they don't exist.. * router/libdb/bsdbtree.c, router/libdb/bsdhhash.c, utils/makedb/makedb.c: Turns out, 'put' function of Sleepycat DB 2.x has changed the return values completely around, and what before was negative for permanent/system error, is now a positive value, and the other way around... (return value for duplicate key is now negative, while it was positive before...) * smtpserver/smtpcmds.c: Count incoming message SIZE= option only *once*, as we may be relaying them all, and not storing locally. For that matter, in ISP mode a limited message size where the message quota limits maximum of rcptcount * msgsize does not make much sense - nor exactly in local storage mode either... 1999-09-30 Matti Aarnio * smtpserver/smtpcmds.c: For ETRN processing be slightly more carefull at (not) accepting '@' and '#' special characters in the beginning of the domain name. * smtpserver/smtpcmds.c: At MAIL FROM we may reject the email reception, and abort the email -- State change is allowed only if we didn't do the abort.. * transports/smtp/smtp.c: Seen a case where immediately successive EHLO + HELO both fail, and at *same* target that very same HELO also succeeds... Most surprisingly, within same session which just rejected one message with an error, simply continuing with an open stream (sort of bug..) we do succesfully deliver email.. 1999-09-27 Matti Aarnio * scheduler/readconfig.c: If the channel/host selector clause contains wild-cards at the 'channel' part, AND the command= parameter contains '$channel' in it (once or more), set CFG_WITHHOST, because such transport agents can't switch over from one channel to another. * transports/sm/sm.c: - If write to pipe fails (except with 'gotsigpipe'), cleanup PIPEs to the child process. - Don't be upset even if write to pipe fails with EPIPE/SIGPIPE. * router/dateparse.c: Certain pathologic inputs are able to feed extra data to the date parsing, and it turned out to run with month zero, which caused division-by-zero --> core dump... 1999-09-21 Matti Aarnio * Makefile.in: Version 2.99.52-pre3 1999-09-17 Matti Aarnio * smtpserver/smtpserver.c: Don't call whoson(), nor tcp-wrapper libraries in case the smtpserver is not running from a network socket. 1999-09-15 Matti Aarnio * configure.in, smtpserver/smtprouter.c: For some proprietary ZMailer instances we realized that we must pass arbitary environment variables down to the router process run underneath the smtpserver. Ergo, ./configure for putenv() function, and use it if available. * utils/policy-builder.sh.in: Disable completely the old (original) spam-address downloading with Lynx, and its associated tests. Time has passed over them... 1999-09-14 Matti Aarnio * proto/smtpserver.conf.in, proto/zmailer.sh.in, smtpserver/zpwmatch-pipe.c, man/smtpserver.8: Documentation entries from Artur Urbanowicz. * router/libdb/bind.c: Limited one potential buffer-overflow into more sensible sizes of input. 1999-09-10 Matti Aarnio * transports/mailbox/mailbox.c, transports/mailbox/fmtmbox.c, transports/mailbox/mboxpath.c, transports/mailbox/Makefile.in: mboxpath-formatter routine from Eugene Crosser. 1999-09-09 Matti Aarnio * lib/symbol.c: Use CRC-31 polynome hash function internally. * transports/mailbox/mailbox.c, transports/mailbox/mboxpath.c, man/mailbox.8, man/mboxpath.1: Added -X[X] option to behave alike -P[P] option, but to use CRC32() function. * lib/symbol.c, lib/crc32.c, lib/hashtest.c: Moved CRC-31 (?) polynome calculator into separate module outside the symbol.c; Entrypoints: 'long crc32(char *s);' and 'long crc32n(char *s, int len);'. 1999-09-08 Matti Aarnio * libsh/interpret.c, libsh/expand.c, include/libsh.c, libsh/shconfig.h, libsh/zmsh.c: Hunted around, and fixed with a KLUDGE a problem in $(elements ..) expansion routine. I am not *quite* sure it is right one for all cases, but detected problems at online "smtp-router" are cured with this. * proto/cf/server.cf, proto/cf/aliases.cf, README.UPGRADING, SiteConfig.in: - Fix previously scrambled server.cf somewhat - Created global settings controlling two parts of the $(routeuser ..) functionality. * smtpserver/smtprouter.c: Cleaned previous hacks. * smtpserver/smtpserver.c, smtpserver/smtprouter.c, proto/cf/server.cf: Noticed several problems: - Router reports were not reported thru the proper socket, and that screw up things.. - Made some wrappers into the server.cf to always issue proper status code in the line; plus always terminating "250 EXPN done", or "250 VRFY done" lines. Also noticed a bug in the zmsh! - smtpserver main loop got spurious EOF result from protocol input line, when the input was in fact just plain CRLF ... Urgle.. * smtpserver/zpwmatch.c: More rework at the SHADOW -branch.. 1999-09-07 Matti Aarnio * smtpserver/smtpserver.c: If (heaven forbid) s_gets() routine notes that the input line length overflows the buffer -- i.e. it filled the buffer, and didn't get an EOF or NEWLINE yet, then it will glob up input, until it sees a NEWLINE (and including that newline..) * smtpserver/zpwmatch.c: Missed one part in previous HAVE_SHADOW_H related patch :-( * proto/smtpserver.conf.in: Place 'ESMTP' at both lines of the 220 message (sigh, PostFix MTA violates RFC 1869 part 4, and probably isn't the only one...) * smtpserver/smtpserver.c, smtpserver/smtptls.c: Missed one instance of read() from converting to Z_read() wrapper for OpenSSL call support... 1999-09-05 Matti Aarnio * Makefile.in: Version 2.99.52-pre2 1999-09-03 Matti Aarnio * doc/guides/smtp-etrn: Spotted a typo error at the ISP side of the document. * smtpserver/zpwmatch-pipe.c, smtpserver/smtpserver.h, smtpserver/smtpserver.c, smtpserver/smtpauth.c, smtpserver/cfgread.c, proto/smtpserver.conf.in, man/smtpserver.8, smtpserver/Makefile.in: Integrated Artur Urbanowicz's "use external program for authentication" approach into the baseline system. (Documents could use some additional words about the topic..) * smtpserver/zpwmatch.c: Fix HAVE_SHADOW_H branch of the code per report from Eugene Crosser. 1999-09-01 Matti Aarnio * transports/smtp/smtp.c: Error reports from "." phase were improper, cured. 1999-08-31 Matti Aarnio * proto/cf/rrouter.cf: Another missed "$(dequote ...)" call. Without this, pipe/file stores don't get "magic" 'pipe.*' and 'file.*' local channel tags so that all other cases can safely be run to procmail as the local delivery agent. 1999-08-30 Matti Aarnio * Makefile.in: Version 2.99.52-pre1 * sfio/* Looking for ways to use this thing... 1999-08-27 Matti Aarnio * contrib/pipeauth-0.55.tar.gz, contrib/rauth-0.56.tar.gz, proto/smtpserver.conf.in, proto/zmailer.sh.in, smtpserver/cfgread.c, smtpserver/fdstatfs.c, smtpserver/smtpcmds.c, smtpserver/smtpserver.h, smtpserver/zpwmath.c, utils/policybuilder.sh.in, doc/guides/smtp-msa-mode: Heaps of changes from an ISP guy Artur Urbanowicz * transports/libta/diagnostic.c: Report the 'host' part of the referred channel/host pair for which the diagnostic is being presented. 1999-08-26 Matti Aarnio * smtpserver/smtpcmds.c: Fix support for SLOPPY mode "MAIL FROM:<>" source address. 1999-08-24 Matti Aarnio * smtpserver/smtpserver.h, smtpserver/smtpserver.c, smtpserver/smtpcmds.c, smtpserver/smtpdata.c, smtpserver/smtprouter.c: Modify (somewhat) buffering issues; small change at s_setup(), all around change putc() to fputc(). * libc/mail.c: Do setvbuf( _IOFBF ) on freshly opened file descriptor. 1999-08-22 Matti Aarnio * transports/smtp/smtp.c: An odd-ball case of lacking RSET in PIPELINING mode when none of the recipients were successfull.. 1999-08-21 Matti Aarnio * proto/cf/fqdnalias.cf: One last missing usage instance of "runas .. cat ..". * scheduler/msgerror.c: Printout the MESSAGE/DELIVERY-STATUS report to the end of the initial text report - because so many (M$) email clients are uterly unable to display them... 1999-08-18 Matti Aarnio * configure.in: For Solaris 2.6 (and Solaris 7 ?) autodetect -lxnet for socket related functions. * smtpserver/zpwmatch.c, utils/listexpand.c: Some cleanup from result of using SUN-CC 4.2 at Solaris 2.6. 1999-08-17 Matti Aarnio * README.TCP-WRAPPER: New file. * lib/rfc822scan.c: Reorganized the code a bit - now the fast path is straighter code with less odd tests. * transports/mailbox/mailbox.c: getpwnam() failure report fix - should not need, but.. * transports/smtp/smtp.c: Solaris 2.5.1 doing getsockname() is doing blocking streams thing even on non-blocking socket :-/ Results are not as nice as with e.g. BSD based systems where same call doesn't block (no need!) Also some errno saving around. 1999-08-16 Matti Aarnio * proto/cf/aliases.cf: When the ssift (and tsift) were changed *not* to strip away quotes, buggeroo... we need *one* place where quotes are stripped -- that is: explicitely.. * lib/rfc822scan.c: _unfold() had problems at detecting the termination condition. * router/rfc822hdrs.c: One fine-tuning of "Received:" header generation; leaves off one space character when no 'from' clause is present. 1999-08-15 Matti Aarnio * transports/smtp/smtp.c: Detect better, when the STDIO fwrite() and friends fail -- bdat_sync() was running wild loop in odd situation when the recipient system died from under- neath, and when coming back up, it rejected the pre- existing connections... 1999-08-14 Matti Aarnio * smtpserver/smtpserver.c, smtpserver/smtpcmds.c, smtpserver/rfc821scn.c: Done work to get the SLOPPY ('S') mode to work properly with MAIL FROM and RCPT TO contained addresses. * utils/makedb/makedb.c: When a duplicate key is give, print out the key value in same form as the debug mode at the smtpserver does it. * utils/policy-builder.sh.in: Remove previous static blocking list source entries. 1999-08-08 Matti Aarnio * man/smtpserver.8: A bug in the document.. TCP-wrapper service is "smtp-receiver", not "smtp-receive" ... * transports/libta/mimeheaders.c: Previous work had some weird problems, reworked, and now it seems to work. 1999-08-06 Matti Aarnio * transports/libta/mimeheaders.c: MIME-conversion labeling had problems after the new form of 'Received:' production at the router. * smtpserver/smtpserver.c: Added couple of setvbuf() calls for stdout, and stderr at the beginning of the smtpserver main() program. Story is long about what and where... look at the ZMailer list archive about ccNotes gateway problem. 1999-08-03 Matti Aarnio * lib/selfaddrs.c: After example code at the list, did some of my own code tweaks to do comparison of address (IPv4 or IPv6) only, not the entire 'struct sockaddr', which contains ports, and other things... (The port might not necessarily be zero in all cases..) * proto/cf/aliases.cf, proto/cf/aliases-new.cf, router/functions.c: A potential security problem in the scripts. I wonder how the "ancient" 2.2.1 did handle this one.. Introduced "cat" builtin to be used within a sequence of: runas $priv cat "$filepath" ... Now strategically placed symlinks can't be used to divulge file contents whose access is prohibited from the symlink creator.. 1999-07-20 Matti Aarnio * configure.in: Minor nit-pick in study.. Must include at first as in many systems all headers are not idemponent, and thus have some pre-requisite headers for inclusion.. * libsh/interpret.c, proto/cf/canon.cf: Don't do implicite dequotation of quotefull strings in ssift and tsift.. ( '"foo@bar"' -> 'foo@bar' ) 1999-07-19 Matti Aarnio * configure.in: Instead AC_CHECK_HEADER() use AC_CHECK_HEADERS() macro. Former *should* work with single headers, but apparently not.. 1999-07-19 Matti Aarnio * Makefile.in: Version 2.99.51 * configure.in, smtpserver/Makefile.in, smtpserver/zpwmatch.c, doc/guides/smtpauth-login-pam-support, man/smtpserver.8: Autoconfigure PAM support for the system. See the smtpauth-login-pam-support for mandatory bit of support file(s)... * smtpserver/smtpauth.c, smtpserver/zpwmatch.c: Pass also user's UID -- although PAM code does not pick it (yet). * compat/rmail/rmail.c, include/libz.h, lib/linebuffer.c, router/functions.c, router/rfc822.c, utils/makedb/makedb.c: GLIBC 2.1 with __USE_GNU active has also 'getline()' function, thus we renamed to zgetline() ... 1999-07-12 Matti Aarnio * Makefile.in: Version 2.99.51-pre3 * router/rfc822.c: - Empty headers caused problems in printout - At errprint() did some token double-printing, now fixed. 1999-07-08 Matti Aarnio * root/prototypes.h, root/rfc822hdrs.c: Still more finetuning of the RFC-822 output folding rules. Now things look nice in all cases that I have encountered/imagined. 1999-07-07 Matti Aarnio * Makefile.in: Version 2.99.51-pre2 * transports/smtp/smtp.c: - Reworked even more that reconnect mechanism. (core drops) - Noticed that 'No MX, no address' reports are coming out more and more regularly in cases where they should *not* occur. As I don't seem to be able to coerce the system to do the *right* thing, let it be TEMPFAIL once again... Rather delayed erroring attempts, than rejections because of wrong reports. * scheduler/threads.c: Made debugging zsyslog(()) call to use LOG_ERR instead of LOG_EMERG; latter one broadcasts to everybody in one of my test systems... * include/libz.h, lib/rfc822scan.c, lib/token.c, libc/mail.c, router/prototypes.h, router/rfc822.c, router/rfc822.ssl, router/rfc822hdrs.c, router/libdb/header.c: Lots of experimenting with code to get proper RFC822 tokenization, and proper folding at the output. Most difficult part seemed to be rfc822scan.c ... 1999-07-05 Matti Aarnio * transports/smtp/smtp.c: Reworked the EHLO/HELO reconnect mechanisms. Now it reconnects to the *same* address where EHLO failed just previously. Otherwise it could connect to some other address in case the target has several addresses.. * lib/rfc822scan.c: MKERROR() did crash for some reason in previous form. Modified call a bit and changed it into a static function of its own. Now it seems to work. (Compiler bug ??) * Makefile.in: Version 2.99.51-pre1 * smtpserver/smtpcmds.c, smtpserver/smtpserver.h: Limit the length of the "HELO/EHLO" string, and the length of reported reversed hostname in the "rcvdfrom" envelope header, and thus in the resulting "Received:" header. 1999-07-04 Matti Aarnio * router/rfc822.c, router/rfc822hdrs.c, router/libta/header.c: A bit more gymnastics to get rid of: Received: from localhost (user '...' uid#1234) by ... and to replace it with: Received: (from localhost user '...' uid#1234) by ... * transports/smtp/smtp.c: Fixed the errorneous 'Final-Recipient:' lines in the delivery reports. * scheduler/threads.c: Re-enable the kludge-report; if they don't happen, the smtp.c is ok, if they do... * lib/rfc822scan.c, lib/token.c, router/rfc822hdrs.c: Multiline compound tokens (topic for recent activities) had still major problems. Did also modify "Received:" header content generation, although not quite sufficiently in all situations (considering when/how to fold commentary material in 'from' entry.) 1999-07-03 Matti Aarnio * man/smtpserver.8, smtpserver/smtpserver.c, smtpserver/smtpcmds.c, smtpserver/smtpserver.h, smtpserver/smtpdata.c, smtpserver/cfgread.c, proto/smtpserver.conf.in: "PARAM min-availspace 5000" -- specify the minimum required available space in the $POSTOFFICE/ directory. In KILOBYTES. 1999-07-02 Matti Aarnio * smtpserver/smtpserver.c, contrib/whoson-1.08.tgz: Arkadiusz Mi?kiewicz : - update to newer whoson package - adding option "-w" to smtpserver to actually do the query of Whoson. Allows code being compiled in without it being used... * proto/cf/canon.cf, lib/rfc822scan.c: Turns out the best way to fix the 'foo\!faa!fii' problem is not to tinker with tsift patterns at canon.cf, but instead have '\!' pair scanned into token of its own. 1999-07-01 Matti Aarnio * proto/cf/canon.cf: Received a patch to UUCP routing problem introduced by the addition of forbidding to spot '\!' pair.. * include/libz.h, lib/rfc822scan.c: Fix the way how _unfold() works. Pointer comparison was certainly *wrong* way... * router/libdb/ldap.c: Reorganized things: Do connect+bind to DB at ldap_open() Do ONLY queries at ldap_search() Do dissociation at ldap_close() * libsh/builtin.c: A bug in handling of $(lreplace ... ) internal conscell chains polluted wrong objects with call parameters. Adding one strategic copy-chain solved it. 1999-06-30 Matti Aarnio * smtpserver/smtpauth.c, smtpserver/smtpcmds.c, smtpserver/smtphelp.c, smtpserver/smtpserver.h, smtpserver/smtpserver.c: A change in logging; make individual input sessions stand out better by using constructed log line prefix; pppppTTTT[rw#-] where "ppppp" is "%05d" of process pid. "TTTT" are chars in range A thru Z, and are derived from log open time system time- stamp as lowest 4 of base26 "digits" (0 = "A", 25 = "Z"). Now doing "egrep ^pppppTTTT smtpserver.log" should pick only the interesting session... 1999-06-29 Matti Aarnio * router/dateparse.c: Fixes for Y2K-correct date parsing; although this is dead code whose result is ignored, it still apparently gets called... Thanks Ken! * TODO: Found major pain (problem, if you will) in the LDAP interface code for the router. It does server connection and DB bind for *every* query :-( * transports/smtp/smtp.c: Arkadiusz Mi?kiewicz reported compilation and operation problems in his IPv6 environment. Turns out a getnameinfo() request block parametrization bug prevented 'IPv6 AAAA only' hosts from even being *found* for the data lookup. * configure.in, config.h.in, acconfig.h, libc.h: Autoconfigure compilation sense of 'socklen_t' for possible use of --with-ipv6-replacement-libc and then smtp.c in mode where getnameinfo() routine can be debugged... * scheduler/threads.c, scheduler/transports.c: Kludgeish "cure" to ``OF=0, HA=lots..'' problem. If the HA exceeds now fixed value (3600 sec), the system does client shutdown, and then recovers. * lib/rfc822scan.c: Patch from for RFC822 comment nesting. In the end result the comments disappear entirely (not so great) but now properly nested comments are not flagged as an error! 1999-06-27 Matti Aarnio * smtpserver/policytest.c, smtpserver/smtpcmds.c, proto/db/smtp-policy.src, utils/policy-builder.sh.in: Tried to use delayed "rcpt-dns-rbl + test-rcpt-dns-rbl +" policy attributes, and noticed that things didn't work quite the way I had planned . Ah well, some testing, and patching, and now it works... Noticed also that the report out of the machinery used wrong variable in the RCPT TO report code. 1999-06-26 Matti Aarnio * configure.in, ChangeLog, compat/rmail/Makefile.in, compat/rmail/rmail.c, compat/sendmail/Makefile.in, compat/sendmail/sendmail.c, include/listutils.h, include/mailer.h, include/malloc.h, include/zmalloc.h, lib/Makefile.in, lib/splay.c, lib/symbol.c, libc/Makefile.in, libc/mail_alloc.c, libident/Makefile.in, libmalloc/maltrace.c, libmalloc/simumalloc.c, libmalloc/testmalloc.c, libmalloc/teststomp.c, libsh/Makefile.in, libsh/expand.c, libsh/shconfig.h, libsh/trap.c, router/Makefile.in, scheduler/Makefile.in, scheduler/mailq.c, scheduler/scheduler.h, smtpserver/Makefile.in, smtpserver/mxverify.c, smtpserver/smtpserver.h, transports/errormail/Makefile.in, transports/errormail/errormail.c, transports/expirer/Makefile.in, transports/expirer/expirer.c, transports/fuzzyalias/Makefile.in, transports/fuzzyalias/fuzzyalias.c, transports/hold/Makefile.in, transports/hold/hold.c, transports/libta/Makefile.in, transports/libta/ctlopen.c, transports/libta/diagnostic.c, transports/libta/dnsgetrr.c, transports/libta/mimeheaders.c, transports/libta/routermxes.c, transports/libta/tasyslog.c, transports/libta/writeheaders.c, transports/mailbox/Makefile.in, transports/mailbox/mailbox.c, transports/mailbox/mboxpath.c, transports/sm/Makefile.in, transports/sm/sm.c, transports/smtp/Makefile.in, transports/smtp/smtp.c, utils/makedb/Makefile.in, utils/vacation/Makefile.in, utils/vacation/vacation.c: Arkadiusz Mi?kiewicz reported compilation problems in his environment. - About all Makefile.in template files did need changes so that ZMailer's own include paths preceed ALL other - Changed all include instances of "malloc.h" to "zmalloc.h", although that probably is unnecessary... (and renamed the file, although CVS will show deletion and adding..) - Manage to autodetect OpenSSL shared objects, although does not do full blown shared-library runtime load path define, as that detail tends to vary from system to system :-( 1999-06-25 Matti Aarnio * transports/mailbox/mailbox.c, configure.in, acconfig.h: configure time option "--with-mboxquotacheck" 1999-06-24 Matti Aarnio * transports/smtp/smtp.c: More of "mailq -Q | grep OF=0" tracking. Also several cases of eradicting extra fflush() calls. * smtpserver/smtpdata.c: Do "TRANSLATION" facility with PARANOID mode; and some Asperin, or equivalent to Eugene Crosser... 1999-06-23 Matti Aarnio * transports/libta/mimeheaders.c: Do downgrade to QUOTED-PRINTABLE in all cases where it is needed. Convert to C-T-E: 7BIT and/or C-T: ... ;charset=US-ASCII only when the message is TEXT/PLAIN type. * transports/smtp/smtp.c: Change the way how timeouts are handled for the connect(); now does connect() in NON-BLOCKING mode, and uses select() to see when the connection has completed, or timeout occurs. 1999-06-22 Matti Aarnio * transports/sm/sm.c: Reworked the 'm'ultiple recipients processing. The previous patch looped in place and kept allocation more and more memory. Outch.. 1999-06-18 Matti Aarnio * transports/sm/sm.c: Internal users reported that somewhen since 2.99.50s5 the 'm'ultiple recipients mode has been broken. A likely fix is supplied. (I haven't had change to test this.) 1999-06-16 Matti Aarnio * smtpserver/policytest.c: More from Eugene Crosser; process %-hack also for those domains we are a relay-target for... * router/libdb/ndbm.c, router/libdb/ldap.c, router/libdb/dbm.c: Eugene Crosser spotted that I had forgotten to change some close_*() parameters to match new prototypes, which I introduced while debugging BSD DB 2.something at Linux Alpha RedHat 6.0 system. Eugene sent patches for ndbm.c, and ldap.c. * configure.in, smtpserver/Makefile.in, utils/makedb/Makefile.in, utils/vacation/Makefile.in: Introduce new '@LIBLOCALDBMS@' substitution which does not include possible LDAP for applications which *don't* use LDAP in themselves. (Eugene Crosser report about missing @LIBSOCKET@ substitution because of ldap things at makedb..) 1999-06-14 Matti Aarnio * scheduler/transport.c: To sub-process (Transport Agent) pass environment variables: TZ, PATH, ZCONFIG I had thought that syslogd does timestamp formulation, but no... It is within the syslog() library routine where that happens! Thus our childs will need at least TZ (SysV-like systems) to know what the local timezone is :-/ * transports/smtp/smtp.c: Treat "No MX, no A" as a serious error worth discarding the recipient address. 1999-06-13 Matti Aarnio * router/libdb/bsdbtree.c (and others during debug): Spotted strange leakage of 'file' objects in the Linux kernel. Turns out that the BSD DB 2.x at Linux/Alpha glibc-2.1.1 leaks one mmap() page at every db open. Turns out also that every search_btree() which misses, did a retry *after* close_btree() call. As a result, routing did need terrible things. Now won't do that db-reopen at search miss with bsd-btree and bsd-hash codes. * scheduler/scheduler.h, scheduler/transport.c: Spotted a case of 'very long' "host" information, which naturally crashed the scheduler when scribling over fixed buffer in the stack. Now using malloc()ed buffers, which are grown in case a need arises. 1999-06-12 Matti Aarnio * libsh/interpret.c, router/functions.c, router/rfc822.c, router/libdb/bind.c: Code massage to silence the egcs-1.1.2 about "possibly unset variable" -- if there is any code flow before variables are set, then it doesn't know, what has happened. * scheduler/readconfig.c, scheduler/scheduler.c: Removed/moved unused variable. * transports/smtp/smtp.c: Sigh.. "warning: variable `channel' might be clobbered by `longjmp' or `vfork'" * utils/vacation/Makefile.in: [ -l somepath ] is NOT known function, it is: [ -h somepath ] 1999-06-10 Matti Aarnio * README.UPGRADING: Explicitely list all standard commands for different database regenerations. * contrib/zmailstats: RE change in the main syslog pattern rule to handle the timestamp format at the beginning of the syslog line. * libsh/interpret.c: Tracked memory leakage in case the script did following: ssift "$somevar" in (.*)@(.*) return "$somevar" ;; tfiss * router/functions.c: "stableprocess" routine used wrong argv[] element to pass along to the real "process" script. * router/rtsyslog.c: A bit more of compiler pleasing; smaller stack frame when the 8000 byte buffer isn't off the stack.. * smtpserver/policytest.c: Silenced compiler about storing into constant variable. 1999-06-07 Matti Aarnio * lib/selfmatch.c: Revoke the change I did while tinkering with autodetecting address being our local one.. * utils/vacation/Makefile.in: "test -e ..." is perhaps POSIX thing, but not at Solaris /bin/sh :-/ * router/rfc822.c: A case of NULL pointer referral in form which I hadn't encountered before... (fqalias.ldap induced case) * Makefile.in: 'make dist' should dump also symlinks! Boo! * ChangeLog, configure.in, acconfig.h, router/libdb/bsdbtree.c, router/libdb/bsdhash.c, smtpserver/policytest.c, utils/makedb/makedb.c, utils/makedb/dblook.c, README.UPGRADING, utils/vacation/vacation.c: BSD DB 2.* interface had subtle difference; KEY and DATA "DBT"s must be memset(&key,0,sizeof(key)) to avoid nasty surprises... Wow, that API is -- versatile ... * router/libdb/bsdhash.c: Fix of old BSD DB 1.85 type behaviour. 1999-06-06 Matti Aarnio * Makefile.in, ChangeLog: Version 2.99.50-s19 * ChangeLog, utils/vacation/Makefile.in, compat/rmail/Makefile.in, compat/sendmail/Makefile.in, doc/guides/openssl: At places 'make install-bin' complained, review/fixup of some Makefile.in:s. OpenSSL referral document finetuning. * ChangeLog, configure.in, acconfig.h, router/libdb/bsdbtree.c, router/libdb/bsdhash.c, smtpserver/policytest.c, utils/makedb/makedb.c, utils/makedb/dblook.c, README.UPGRADING, utils/vacation/vacation.c: Autoconfigure BSD DB 2.* usage into ZMailer. The resulting databases are BINARY INCOMPATIBLE, and thus must be recompiled when upgrading! * ChangeLog, configure.in, router/Makefile.in, smtpserver/Makefile.in, router/libdb/Makefile.in, utils/makedb/Makefile.in, utils/vacation/Makefile.in: Two overriders: --with-generic-include="-I/path1 -I/path2 .." --with-generic-library="-L/path1 -L/path2 .." 1999-06-03 Matti Aarnio * smtpserver/policytest.c: - Ignore possible 'relaycustomer +' attribute - Autodetect locally active IP interfaces so that a recipient address of will be successfully matched without having that entry in the smtp-policy.mx file.. - Do processing fully, and detect that "foo.fi" is not acceptable target domain... (For RCPT TO, that is..) - Some level of '!' path processing implemented.. * include/libsh.h, libsh/builtins.c, libsh/expand.c, libsh/interpret.c: - Noticed that 'case' statements use squish() function, which has its glob-character matching effectively disabled if 'set -f' mode is active, thus 'case' breaks.. - Reworked quite a bit on how the expand() handles input conscell chain contained strings, and splits them into white-space (per IFS) separated sequences. Now the 'toplevels' variable in the standard router scripts does not split into lots of *huge* strings of which only the initial few chars are used, and then rest are 'truncated' by means of resetting the 'slen' field; no, now only the initial segment of the string is treated that way (no need to copy it), but all of the subsequent ones are copied over into their separate size matched strings. * utils/policy-builder.sh.in: Change a bit comments regarding 'smtp-policy.relay' file content treatment! 1999-06-01 Matti Aarnio * scheduler/Makefile.in: Remove @GETPWLIB@ from scheduler program linkage set. Users used by the scheduler are *only* those in the local system getpwnam() yields! 1999-05-30 Matti Aarnio * configure.in: Store into '#define CONFIGURE_CMD' also CC and CFLAGS used at the compilation. * libsh/expand.c + include/listutils.h, libsh/builtins.c, libsh/interpret.c, libsh/listmalloc.c, libsh/listtrees.c, libsh/variables.c, router/db.c, router/functions.c, router/shliaise.c: After much wondering about profiles (-pg), realized that about 20% of total runtime is spent in fairly complicated recursive list object copying, while the default case of it is a linear list, which should be really simple... Also, there is absolutely *no* reason to copy CAR() branch in most cases! In fact we seem NEVER have need for full TREE copying, just current CDR chain! This turned up to supply router speedup of about factor 1.14 with vger.rutgers.edu's workloads. (about 12 %) expand(), and its subroutines, are still consuming about 16.8% of the total runtime. (860 recipient addresses in the test set, with crossbar() total of 1737 $(router ..) calls, 228k calls of expand() -- after all, one for each function call within the script.) Out of that, perhaps about 7% of the total runtime is amendable to changes without changeing overall system semantics radically... ... for that matter, perhaps for security point of view it could be better to explicitely execute calls that invoke subroutine IO redirection, and shell glob expansion.. With that, the shell script language could be turned into more secure one, and perhaps this 16% could be eliminated entirely ? A historical note from about 1994 shows that router speed for a reasonable test run was about 120 k messages per 24h, or 1.39 recipients per second at a SparcStation-10/50. Syslog monitoring and load factor estimating at vger.rutgers.edu with (on average) 50 recipient messages seems to indicate a speed of about 15-20 recipients per second at similar hardware, although 1) compiler has evolved, 2) workload is different, 3) router internal machinery has had *radical* changes. * include/ta.h, include/libz.h, lib/taspoolid.c, router/prototypes.h, router/rfc822.c, router/rtsyslog.c, scheduler/scheduler.h, scheduler/scheduler.c, scheduler/mailq.c, scheduler/update.c, scheduler/msgerror.c, scheduler/threads.c, smtpserver/smtpdata.c, transports/errormail/errormail.c, transports/libta/tasyslog.c, transports/fuzzyalias/fuzzyalias.c, transports/libta/ctlopen.c, transports/mailbox/mailbox.c: A cascade of changes from wanting to change the basis of taspoolid() generated string.. Now it is parsable back to inode number and input file mtime -- even readable online, although obfuscated... Also renamed lots of variables in the scheduler from 'ctime' to 'mtime', as their function has been for the last few years... * transports/smtp/smtp.c: Cleanup slightly the error report when no MXes are found, *nor* address, and yield INSTANT error. * libsh/interpret.c: fixing indentation, it seems... * proto/db/smtp-policy.src: Per default, don't block even "private internet" networks. 1999-05-28 Matti Aarnio * libsh/expand.c: Remove dupnstr() + freestr() calls by changeing code slightly.. (Circa 5% of runtime in that junk.. For that matter, about 13% of runtime is s_copy_tree() calls by that same expand() function... Better to rewrite it alltogether.) * libsh/interpret.c: Removed unnecessary strsave() calls from within debugging. * router/router.c: With received SIGEXIT/SIGTERM, DON'T die(); this is about getting the profiling data out... 1999-05-27 Matti Aarnio * smtpserver/smtpserver.h, smtpserver/policytest.c, smtpserver/policytest.h, proto/cf/canon.cf: - Implemented "simple" '%'-hack test (also '!'-hack) - Implemented 'message "text string"' attribute handling - Canon code ignores '\!' pair in address, however that actually breaks UUCP bang-path processing slightly: broken\!case!is!like@this working!case\!is!like@this (It is about difficulties of writing an RE for finding the left-most non-quoted bang ...) * libsh/tregexp.c, libsh/interpret.c: A bit less copying around, and way smaller initial malloc to avoid need for big memory blocks, which don't seem to be available for later reuse -- memory fragments, you see.. (At least I think so..) * configure.in: Improve a bit of the testing of vs. , and -ldb1 vs. -ldb -- perhaps this is glibc 2.1 systems thing only, perhaps this is more common, don't know.. * utils/vacation/Makefile.in: When installing under a prefix, create $(prefix)$(VACATIONDIR) in case it doesn't exist before storing anything into it. Also corrected the installed symlink a bit. * transports/smtp/smtp.c: One semi-obscure bug in pipelining related smtp_sync() code yielding false positive delivery report, if the remote did support CHUNKING, and the message was sufficiently small, and the remote closed session before replying to MAIL/RCPT*/BDAT sequence completely... ( read: "small" = under 64 kB, which means almoast all emails ... ) 1999-05-26 Matti Aarnio * scheduler/readconfig.c, scheduler/transport.c: Reworked the handling of command="..." tags in the scheduler.conf file. Now it handles properly the initial literal environment variable settings. * smtpserver/smtpserver.c, scheduler/scheduler.c, include/libz.h, lib/killprev.c: Remove the PIDFILE when the server goes away. (This isn't perfect method, but let it be..) * router/functions.c: Created router function: stableprocess which is a wrapper on process.cf defined process() with cleanup of gensym generated symbols. (Work In Progress; part of my ideas to completely rework router queue...) * router/db.c: Modify cache definitions -- zero them at the local db's * router/rfc822.c, transports/libta/ctlopen.c: Apeace Solaris 2.6 / gcc 2.7.2.3 "-O2 -Wall" compilation * proto/smtpserver.conf.in: Cleanup (a bit) the default settings 1999-05-25 Matti Aarnio * smtpserver/smtpauth.c: Don't log incoming BASE64-encoded passwords. 1999-05-17 Matti Aarnio * smtpserver/smtpdata.c: Be a bit sloppy with DATA statement which follows with white-space.. (or whatnot..) On the other hand, if "-s strict" mode is on, complain! 1999-05-16 Matti Aarnio * transports/mailbox/mailbox.c: Add '-U' option for selectively turning on the 'X-UIDL:' creation while storing to mailboxes and files (but it won't be created while sending to a pipe). * libsh/listmalloc.c: Some ANSI-C present, which the K&R compiler at SunOS 4.1.4 didn't like at all.. 1999-05-15 Matti Aarnio * README.CRYPTO, bin/usa-itar-stripper.sh, Makefile.in, smtpserver/smtpauth.c, smtpserver/smtpcmds.c, smtpserver/smtpserver.c, smtpserver/smtpserver.h, smtpserver/smtptls.c: Modifications and tools to allow package management to be done with command: 'make dist-usa' The resulting TAR file should be US ITAR compliant. * include/listutils.h: K&R compiler reminder; all the world isn't ANSI... * transports/mailbox/mailbox.c: Don't generate X-UIDL: header when sending the message thru a pipe... * transports/smtp/smtp.c: Hunting the reason for SEGV in case ALRM triggers in a few odd situations. * scheduler/transport.c: If the command= line for some channel/host selector tag contains $-expression for ZENV variable, and that variable does not exist, the scheduler did crash.. 1999-05-14 Matti Aarnio * VERSION 2.99.50s18 * smtpserver/smtpdata.c: BASE64 encoding of 7 chars long name yields a BASE64 string with two '=' chars, that was considered bad by the decoder.. ( I think there should be only one '=' char, but Netscape Communicator 4.51 does it that way :-( ) * router/functions.c: Old (very old!) 'old-vs-new' scheduler locking scheme causes problem at Solaris 2.6. Peter IP (Toronto) got ZMailer router process pids 6/7/8/9, which the router treated somewhat poorly.. * configure.in: Autosense OpenSSL at default location, if not otherwise defines. 1999-05-13 Matti Aarnio * router/router.c: Move openlog() a couple lines down so that also interactive routers will use it. * router/rfc822.c: Change the Received: header 'id' information to be the same which syslog() uses, and which is reported at the smtpserver. Eases looking for message spoolids at syslog outputs. * transports/smtp/smtp.c: Overhauled error processing in various situations, now an EX_TEMPFAIL condition will use next MX server (if one is available) very much independent of where along the code the problem occurred. Originally a problem in e.g. connection setup until EHLO/HELO probe was retryable to an MX, but now a much later problem phase (DATA + . yielding 4** code, for example) can do the same. Later continued the theme with logging of the smtp client session to one client with weird looking problem: Ah, finally some clue on the PIPELINING problem; smtp_sync() ignored (or treated poorly) certain way how EXIM responds when it has disk full. We missed the initial MAIL FROM-> 4** code, and errored with the later "503 No sender yet given"... D'uh.. (Likely this was entirely ZMailer problem, but was it only introduced yesterday, or has it been lurking there all along with PIPELING, and never spotted before ??) It manifested as never accumulating queue to sites using PIPELINING, and behaving like above.. * smtpserver/smtpserver.h, smtpserver/cfgread.c, smtpserver/smtpserver.c, smtpserver/smtptls.c, smtpserver/smtpsmds.c, proto/smtpserver.conf.in: Much work again, now to parametrize ways of selectively adding pieces of "Received:" header contained commentary information. Features: ident, whoson, auth-username, tls-cipher, tls-client-cert Added also: "PARAM AUTH-LOGIN-also-without-TLS" which allows paintext "AUTH LOGIN" to be used without successfull STARTTLS. 1999-05-12 Matti Aarnio * smtpserver/smtpserver.c: Fix introduced accidental NULL referral. * SiteConfig.in: Partially undo previous change to SiteConfig.in; let the FORCEPUNT to be there... * scheduler/scheduler.c, scheduler/update.c: Fixed the 'stored recipient count' gauge leakage. 1999-05-10 Matti Aarnio DISTRIBUTION: version 2.99.50s17 * SiteConfig.in, proto/cf/rrouter.cf: Moved "SMARTHOST" into historical bit-bucket. Suggested way is to use "dot-route" to default all non-local traffic to some server. * smtpserver/smtpauth.c: The "ok" reply was wrong (253), and M$ clients barfed at the login. Change that to 235 and it works just fine. * proto/smtpserver.conf.in, doc/guides/openssl, smtpserver/cfgread.c, smtpserver/smtpserver.h, smtpserver/smtpserver.c, smtpserver/smtptls.c, smtpserver/smtpcmds.c: Interoperability testing dig various weird things. Now I know lots of details about the certificates :) For the juicy bits about server certificates, see file: doc/guides/openssl 1999-05-09 Matti Aarnio * Makefile.in: Version 2.99.50s17 * doc/guides/openssl, doc/guides/configure: A bit of updates... * smtpserver/cfgread.c, smtpserver/smtpserver.h, smtpserver/smtpserver.c, smtpserver/smtpcmds.c, smtpserver/smtptls.c, proto/smtpserver.conf.in, doc/guides/openssl: Final hurdles at getting the TLSv1/SSLv[23] support to work at the smtpserver. Used lots of POSTFIX specific code here, and wondered at several document bugs... * here and there...: Some minor things to silence "egcs -Wall -O3 -g" * configure.in, acconfig.h, smtpserver/Makefile.in, smtpserver/smtpserver.h, smtpserver/smtpcmds.c, smtpserver/smtpserver.c, smtpserver/smtptls.c: Configuration options to take into use the OpenSSL TLS code -- although this is not yet complete thing.. 1999-05-08 Matti Aarnio * smtpserver/smtpserver.h, smtpserver/smtpserver.c, smtpserver/smtpcmds.c: Place-holder hooks for 'STARTTLS' related stuff... (once I get the OpenSSL to work at my Alpha ..) 1999-05-07 Matti Aarnio * compat/sendmail/sendmail.c, include/mail.h.in, smtpserver/smtpdata.c: Define _MAILPRIO_HIGH, and sense it in inputs, plus sense "Precedence: high/normal/bulk/junk" to set the mail_priority of zmailer(3) library. * transports/mailbox/mailbox.c: Implemented X-UIDL: header for POP3 UIDL command support; may some day have option which disables creation of that local storage string.. Also, did same thing as sm.c below: * transports/sm/sm.c: Cleaned up (a bit) certain has_header()/delete_header() processing so that all possible multiple instances of same ``header to be deleted'' will be deleted. 1999-05-05 Matti Aarnio * smtpserver/fdstatfs.c: A sidestep to a problem with fstatvfs() at glibc 2.1.1 at Linux/i386/2.1.122 kernel (any 2.2.* kernel, I guess.) * configure.in, smtpserver/Makefile.in: Autoconfigure "-lcrypt" library for the use at @GETPWLIB@ cases. Also have '--with-privateauth' option for the smtpserver in case some magic private things are desired (it is possible..) * smtpserver/smtpauth.c, smtpserver/zpwmatch.c, smtpserver/cfgread.c, smtpserver/Makefile.in, man/smtpserver.8: Separate zpwmatch() function into its own file. Have 'PARAM smtp-auth' configuration parameter (and document it at the smtpserver man-page) which enables use of SMTP AUTH. The default system uses classical UNIX password analysis. (but see the comments at zpwmatch.c!) 1999-04-26 Matti Aarnio * Makefile.in: Version 2.99.50-s16 * proto/cf/fqdnaliases.cf: Typo: Missing ending ']' at a '[' test clause. * smtpserver/policytest.c: Policytest should accept everything when running in non-network interactive mode. (Ok, Syntax issues aside..) * smtpserver/smtprouter.c, smtpserver/cfgread.c, smtpserver/smtpserver.h, smtpserver/smtpserver.c, proto/smtpserver.conf.in: Supply "PARAM enable-router" parameter to control the availability of interactive router subsystem. Also strengthened the way how the user input is passed to the interactive router so that it should not allow bad inputs to cause escape into shell at any moment.. However this isn't the only place where the danger may lurk, so per default the "enable-router" is OFF. 1999-04-25 Matti Aarnio * smtpserver/smtpcmds.c, smtpserver/smtpserver.c,smtpserver/smtpdata.c: Change the 'multilinereplies' processing around "a bit". Have sensible reports even when single-line mode is used. * proto/smtpserver.conf.in, man/smtpserver.8, smtpserver/cfgread.c, smtpserver/smtpserver.h, smtpserver/smtpdata.c, smtpserver/smtpserver.c, smtpserver/smtpauth.c, smtpserver/Makefile.in, smtpserver/smtpcmds.c, smtpserver/policytest.c, smtpserver/policytest.h: Created base work for ESMTP "AUTH LOGIN ..." command processing. It does receive the input parameter, and decodes the initial input, plus asks for the password, plus decodes that, but then the final hurdle of checking the password is not completed -- all the kwazillion ways of doing it in different systems ... THIS EXTENSION IS CURRENTLY NOT ENABLED IN THE SYSTEM! * router/rfc822.c: Converted from symbol() to symbol_lookup_db() in place where the data to be looked up is stored in specific spl set. * include/ta.h, scheduler/mailq.c, scheduler/msgerror.c, scheduler/prototypes.h, scheduler/qprint.c, scheduler/scheduler.c, scheduler/scheduler.h, transports/libta/ctlopen.c, transports/libta/lockaddr.c: I decided that there is no point in using 'long' to store offsets to transport specification file -- if that file exceeds 2G in 32-bit systems we are in deep trouble anyway -- and there is no point in trying to fit this animal into 16-bit machines... Could squeeze around 16 bytes per recipient in the scheduler memory use. (Consider having a million recipients in memory data...) * smtpserver/smtpserver.c, smtpserver/smtpserver.h, smtpserver/cfgread.c, smtpserver/smtpcmds.c, proto/smtpserver.conf.in: "PARAM no-multiline-replies" which hides away all RFC-821 Appendix E described multiline replies, in case you have no other way to fix your non-conformant clients... The result is often rather jumbled as the last reported line usually isn't very usefull.. * include/mail.h.in, include/ta.h, transports/libta/ctlopen.c, scheduler/scheduler.c: Reserve some space into "r" entries of the TA files for delay state reports. * scheduler/scheduler.c: Don't unnecessarily unlink() ETRN requests twice.. 1999-04-21 Matti Aarnio * ChangeLog: Log these last changes.. * include/listutils.h: A surprising macro in FreeBSD/OpenBSD clashed with s_push() gcc extended macro code contained variable name (_X), replaced with another name (_tmpX), which hopefully doesn't clash... * scheduler/threads.c: Change 'ulong' type to 'u_long', which should please also FreeBSD and OpenBSD.. * configure.in: Modify tests for , and resolver so that systems where all headers are not idemponent will be able to do the test compiles successfully, and thus avoiding false negatives.. * smtpserver/mxverify.c: Changed IPv6 RBL tests to use "x.x.x....x.x.ip6.rbl.maps.vix.com" domain format. Those small digits at the top of the address range without "ip6" are clashing with various IPv4 "banned" address spaces... 1999-04-14 Matti Aarnio * proto/cf/fqdnaliases.cf: Create fqdnalias: foo@bar: foo@bar resolve it.. Without this change, the router will recurse infinitely and crash. With this, the router will break the loop, but the result will still be stupid: smtp bar foo@bar ... * smtpserver/mxverify.c: Fatal RBL lookup error -- caused effectively all instances to try for IPv6 lookups which failed consistently. * Makefile, Makefile.in: Release 2.99.50s15 (to have a clean point) 1999-04-12 Matti Aarnio * include/listutils.h: Small macro change to please gcc 2.7.2.1 at FreeBSD 3.1; but perhaps it is due to some unfortunate #define overwriting my macro contents somewhere else... I am listening on the reports of the issue. 1999-04-10 Matti Aarnio * smtpserver/policytest.h, smtpserver/policytest.c, smtpserver/mxverify.c: A surprise was found in pure IPv6 operational environment regarding RBL policy testing. Changed things to also do IPv6 address RBL lookups in case the connection was not by IPv4-MAPPED addresses. (E.g. Lookup for A with zone in same manner as IP6.INT reversers construct it from the address, and then just add 'rbl.maps.vix.com' to the reverse address string in similar manner as IPv4 lookup works.) Of course I don't know what exact format the RBL will use for IPv6 addresses, my quesses are: b.3.b.7.5.7.e.f.f.f.8.f.0.0.2.0.0.0.e.f.2.0.0.0.0.1.6.2.e.f.f.3.rbl.maps.vix.com. b.3.b.7.5.7.e.f.f.f.8.f.0.0.2.0.0.0.e.f.2.0.0.0.0.1.6.2.e.f.f.3.ip6.rbl.maps.vix.com. and the implemented thing is the first one of these. The address in question is following: mea.ipv6.tmt.tele.fi IN AAAA 3ffe:2610:2:fe00:200:f8ff:fe75:7b3b * transports/mailbox/mailbox.c: Another way how getpwnam() surprised -- errno value isn't all zero in Linux with glibc 2.1 in all conditions; username contained multiple dots ("a.b.c") yielded errno value ENOENT :-O (At OSF/1 EINVAL can be had for valid lookup without found entry...) Sigh. * include/shmmib.h, scheduler/scheduler.c, scheduler/scheduler.h, scheduler/threads.c, scheduler/update.c: Modified monitoring MIB set a bit for the uses of the scheduler. Made a line in "mailq -Q" to show Counter32, and Gauge32 type variables. (On the way towards the SNMP MIB..) 1999-04-06 Matti Aarnio * libsh/builtins.c: $(elements ...) tinkering; code "cleanup". (This wasn't the real problem..) * include/listutils.h: Test of ``LISTness'' vs. ``STRINGness'' became a bit confused with ELEMENT flag set. Mask that off when comparing. (THIS was the problem that plagued $(elements ...) ) * libsh/execute.c: Just small cleanup. * libsh/expand.c: glob() expansion list construction straightening. Pre-Sort items in reverse, and the list construction will be easy.. * libsh/interpret.c: functype() Remember always to set the return value variable -- if its pointer is available, otherwise some things behave strangely depening on what happens to be at the stack frame... ( + removed some dead code from the interpret() proper.) * libsh/listtrees.c: _s_copy_tree() Added a comment (for myself) to explain things a bit. * proto/cf/server.cf: Some extra tinkerings for those who want interactive router at their smtpserver. * smtpserver/policytest.c: _addrtest_() Selectively test blanket allowance/denial things, so that when e.g. MAIL FROM: the literal IP address can be looked up from the RBL/DUL/whatnot database while not allowing e.g. MAIL FROM to suddently allow input from any possible host just because once a suitable address is given in MAIL FROM data... * smtpserver/smtpchild.c: Count the number of childs properly, and always remember to return a value thru the childcntp pointer. 1999-04-03 Matti Aarnio * Makefile.in: 2.99.50-s14 * proto/cf/aliases.cf: Fixed a typo, which doesn't affect most people at all.. * scheduler/transport.c: Allow the command="..." to define at first some environment variables, before starting the program. Syntax something like: command="VAR1=value1 VAR2=value2 command -params" No white-space is placeable into the values, all expansions that work for the command parameters will also work at values. * transports/mailbox/mailbox.c: A strange DEC Tru64 (a.k.a. DEC OSF/1 a.k.a DEC UNIX) feature in getpwnam(), which seems to return errno==EINVAL, when no entry is found at /etc/passwd -- getpwnam() return value is NULL, of course. * router/db.c: Changed the way how D_db ("trace db") works; shows the query- in-progress proceeding. * router/router.c: Turn stdout/stderr buffering from NON-BUF to LINE-BUF mode. * proto/db/smtp-policy.src: Removed "relaycustomer +" from the "_full_rights" set. * man/Makefile.in, man/smtpserver.8, proto/smtpserver.conf.in, smtpserver/cfgread.c, smtpserver/smtpchilc.c, smtpserver/smtpcmds.c, smtpserver/smtpserver.c, smtpserver/smtpserver.h: - added new PARAMs to disable selected facilities - added new PARAMS to tune a few other facilities - allows limiting the number of parallel smtpserver processes - updated smtpserver.conf.in sample file - updated man-page * Makefile.in: 2.99.50-s13 * include/listutils.h, libsh/builtins.c, libsh/execute.c, libsh/sh.h, libsh/expand.c, libsh/interpret.c, libsh/listmalloc.c, libsh/trap.c, libsh/listtrees.c, libsh/listutils.c, libsh/tregexp.c, libsh/zmsh.c, libsh/variables.c, router/db.c, router/functions.c, router/rfc822.c, router/router.c, router/router.c, router/shliaise.c, include/libsh.h, router/libdb/bind.c, router/libdb/bsdbtree.c, router/libdb/bsdhash.c, router/libdb/dbm.c, router/libdb/gdbm.c, router/libdb/header.c, router/libdb/hostsfile.c, router/libdb/incore.c, router/libdb/ldap.c, router/libdb/ndbm.c, router/libdb/ordered.c, router/libdb/yp.c, router/libdb/unordered.c: A MASSIVE rewrite of internal conscell processing facilities. * smtpserver/cfgread.c, smtpserver/smtpserver.c: Accept PARAM keywords even if they are preceded with some white-space at the beginning of the line. * router/rfc822.c: Modify privilege data verification; accept also negative values. * transports/libta/tasyslog.c, utils/listexpand.c, smtpserver/smtpserver.h: Minor things to apeact some compilation environments * compat/sendmail/sendmail.c, configure.in: Use mkstemp() function, if it is available * smtpserver/smtpserver.c: Utter strictness in local mode is not so good.. Turn the strictness selectively off in local mode. * Makefile.in: Incremented version number * proto/cf/server.cf: Accept "MAIL FROM:<>" from the client. * INSTALL: Added a note about UUCICO/uucp uids, and 'zmailer'-group * transports/mailbox/mailbox.c: Moved "running as root test down a bit, changed slightly some debug related diagnostic messages. * transports/libra/diagnostic.c: If scheduler diagnostics are suppressed, suppress also syslog messages. * transports/libta/tasyslog.c: Allow a bit longer diagnostic string to be sent... * */Makefile.in, */*/Makefile.in: Change 'libz.a' to 'libzm.a' -- until somebody uses thus named library in some other system, and we can't sidestep it... * smtpserver/policytest.c, smtpserver/smtprouter.c: Fix debug related buffer overflow in policytest.c (debug messages scribled over memory -- no user data at the writeout) add a bit more debug things into the smtprouter.c so that we can see what is sent to the router process. * smtpserver/policytest.c, smtpserver/policytest.h, smtpserver/mxverify.c, proto/db/smtp-policy.src: Updated commentary documents a bit at smtp-policy.src file. Made policytest stuff to use arbitary size strings as options, and pass such to rbl_dns_test() function as a colon-separated list of domain suffixes for where to query for RBL-like data. * transports/mailbox/mailbox.c: Hunting an ellusive problem reporting [exit status 0] and not knowing which branch of the thing it really is from. Also matched UTIME/UTIMES test order with other instances in the same code. * transports/sm/sm.c, transports/smtp/smtp.c, utils/pop-proxy/pop-proxy.c, scheduler/scheduler.c: Moving arount the setvbuf() calls a bit, probably mostly safe without this, but better be absolutely sure.. * libc/md5c.c: Cleanup of 32- vs. 64-bit platform portability issues. * transports/libta/mimeheaders.c, transports/mailbox/mailbox.c, transports/smtp/smtp.c, transports/sm/sm.c, include/ta.h: A bug in how 8-bit containing body was downgraded to Quoted-Printable when the message in question was in fact NOT TEXT/PLAIN ! Aargh!.. (MULTIPART/ZZZ things..) * transports/smtp/smtp.c: A Q-P encoding change from Arnt Hulbrandsen * transports/mailbox/mailbox.c, router/rfc822.c, lib/selfaddrs.c: Adaptation to match some glibc 2.1 beta compability issues. * proto/cf/crossbar.cf * utils/listexpand.c Modernization work * proto/smtpserver.conf.in Per default turn off 'DEBUG' command * contrib/whoson-protocol.txt Protocol document (draft) * INSTALL: Added a word about scheduler.conf file. * doc/guides/smtp-etrn, doc/guides/offline-operation Generalized guide on ETRN usage with ZMailer, and one special case.. 1998-11-19 Matti Aarnio * Makefile.in: Upped version number into 2.99.50-s11 * utils/vacation/vacation.c: Fixes for DB(M) handling in the utility, also automated the installation procedure into userspace a bit. * configure.in, router/libdb/bsdbtree.c: A bit of experimenting with BSD DB 2.*, which apparently needs major rework to operate properly. Using if it is available in place of .. * libc/strsignal.c: FreeBSD 3.0 portability problem fix * transports/smtp/smtp.c: - reworked on how much of the message has been transported over the connection at the time of the timeout/whatever failure in the DATA-phase. - report MX-lookups at 'ps' too - setjmp() nesting fixes * configure.in, proto/newdb.in proto/Makefile.in: - Autoconfigure the path of 'true' into newdb script - FreeBSD reported problems about shell variable substitution during the install phase * smtpserver/smtpdata.c, smtpserver/smtpserver.c, smtpserver/smtpserver.h, smtpserver/smtpcmds.c: - Extra-vigorous protocol syntax checking mode * smtpserver/policytest.c: Policytesting failed with 'case 2' of the boilerplate config. * man/mailbox.8, transports/mailbox/mailbox.c: - Automatic /bin/sh-less execve() for pipe subprograms which falls back to /bin/sh-full in case some parts of the input string prevent the use of the pure mode. (like '>' and '$' characters in the cmdline.) - pass our locally generated recipient-specific environment into the execve() every time.. - updated man-page about the subprogram execution environment * smtpserver/cfgread.c, smtpserver/smtpserver.c, smtpserver/smtpserver.h, proto/smtpserver.conf.in: The example tells more (in smtpserver.conf) # # HDR220 metatags: # %% -- '%' character # %H -- SS->myhostname # %I -- '+IDENT' if 'identflg' is set # %V -- VersionNumb # %T -- curtime string # %X -- xlatelang parameter # PARAM hdr220 %H ZMailer ESMTP-server %V running at ... PARAM hdr220 %H (NO UCE)(NO UBE) local time is now %T # Also did a bit of cleanup * SiteConfig.in: Added a couple of previously undocumented ZENV variables, which are used in the binary code in some places. * lib/nobody.c, lib/selfaddrs.c, lib/trusted.c: Some setreuid() code changes pulled back from way back. SunOS 4.* didn't consider "(uid_t)-1" to be the same as "-1" ... * smtpserver/smtpserver.c: Missed the call of settrusteduser() in one codepath. 1998-10-10 Matti Aarnio * configure.in (and configure), utils/vacation/Makefile.in: Autosense possible location of 'vacation'. If none is found, assume "/usr/bin/". At 'make install', install a symlink at that user-space location to point into MAILBIN/vacation.sh * utils/vacation/vacation.c: Cleanup of database open/close routines after seeing one case of CORE-dump at nic.funet.fi when a new user took this vacation into use.. * Makefile.in: Version 2.99.50-s10 * transports/mailbox/mailbox.c: - Pass "TZ" environment variable to subprocess - Lots of playing with 'sieve' interface * transports/smtp/smtp.c: Allow address-literals of form [addrlit]:portnum, that is you can define explicite route of style: sometarget smtp![ip.addr.lit]:5678 which will then connect to given host and port. The default port is 25. * proto/newdb.in: "while true" does not work everywhere, use: "while /bin/true" * proto/cf/rrouter.cf: Support for channel-group 'smtpgw*' with its own peculiar rules. Originally Sonera internal stuff. * configure.in (and configure): "--with-privatembox" option added * packaging/solaris/*: Solaris 'pkgadd' tools * scheduler/scheduler.c, scheduler/scheduler.h, scheduler/threads.c, scheduler/transport.c, scheduler/servauth.h, scheduler/servauth.c: Two new files (servauth.[ch]), and changes elsewere to support them. Incomplete. 1998-08-14 Matti Aarnio * man/mailbox.8, doc/manual/zref-transport-agents.tex: nroff formatting corrections, spelling fixes * man/router.8, router/router.c, router/rfc822.c, router/prototypes.h, router/rfc822hdr.c: Router got new commandline option: '-W' Existence of that option activates OLD behaviour of the router where it writes 'Illegal-Object:' headers to the file... The new behaviour passes the faulty lines unmodified; "as is". * utils/policy-builder.sh.in: Modifications to handle non-clean spam-domain input better. 1998-08-09 Matti Aarnio * transports/smtp/smtp.c: While reviewing new draft-ietf-drums-smtpupd-08.txt, I found out that occasionally some SMTP servers do reply just '250\r\n', which would confuse PIPELINING mode codepaths, very least. (Although with the ENHANCEDSTATUSCODES that is an unlikely occurrence..) 1998-08-03 Matti Aarnio * include/malloc.h, include/zsyslog.h, libsh/builtins.c, libsh/test.c, router/libdb/header.c, libsh/shconfig.h, lib/strmatch.c, lib/pjwhash32.c, lib/selfaddrs.c, libc/fullname.c, include/md5.h, libc/__fopen.c, libc/md5c.c, libc/strtoul.c, libc/gai_strerror.c, libc/getaddrinfo.c, smtpserver/smtpserver.h, smtpserver/smtpserver.c: A burst of activity to fix compilation at SunOS 4.1.* systems with the system's K&R compiler. (Sigh what a moron for a compiler... Was the pre ANSI-C era really that awfull ???) (This is not complete work, more needs to be done before it all goes thru..) * utils/policy-builder.sh.in: Add use of '-A' (append) to the policy-building parameters. * proto/scheduler.conf.in: Add some comments into the beginning concerning resource usage limitation parameters: maxta, maxchannel, maxring * include/zmsignal.h: Change name of variables from 'sigmask' to 'sigmsk'. SunOS 4.1.* defines 'sigmask' in its as a MACRO... 1998-07-30 Matti Aarnio * Makefile.in, compat/Makefile.in, compat/rmail/Makefile.in, compat/sendmail/Makefile.in, proto/Makefile.in, transports/Makefile.in, utils/Makefile.in, utils/makedb/Makefile.in, utils/vacation/Makefile.in: - Adding quite many "@" characters at critical places to SILENCE the "make install" process so that people can have it easier to SEE when something generates error reports, or otherwise wants attention - Those places which do rather unconditionally store new files into $MAILSHARE/ datasets use following new rule: Don't overwrite existing files (scripts especially) if they ``cmp'' the same as our current file. If they are different, move old to backup, and then write new one there * smtpserver/mxverify.c: When compiling on SunOS 4 the T_TXT definition does not exist in system includes, define it ourselves. * router/libdb/unordered.c: ("Matthias Urlichs" ) An error in open routine caused lookup always to fail on the FIRST lookup. Oops... * include/libsh.h: Had ANSI prototype there, and a K&R compiler "barfed"... 1998-07-28 Matti Aarnio * proto/Makefile.in: (Mark Visser) Missed " ; \ " from the middle of "make db" sequence. 1998-07-27 Matti Aarnio * Makefile.in: Version number: 2.99.50-s8 * configure.in, acconfig.h: Autoconfigure Eugene Crosser's whoson system into the ZMailer smtpserver. * configure.in, proto/sm.conf.in: Autoconfigure ``procmail'' path. * router/libdb/Makefile.in: Use of "::" on dependency tupe caused unnecessary recompilation of router/libdb/*.c files. 1998-07-26 Matti Aarnio * smtpserver/policytest.c, smtpserver/mxverify.c: Prefix all debug-mode outputs with '000-', which (perhaps) will make it usable even during normal SMTP server session. * transports/mailbox/Makefile.in: At "make clean" do remove 'mboxpath' binary too... * router/lib/db/ordered.c: Removed "reopen:" goto label errorneously in the previous "compiler pleasing" round as "unused label".. Sat Jul 25 21:34:11 1998 Matti Aarnio * proto/db/smtp-policy.relay (new): Added a default boilerplate edition of this file into the package (all comment -- but a lot of it..) * proto/db/smtp-policy.src: Edited the comments a bit * proto/Makefile.in: Conditional install in 'make db' the above created smtp-policy.relay file. 1998-07-13 Matti Aarnio * utils/policy-builder.sh.in: Mark Visser reported about the clumsiness of the spam-list retrieve and merge code. His patch went in almoast as is. * aclocal.m4, transports/mailbox/mailbox.c: Debian hits again; I don't know which version this is about, but some recent ones have header, and library libfilelock.{a,so} with similar code to that of the Solaris libmail.a's maillock() stuff. * proto/cf/rrouter.cf: Move the "new localpart processing" a bit downwards so that "pipe.*" and "file.*" tricks used with the scheduler configs still work. 1998-07-08 Matti Aarnio * */*.h */*.c */*/*.c: (all over the place) Did use "egcs -Wall -W -W***" to get lots of warnings. Tried to fix most of them. (later heard that the CVS dumpped source is a bit broken in Solaris 2.5.1 ... will have to return to this...) * transports/mailbox/mailbox.c, transports/sm/sm.c: Internally the append_header() routine has a buffer of 2kB size. Under no circumstances allow that to be exceeded, thus use patterns like: "%s.999s" while formatting something. 1998-07-02 Matti Aarnio * transports/smtp/smtp.c: An error report from Serge Lozovsky did show that in non-pipeline mode the smtp-ta did errorneously consider successfull a delivery to remote system that has rejected the only (or all?) recipient address(es). 1998-07-01 Matti Aarnio * transports/mailbox/mailbox.c: Lock acquisition code had "faulty" way to unrolling itself. I begun to suspect this while my ELM at rare times complained that it can't remove the .lock file it constructed itself :-O Turns out, the lock unrolling did unroll also the failed lock -- in case of "dot-lock" that is the "mbox.lock" file.. 1998-06-29 Matti Aarnio * man/scheduler.8, man/smtpserver.8: Added .SH paragraphs about TCP-WRAPPER into them. 1998-06-27 Matti Aarnio * TODO, include/libsh.h, include/listutils.h, libsh/builtins.c, libsh/expand.c, libsh/interpret.c, libsh/listutils.c, libsh/regex.h, router/funcsions.c, router/shliaise.c: Conditional removal of code handling "prev" pointer for "setf" function uses at the conscell objects. Default is NOT to remove them (yet), although experimental systems running code without any "setf" anywhere seem to be successfull, and even reduce the router memory footprint by 2-3 % ... 1998-06-26 Matti Aarnio * obsolete/* Removed this tree from the CVS, perhaps soon also from the distribution tar files... 1998-06-25 Matti Aarnio * libc/md5c.c: Some fine-tuning added so that the module will compile at most platforms, but perhaps linking will cause indigection at some systems still (Solaris -lnsl has md5 routines in...) 1998-06-23 Matti Aarnio * smtpserver/policytest.c: For a "always_accept" sender, turn DNS based rejections (NXDOMAIN errors) to be hard errors. * smtpserver/smtpcmds.c: A mistake in recipient policy analysis turned soft DNS error into hard policy rejection... Outch. 1998-06-22 Matti Aarnio * smtpserver/Makefile.in, smtpserver/smtpserver.h, smtpserver/libwhoson.c, smtpserver/whoson.h, smtpserver/whosond.c: Initial hooks/codes for "who-is-on" daemon. * include/splay.h, lib/pjwhash32.c, lib/symbol.c: Created symbol-routines which handle arbitary binary key values; key hashes are a bit simple- minded, though... * scheduler/scheduler.c, man/scheduler.8: Runtime option '-n' -- see the man-page. * include/policy.c, smtpserver/readpolicy.c, smtpserver/policytest.c, smtpserver/policytest.h: Added 'trustrecipients' attribute. * smtpserver/rfc821scn.c: Additional texts to match several error cases that I keep seeing around at SONERA's systems.. * doc/guides/smtp-policy: Updated the contents to match system a bit more * include/md5.h, include/md5-global.h, libc/md5c.c: Included generic MD5 code into the ZMailer's libc. To be used in a few server applications... (including APOP-like session authentication for scheduler cmd interface..) 1998-06-16 Matti Aarnio * Makefile.in: PATCHLEVEL 50-s6 * configure.in, nearly all Makefile.in files: Configure option: --with-generic-include="-I/..." which is added into ALL compilation statements 1998-06-15 Matti Aarnio * smtpserver/policytest.c: Moved "P_A_FullTrustNet" attribute recognition up a bit so that "P_A_RelayCustNet" does not blank full_trust mode! * utils/policy-builder.sh.in: Added another retrievable spam domain list * proto/cf/aliases.cf: A fragment of script (in comment) which is specific for listserv.funet.fi system. * scheduler/msgerror.c: When reporting "Multiple-fault:" type error, tell also which template file is not openable, and ask the recipient to report the case to the postmaster... * router/functions.c: - rd_doit() -- don't rename "core*" file into normal spooled file name - run_listexpand(): Did complain when it got 5 arguments and refused to do the work :-( 1998-06-14 Matti Aarnio * proto/cf/aliases.cf, proto/cf/fqdnaliases.cf: Match new listexpand parameters. * router/functions.c: Added two new parameters to function listexpand to match underlying rrouter script function. * libsh/regex.h, libsh/regex.c: Matched against GLIBC type defines. * transports/smtp/smtp.c: A screwup in DBAT dot-processing algorithm. Basically BDAT phase did line-start dot-duplication, which is all right for DATA phase, but a no-no for the BDAT :-( 1998-06-13 Matti Aarnio * proto/cf/aliases.cf: Removing "sender" parameters from alias processing cases. Copying them from other codes was a mistake with subtle results. * proto/newaliases.in: Parameter parsing error; didn't honour '-s' option. 1998-06-12 Matti Aarnio * smtpserver/smtpserver.c, smtpserver/smtpserver.h, smtpserver/smtpcmds.c: Reviewed, and changed quite many of the ENHANCEDSTATUSCODES code values presented in replies. Not all DNS errors are POLICY (*.7.*) errors although policy analysis detects them.. 1998-06-11 Matti Aarnio * proto/makedb.in: Altered the parameter pickup routines to be more generic in view of what *sh shells support: while true; do .. break .. done * transports/mailbox/mailbox.c: Mistakes after code movements in program() function :-( * smtpserver/smtpdata.c, smtpserver/smtpserver.c, smtpserver/smtpcmds.c, utils/smtpserver-log-parser.pl: Make the verbose debug log more readable -- guaranteed session end-points to be findable: ######w \t 221 ... ######- \t ... * smtpserver/smtpcmds.c: Change ETRN reply texts a bit; don't anymore speak of TURNME.. 1998-06-09 Matti Aarnio * proto/cf/aliases.cf, proto/cf/fqdnaliases.cf, proto/cf/p-usenet.cf, proto/cf/rrouter.cf, proto/cf/standard.cf, proto/db/aliases.zmsh.example: Major reorganisation of scripts; - All things that have something to do with local users are clustered into aliases.cf. (Away from standard.cf, and from rrouter.cf) - A new calling arguments for routeuser (aliases.cf, and rrouter.cf) - Smart "+plustail" processing into fqdnaliases.cf - Hooks in between p-usenet.cf, and aliases.cf to handle 'newsgroup' database. With these the system can handle usernames that have dots in them, and unifying multiple (in theory: arbitary number of: aliases.zmsh.example) aliases, and other databases. 1998-06-08 Matti Aarnio * include/policy.h, smtpserver/policytest.c, smtpserver/policytest.h, smtpserver/readpolicy.c: Altered "relaycustnet" semantics a bit. Even though it will still allow unlimited inbound feed, it will at first fo DNS verification on MAIL FROM, and RCPT TO headers -- accepts recipients if said recipients have any DNS A/MX entry at all, and likewise with sender... Introduced new attribute: "fulltrustnet +" which can be used in case there absolutely is need for not checking input addresses (like in a very high-volume email traffic in between vger.rutgers.edu and nic.funet.fi ...) * transports/mailbox/sieve.c, transports/mailbox/mailbox.c, include/sieve.h, transports/mailbox/Makefile.in: Hooks for 'SIEVE' things; ways for the user to define message arrival time filtering. 1998-06-06 Matti Aarnio * contrib/spamlist.py, utils/policy-builder.sh.in: New version of spamlist.py from Roy Bixler, a bit support for its output format into policy-builder. 1998-06-05 Matti Aarnio * transports/sm/sm.c: Added "H" flag to the "sm" facility flags. Altered here the way the BSMTP mode has worked for ages. Specifically now it requires EXPLICITE 'H' flag to produce HELO/EHLO to the begining of the BSMTP datastream. * utils/rotate-logs.sh.in, configure.in, utils/Makefile.in: Made "rotate-logs" utility to be 1) autoconfig, and 2) installed every time a "make install" is done. * smtpserver/smtpserver.c: Some clients seem to send "QUIT\r", and then close the socket. This caused indigestion and error logging. Now it will at least tell what the input was, and then have indigestion complaints.. * proto/zmailer.sh.in, proto/Makefile.in, proto/newdb.in: - Alterations on "zmailer" command, streamlining, adding a "zmailer makedb" function to compile all common database. - Option processing improvements on newdb to allow arbitary order of options affecting the underlying makedb. - Makefile changes to (possibly) improve "make forms" installation processing to be more generic. * utils/vacation/vacation.c: Add HAVE_DB_H test to be third known/supported database. Will it be sufficient for all environments shall be seen.. * utils/vacation/convtime.c: Add a couple autoconf includes 1998-06-04 Matti Aarnio * proto/newdb.in: Smarter option processing, usage() function. * compat/sendmail/sendmail.c: Recognize when no message was collected, and silently ignore such input. * compat/sendmail/sendmail.c: - Parse and verify option parameters for -B and -p options - implement -s option (save From_) - wrote a REAL usage() report -- try 'sendmail -?' to see * smtpserver/smtpcmds.c: When presenting diagnostics on bad parameters on MAIL command, the server used occasionally response number 455, where it should use 501 -- cosmetic bug, I would say.. 1998-06-03 Matti Aarnio * proto/zmailer.sh.in, proto/newaliases.in, proto/newdb.in, proto/newfqdnaliases.in, utils/makedb/makedb.c: - "makedb" got "-s" option so that in alias mode (-a) it will not utter final statistics report - Created command "zmailer newdb" which generates four well known flatfile databases into their respective binary views. (localnames, routes, aliases, fqdnaliases) If all goes ok, this command prints no statistics of databases. (You could guess: somebody asked for this :) ) - Others just recognize the "-s" option, and pass it around properly. * router/functions.c: Removed large chunk of dead code (#if 0 ..) * proto/Makefile.in: Fixed configuration file back-reference * smtpserver/smtpcmds.c: When printing HELO-string into rcvdfrom comment, quote also "(", and ")" characters. * smtpserver/smtpserver.c: Revised that 'smtpserver -i' mode code so that it should work on a wider array of systems. 1998-06-02 Matti Aarnio * smtpserver/smtpserver.c, smtpserver/smtpcmds.c: More Roy Bixler patches. Moved some code around so that the system will always know when it truly is connected to a network socket. The server will not produce "rcvdfrom " envelope header in case it is not running on a network socket. * Most Makefile.in files: Reworked on "prefix=nnnnn make install" installation. Now it creates the target parent directories if needed. * Makefile.in, ChangeLog: Version 2.99.50-s5 -- are we now safe ?? * transports/mailbox/mailbox.c: Added a bit earlier test for "user" /dev/null so that that target can be "delivered" to without privilege check. * router/rfc822.c: More gymnastics on getting the possibly missing "From: " header generated in all possible situations. 1998-06-01 Matti Aarnio * proto/newaliases.in, proto/newdb.in, proto/cf/i-routes.cf, proto/cf/standard.cf, proto/newfqdnaliases.in: Synced 'newdb' usages to supply explicite '-l' option, Modified scripts to require /bin/sh, not MAILBIN/router ! * router/rfc822.c: If the message had no 'From: ' header, generate one for it. * proto/cf/standard.cf: Add "bitnet" and "uucp" into the default toplevels list. * proto/cf/rrouter.cf, proto/cf/i-bitnet.cf: Updated BITNET routing rules; nic.funet.fi is Finland's *only* BITNET host these days, and works as BITNET-bound email gateway. * libsh/builtins.c, proto/cf/aliases.cf, proto/cf/aliases-new.cf: Whoops, two mistakes in aliases.cf -- wrong name at lappend parameter, and lappend internal functionality at empty target variable content was faulty (SIGSEGV..) * configure.in, libc/setregid.c, libc/Makefile.in, transports/mailbox/mailbox.c: Within mailbox.c there is one instance of fork(), and following it are some setgid()/setuid() statements. Setting real-uid/real-gid is not sufficient for systems with NFS-mounted home directories... Changed those calls to be setregid() and setreuid(), which should work far better, but I have no idea about its portability to for example SunOS 4.1.* series. * proto/Makefile.in, proto/cf/SMTP.cf.in, configure.in: Default router.cf is altered to have protocols 'routes', and 'smtp'. Per default the system install will not have UUCP in routing stage anymore. * smtpserver/smtpdata.c: Oops: At DATA processing the content-policy was tested only if the last RCPT policy analysis yielded All-OK (0) status. 1998-05-26 Matti Aarnio * proto/newdb.in: The script is in fact pure /bin/sh script, don't ask for running the router on it. * proto/cf/aliases.cf, proto/cf/consist.cf, proto/cf/crossbar.cf, proto/cf/aliases-new.cf: Replaced all uses of 'setf' with 'lappend' and 'lreplace', where appropriate. 1998-05-25 Matti Aarnio * libsh/builtins.c: Finetuning of sh_lreplace() gadget. 1998-05-19 Matti Aarnio * proto/db/smtp-policy.src, utils/policy-builder.sh.in: Define a new alias tag: _localnames to be used with "localnames" source file. It does not have "relaycustomer +" tag in it, thus just simply doing "MAIL FROM:" will not allow wide-open relaying... * libsh/trap.c: printf() argument castings into ones that are likely supersets of all existing instances -- mainly casting to long ... * proto/cf/standard.cf: relation 'localnames' got '-m' option, which was missing :-/ (detected when one system didn't react at changes in the localnames file/database) * lib/rfc822scan.c: A bug in error processing caused certain pathologic cases to follow NULL pointer... Outch. Sample case detected by Alexis Yushin. * lib/symbol.c: Removed conditionally builtin "pjwhash32()" function in favour of same function in separate file. * libsh/builtins.c: First stab at 'lappend' function * Makefile.in: Don't demand automatic dependency on "configure: configure.in", which would run autoconf, as CVS extraction gets dates reversed (or otherwise screwed up..) * configure.in, smtpserver/Makefile.in, smtpserver/smtpserver.c: Comment away "-lelf" and "-lkvm" autoconfig tests, and removed all instances of loadaver_current() uses. I don't believe there to be any worthwhile point in it. * configure.in: Fixes for "-lresolv" tests so that IRIX6.4 is happy. 1998-05-14 Matti Aarnio * All (very nearly) Makefile.in files, plus configure.in: Support CPPFLAGS input environment for the configure, and store it into all Makefiles that do compilation. It is intended to carry "-I/.." options for CPP phase so that some illogically placed headers can be found. * utils/makedb/makedb.c: Missed 'l' and 'u' from getopt() parameters; Oops.. (Andrew Tridgell reported) * scheduler/msgerror.c: Had one fseek() lying around from another branch of flow control -- but the "fp" parameter pointed to NULL :-/ (Andrew Tridgell reported) * transports/smtp/smtp.c: Two instances of zsyslog() with LOG_ALERT changed to LOG_ERR -- the detected errors are not ALERT level stuff. (Andrew Tridgell reported) 1998-05-13 Matti Aarnio * utils/makedb/makedb.c: Fixes on policy parser error reports, and generic duplicate key error's key printing routine. * proto/newaliases.in, proto/newdb.in, proto/zmailer.sh.in: As "makedb" got explicite "-l/-u" options, so did "newdb", and their callers. * proto/cf/standard.cf: Relation parameters for db FILES need to have "m" in them, we learned.. (via the hard way, of course). Now added it. * utils/Makefile.in: Conditionalize installing of "policy-builder.sh" script so, that it is no longer overwritten blindly by "make install". * smtpserver/smtpcmds.c, smtpserver/policytest.c, smtpserver/policytest.h, smtpserver/readpolicy.c, include/policy.h: Implemented smtp-policy attribute "maxinsize" which has numeric value, and can be used to limit (in source specific manner) incoming message sizes. Implemented also provision for "maxoutsize", which use remains to be seen.. (possibly for running inside smtp transport agent) * utils/makedb/makedb.c: Runtime options '-l' and '-u' for explicite lower-/upper- casification of key strings before storing them into db. 1998-05-05 Matti Aarnio * smtpserver/mxverify.c, smtpserver/smtpcmds.c: Spotted a case of certain timeouts slipping into hard errors when they definitely should not. Added new report for "Detected temporary DNS error, retrying may help.." * scheduler/readconfig.c, scheduler/scheduler.h, scheduler/transport.c, proto/scheduler.conf.in: Selector parameters: priority= nice= syspriority= sysnice= Further into at the default scheduler.conf(.in) comments. Idea thanks to * smtpserver/smtpcmds.c: If ident lookup is not done, don't add ' ident : ..' stuff into the 'rcvdfrom ' envelope header. 1998-05-04 Matti Aarnio * Makefile.in: Version 2.99.50-s3 * transports/smtp/smtp.c: A few cases of calls to smtp_sync() while SS->smtpfp was already NULLed are now protected. One can't do any remote syncing when there is no remote... 1998-04-27 Matti Aarnio * libc/Makefile.in: Fixes for installation of zmailer.h, and libzmailer.a (Noted by Eugene Crosser) * compat/sendmail/sendmail.c: Usage of "-f<>" option (source as the 'box') creates message with "channel error" for the reported envelope source. This was for easier debugging of 'channel error' things, but YMMV as the saying goes... * proto/Makefile.in: Weird problems in "make mailshare" installing *.conf files. Perhaps fixed now. (Noted by Eugene Crosser) * router/rfc822.c: Alternate scheme for dealing with SMTP MAIL FROM:<>, and lack of "From:" header in the message. * Makefile{,.in}: Version: 2.99.50-s2 1998-04-24 Matti Aarnio * configure.in, acconfig.h, lib/prversion.c: Store ./configure command parameters into a #define, and present them in places calling prversion(). (Makes sense in several cases, IMO..) * configure.in, (most)/Makefile.in: New configuration option: --with-getpwnam-library="-L.. -l.." helping overloading of getpwnam() (and getpwuid() ?) library calls with your own special ones; if you have some ISP system with backend databases, etc.. 1998-04-09 Matti Aarnio * configure.in: RedHat 5.0/glibc-2.0.7-6: found res_init() in libc, but not others, like res_mkquery(). * transports/mailbox/mailbox.c, man/mailbox.8, README.UPGRADING: New option: "-C" to canonify username to what pw_name from getpwnam() call yields. Previously this canonification was done blindly -- does not affect most people, but there are a few weird fellas whose getpwnam() uses key A, and yields string B in pw_name field.. (These are not only Telecom Finland weirdos..) 1998-04-08 Matti Aarnio * Makefile.in: Version 2.99.50-s1 * transports/mailbox/mailbox.c: New hash-directory option: -P (and -PP) This uses pjwhash32() on the username and then calculates a sub-directory hash from that -- in range of 'A'..'Z'. With two -P's, the hash is two-level deep: /dir/X/Y/uname where X is calculated as 'A'+(hash/26)%26, and Y is 'A'+(hash%26) * transports/mailbox/mboxpath.c: A new command with options '-d dirpath', '-P[P]' and '-D[D]' and username parameter. It calculates same hashes, and reports to stdout the path of the mailbox. This is intended for system-wide startup script usage with code something like following for sh: MAIL=`mboxpath -P $USER` It is important to use SAME options here as are used at running the mailbox program. * lib/pjwhash32.c: Pulled from elsewere a hash-routine to convert a string into a hash-value. Used at mailbox subdirectory hashing.. * transports/smtp/smtp.c: "-L localidentity" option fix. Now it can select from possible multiple local identities. Also corrected similar problem with '-F' option processing. 1998-04-03 Matti Aarnio * smtpserver/smtpserver.c, smtpserver/smtpdata.c: Report end of "DATA" and "BDAT" commands too -- the server may be sleeping around on idle-states for a long time.. 1998-03-27 Matti Aarnio * transports/mailbox/mailbox.c: Corrected the use of x.2.2 type of mail status code from "5.2.2" to "4.2.2". (Persistent transient error.) * INSTALL: Warn about configuration phase --prefix="parameter" string size limitations due to usual shell-script execution rules on UNIXes. * router/libdb/header.c: Alter the syntax of "Reply-To:" header match that of RFC-822. Oops since 2.2.1 (and probably before!) * router/libdb/bind.c: Just nice for debugging -- report DNS lookup error codes in text instead of "magic" numbers -- "NXDOMAIN" instead of "3", for example. 1998-03-26 Matti Aarnio * transports/mailbox/mailbox.c: A quota-check hook from Eugene Crosser 1998-03-25 Matti Aarnio * smtpserver/smtpcmds.c, smtpserver/smtpserver.c: Limit a bit about how much meta-lines are logged. If policyresult status is zero, don't report it. * libident/identuser.c: When connect() yields "ECONNREFUSED", present report "NO-IDENT-SERVICE" instead of "SOCKFAULT1". (Local users at a large SPARC server running 2.6 and sending email with PINE thru SMTP.) 1998-03-22 Matti Aarnio * libsh/builtin.c: Function sh_echo() got slight alteration. Now it detects options: [-n] [--] [rest of args] in this order. This in order to be able to write script statements of style: echo -- "$@" where the arguments MAY start with first component of "-n" ... * proto/cf/canon.cf: Rearrangeing the script a bit, do $(canon ..) calls in canonical () in case the hostname has at least one dot in it, and the top-level part is not know one. * router/db.c: Somebody asked for a method for querying available databases in the router scripts. Method is: relation -T -t dbtype dummyname Return-code 0 means the 'dbtype' is known, and non-zero means it is unknown. * lib/selfaddrs.c, transports/smtp/smtp.c: Modify function matchmyaddress() to return value 3 for such destination addresses which are of invalid network (Valid range being: 1..223; out of which 127 yields value 2) This was prompted by noticing appearance of faulty DNS data in larger amounts in the network. Lots of A 0.0.0.0 entries especially. 1998-03-20 Matti Aarnio * INSTALL, doc/guides/configure, doc/guides/smtp-policy: A bit of rewriting on INSTALL file, new guide file, and update of smtp-policy guide. * ChangeLog, Makefile.in: Pushing out version 2.99.49p10-s11 1998-03-17 Matti Aarnio * proto/cf/rrouter.cf, INSTALL: Match $hostname also with non-lowercase domain address. 1998-03-16 Matti Aarnio * scheduler/msgerror.c: In case boiler-plate template was not found, writeheader() would produce bad error report message which router would not accept for processing. 1998-03-12 Matti Aarnio * scheduler/qprint.c: A patch from Andrey Blochintsev to restore the functionality of "mailq -v" with scheduler running with "-H[H]" mode. * proto/cf/rrouter.cf, proto/cf/aliases.cf: When user has .forward but it is empty, aliasing pushed said user to bitbucket. Now the .forward is expanded to be "\username", and no recursed .forward will be entered. Thus the user will stay in the local system. 1998-03-11 Matti Aarnio * router/rfc822hdrs.c: A user found pathological case of errors which caused his router process to blow up. Turned out to be overflow of stack contained array :-( Outch! Fixed/limited several array/buffer references in the module. * transports/smtp/smtp.c: Full-blown pipelining -- will collect replies asynchronously, and merge results into smtp_sync() states while still sending things out. This limits (sometimes) amount of things kept in the reply pipeline buffers. * proto/cf/aliases.cf: Backed off some changes; pushed too much into bitbucket :-( * transports/smtp/smtp.c: Alexis Yushin reported that BDAT didn't work in non-pipelined mode :-( Corrected that. 1998-03-10 Matti Aarnio * proto/cf/aliases.cf, proto/cf/rrouter.cf, proto/cf/fqdnaliases.cf: Rewamped $(rrouter ..) and some other scripts after reports regarding mysterious problems with $PUNTHOST. * include/shmmib.h, scheduler/prototypes.h, scheduler/scheduler.c, scheduler/scheduler.h, scheduler/update.c: Hooks/preparations for RFC 2249 * smtpserver/smtpdata.c: Some anti-spam hacks merged in 1998-03-04 Matti Aarnio * transports/smtp/smtp.c: Correction to EHLO SIZE processing, aborting message sending if the remote system presents a size limit, and current message size estimate exceeds that value. Corrections to the way, how active smtp socket peer IP addresses are presented for diagnostic reports. 1998-03-03 Matti Aarnio * transports/smtp/smtp.c: Smarter processing for PIPELINING error reports; if PIPELINING mode yields error on MAIL FROM, use that report at all RCPT TO lines. Also support multiline reports from remote servers (up to a few kilobytes long..) * smtpserver/smtpserver.h, smtpserver/smtpserver.c, smtpserver/Makefile.in, smtpserver/policytest.h, smtpserver/smtpdata.c, smtpserver/contentpolicy.h, include/mail.h.in, include/policy.h: - Modified several multi-line policy report messages into a narrower form with same text folded on shorter lines - Added hooks for contentpolicy() function that will be able to analyze file content for possible acceptable content issues ("SPAM"...) * libc/mail.c: mail_fname(FILE * mfp) A new function returning filename of currently open mail related FILE *. * Makefile.in: Version 2.99.49p10-s10 * proto/cf/canon.cf, proto/cf/rrouter.cf: Modifications on canonicalize() routines so that X.400 gateway address of type: /foo/bar@gate.way will go thru unharmed. * router/rfc822.c: If there is no "From:" header, and the source channel is "error", create a "From: <>" header. At the same time the source ENVELOPE rewrites itself to be postmaster@local.host.name (hmm..) 1998-02-28 Matti Aarnio * transports/smtp/smtp.c: Still problems with PIPELINING; doing smtp_sync() twice wasn't prepared for a thing, and thus in some cases (all cases?) syncing after BDAT didn't quite work ok. * proto/cf/rrouter.cf: "lotusgw!" routes_spec() tag -- exact details are in state of flux, but this is for Telecom Finland operations. Apparently suitable pre-routing at smart MTAs can help a lot with system performance at lotus-smtp-gateways. 1998-02-27 Matti Aarnio * Makefile.in: Version id 2.99.49p10-s9 * scheduler/threads.c, contrib/mrtg/mrtg-zmq: Noticed that "mailq -sQQ" output does not produce consistent results when the scheduler has no jobs in queues, nor any children. Changed the reporting a bit to give out a line of zeroes, if nothing else. Added a bit more comments and explanations into the mrtg-zmq script; it is a tool, but in itself it is NOT sufficient for graph producing, just to produce the data. * smtpserver/smtpserver.c: Recognize several error conditions within s_getc() in connection with read(2) call. * transports/smtp/smtp.c, include/ta.h, transports/libta/writeheaders.c: Implemented the outbound RFC 1830 CHUNKING (BDAT) method for message sending; needed revised methods on header writing -- to a buffer. 1998-02-26 Matti Aarnio * utils/makedb/dblook.c, utils/makedb/makedb.c: printf format string fixes * transports/smtp/smtp.c, transports/libta/diagnostics.c: Carry IP address in WTT data to the DSN reports. Thus will tell exactly which IP addresses tried to connect to each other. * smtpserver/smtpcmds.c: Last of Gabor Kissig's router-driven policy facility patches have been incorporated into the server. 1998-02-25 Matti Aarnio * scheduler/resources.c: Per report from Martin Wendel, AIX 4.2.1 yields surprising values on getrlimit(RLIMIT_NOFILE, ..) result structure. Altered things so that we won't use that facility on _AIX systems -- ever. * router/functions.c: $(syslog ...) and $(logger ...) functions 1998-02-20 Matti Aarnio * scheduler/threads.c (and others): Radical thread-group job-picking reorganizing. More deterministic approach on scheduling with better guarantees of even activation of threads. * scheduler/mailq.c: The "-Q" mode receiving of non-Q-mode vertices data streamlined so that it does not require up to infinitely long linebuffers... * transports/smtp/smtp.c: NULL-ptr reference in logfp output. Oops... 1998-02-17 Matti Aarnio * transports/smtp/smtp.c: A weird error report surfaced, which turned out to be error in PIPELINING smtp_sync() processing. Remote end gave 250 answers to all MAIL FROM, and RCPT TO lines, but then gave 451 for DATA. The code had NO handling of DATA command error :-( (There was apparent assumption that if any RCPT TO is ok, then DATA will be ok too..) 1998-02-16 Matti Aarnio * scheduler/scheduler.c, scheduler/update.c: When #resync:ing a file, read it immediately back, but for those vertices that have a lock, issue a delayed "wakeup".. * proto/cf/aliases.cf: User's domain identity carrying fixed * proto/cf/canon.cf: Unnecessary back-slashes removed * proto/cf/crossbar.cf: Fine-tunings for rewriting outbound email envelope with source domain at the local host * proto/cf/p-usenet.cf: Automatic dbtype configuration * router/db.c: Cleaned up the address processing expressions * smtpserver/smtpdata.c: Corrected message-id analysis routines a bit 1998-02-13 Matti Aarnio * proto/zmailer.sh.in, utils/rotate-log.sh: Updates for log-chop processing with various programs -- routers and the scheduler are slightly different animals in this regard.. 1998-02-12 Matti Aarnio * router/functions.c: Logical expression error within run_dequote() caused faulty processing of string "'" in such manner, which caused process to blow up... 1998-02-05 Matti Aarnio * include/libc.h: Prototypes of inet_ntop() and inet_pton() changed to match the world view -- Solaris 2.6 is now happy with us.. * proto/cf/p-smtp.cf: Patch by Alexis Yushin for UUCP route handling * compat/rmail/rmail.c: Patch by Alexis Yushin for proper use of values * transports/errormail/errormail.c: An off-by-one fix by Alexis.. Wed Feb 4 19:52:36 1998 Matti Aarnio * configure.in: More modifications to support RedHat 5.0 systems, and at same time to support Solaris 2.5.1. * lib/esyslib.c: erename()/eqrename() tricks to handle Solaris returning EBUSY at inconvinient moments.. Now will retry the thing until successfull (or failed with errno != EBUSY/EINTR) * libident/identuser.c: Added setjmp()/longjmp() along with alarm() handling to make sure we have shortish timeout * smtpserver/smtpcmds.c: Cosmetic fixes on verbose log file * smtpserver/smtpdata.c: Recognize a couple of headers that exist on spam email * smtpserver/smtprouter.c: Recognize router subprocess death, and report it with "400 Help...!" message, not silent "250 All fine." * smtpserver/smtpserver.c: Report RBL TXT data to the connecting user * transports/libta/diagnostic.c: Trunc-back the transport-spec file if its appending fails. * transports/smtp/smtp.c: In EHLO/HELO/reconnect/HELO/... connection test attempts do retry all MXes -- but only once each. Sun Dec 21 14:59:04 1997 Matti Aarnio * libc/setreuid.c: AIX code missed "return 0;" at the end. (From: pakrat@raleigh.ibm.com) And also SysVr4 emulation code had missing "return rc;" at its end.. Sat Dec 20 01:05:00 1997 Matti Aarnio * smtpserver/smtpchild.c: Altered the child registration and detection mechanism to test for "sneaky" disappearances of subprocesses before getting them registered into the server childs[] table. Has "tunable" interval on how to monitor on the existence of subprocesses. * smtpserver/cfgread.c: Removed "PARAM toplevels", a stillborn thought. * libsh/sh.ssl, libsh/sslwalter.c, proto/cf/*.cf{|,.in}: Codified error report on use of 'sift' / 'tfis' syntax in place of 'tsift' / 'tfist' names. Report mismatching keywords -- like 'ssift' ending with 'tfist'.. Fixed all prototype *.cf and *.cf.in files in this regard. * libsh/test.c: Don't complain always at failing *stat() calls -- allows things like this without spurious junk at the screen: if [ ! -f $MAILVAR/db/aliases$DBEXTtest -o \ $MAILVAR/db/aliases -nt $MAILVAR/db/aliases$DBEXTtest ]; then # Yes, so update! $MAILBIN/zmailer newaliases fi Wed Dec 3 00:39:45 1997 Matti Aarnio * router/libdb/gdbm.c, router/libdb/ndbm.c: GDBM sets gdbm_errno with value GDBM_EMPTY_DATABASE if the DB is empty; NDBM does not complain of the situation. GDBM and NDBM had poorly (nee, pathologically) functioning search ( = fetch ) retry routines, rewrote them slightly. * smtpserver/smtpserver.c: Lacked stashmyaddresses(NULL) call to prime the interface address data. * smtpserver/mxverify.c: Considered only the first MX record arriving; Duh :-( Sat Nov 29 16:02:48 1997 Matti Aarnio * libsh/interpret.c: Finally! Found the place(s) of leaking memory (conscell + string) in the router. Now nic.funet.fi does not have infinitely growing router :-) Mon Nov 24 20:44:27 1997 Matti Aarnio * configure.in: Re-organize the --with-ipv6-replacement-libc option processing -- now it finally works. * router/libdb/unordered.c, router/libdb/ordered.c: The incore database of filenames (spt_files) reorganization, and code reuse on the open_seq()... Fri Nov 21 19:48:06 1997 Matti Aarnio * utils/makedb/makedb.c: - "-A" option to APPEND more data to same key. This behaves slightly differently for aliases, and for policy -- former being ascii, latter being binary. - Process alias entries in similar manner as the sendmail does them * router/libdb/ndbm.c, router/libdb/bsdbtree.c, router/libdb/gdbm.c, router/libdb/bsdhash.c, router/libdb/dbm.c, router/libdb/ordered.c, router/libdb/unordered.c: Retry the DB file open three times with 1 second sleep in between; this to allow certain small window of db file unavailability during mv operation of the installation of a new version. * libsh/interpret.c: Fix errorneously added s_free_tree() -- it really is ok to free() only the top cell. Freeing more blows things up.. Tue Nov 11 04:16:41 1997 Matti Aarnio * transports/smtp/smtp.c: Extensive revisioning of MX processing in order to eliminate spurious "DNS inconsistensy detected, no MX, no Address" reports. Now it seems the things are under control. * smtpserver/smtpserver.c, smtpserver/policytest.c, smtpserver/policytest.h, smtpserver/smtpcmds.c, smtpserver/smtpdata.c, smtpserver/readpolicy.c, smtpserver/mxverify.c, include/policy.h: - Initial implementation on http://maps.vix.com/rbl/ - mechanisms to pass some data back from policytest to the smtpserver proper, and to add some logs.. * transports/mailbox/mailbox.c: Support "-d" option for defining MAILBOX directory at runtime. * transports/expirer/expirer.c, transports/expirer/manual-expirer.in, transports/expirer/Makefile.in, configure.in: Manual interface for the transport-agent level entity; providing also easier usage for the expirer in case of wanting to expire something from the queues. * transports/mailbox/mailbox.c: Non-zero exit codes from pipe subprocesses were considered all bad. Now things have been altered to treat most of them as EX_TEMPFAIL, and only very few as permanent. Found this problem out by a procmail using collegue, when his home disk filled up, and deliveries bounced with code EX_TEMPFAIL :-O * scheduler/scheduler.h, scheduler/transports.c: Store approximation of the argv[] into a per-process array of subprocess states, and in case the transport sub-process exits with EX_SOFTWARE status code, send syslog() message of MAIL_EMERG, and more verbose stuff to stderr log file. At syslog config you might consider directing MAIL_EMERG to some people directly, and in realtime.. * scheduler/msgerror.c: When reporting errors, use correct pointer to the "plain ascii" error report. * smtpserver/smtpserver.c, smtpserver/policytest.h, smtpserver/readpolicy.c, smtpserver/policytest.c, include/policy.h: - If the policydatabase can't be opened, delay a bit at the connect formation, but always send syslog MAIL_EMERG in hope of it becoming handled in realtime... - Add new policytest attribute: sendernorelay * libsh/interpret.c: In preparation for more debugging, separate things in between sAssign, and sTempAssign executions. * router/libdb/yp.c: "struct search_info *" was old style, declaration, new one is "search_info *" * smtpserver/policytest.c: Search for the first unquoted at-character in the input address -- for things like: <"foo@bar.foo"@bar.foo> (Not that I particularly like the idea of supporting this kind of addresses -- people using them are prone for nasty surprises. And frankly I am afraid they might get past relay-hijack blocking methods too...) * smtpserver/smtpserver.c: The same-ip-connect-source registration, and trapping needs a bit of rearrangements -- namely the subprocess must wait enough for it to become registered for sure.. Now in front of nearly all exit() calls is sleep(2), which tries to ensure that the subprocess will exist long enough for the main server to be able to register it before subprocess' death. * utils/makedb/makedb.c: - policy compilation shall reject all inputs with invalid contents - alias compilation shall treat extended input lines as such where the previous line is has appended "," at the end. * router/rfc822.c: Set trace header ("Received:" header) date to be that of the input spool file mtime. Mon Oct 27 09:51:02 1997 Matti Aarnio * doc/guides/procmail: Condensed wisdom on how to use procmail in place of the "mailbox" local delivery channel for user deliveries. * doc/guides/cyrus-imap: Edit version the "procmail" guide with relevant info regarding CMU Cyrus message store. * scheduler/readconfig.c: Match documents about parameterless clauses. * compat/sendmail/sendmail.c: Ignore (silently) all '-oXXXX' options we don't recognize. * router/libdb/selfmatch.c, proto/cf/rrouter.cf: Corrections for selfmatch routines from Darryl Miles * scheduler/scheduler.c, scheduler/threads.c, scheduler/agenda.c: Use controlled per facility symbol splaytrees, and shrink them when the symbol is no longer needed. (i.e. channel, and hostnames come and go..) Sun Oct 26 03:30:08 1997 Matti Aarnio * include/splay.h, lib/splay.c, lib/symbol.c, router/db.c, router/conf.c, router/prototypes.h, router/router.c, router/libdb/incore.c: Seriously limit the amount of memory needed to handle operative symbols. Perhaps there is (after all) no need to have infinite memory capacities. (Each object cache has its own namespaces, and when the need for some named object disappears, it is removed from the namespace.) Wed Oct 22 22:38:10 1997 Matti Aarnio * Releasing 2.99.49p8 to public after series of compilation, and running tests at various platforms. These include: Linux/ELF/libc5, Linux/AXP/glibc2, Solaris 2.5, DEC UNIX 4.0 * libsh/io.c: glibc loader complained about the use of 'gets()', removed the unused code block. * utils/vacation/vacation.c: #include * utils/vacation/Makefile.in, libc/myhostname.c: Revised the routine not to use gethostbyname(), and finally realized that all it missed was -lresolv library.. * libc/getnameinfo.c, libc/getaddrinfo.c, include/netdb6.h, configure.in: Corrected and completed the --with-ipv6-replacement-libc configuration parameter functionality * smtpserver/rfc821scn.c: Corrected object scanning * proto/db/localnames: Default content of the file is EMPTY, not nic.funet.fi's hostname listing... * proto/Makefile.in: Per default install also files: $MAILSHARE/smtpserver.conf $MAILSHARE/router.cf if they are not present at the system * compat/sendmail/sendmail.c, man/sendmail.8: Sync "-t" option with sendmail, and especially so with the manual page! Mon Oct 20 09:46:04 1997 Matti Aarnio * Makefile.in: Version 2.99.49p8 (way too many p6* versions around..) * Makefile.in, configure.in, utils/Makefile.in, utils/makedb/Makefile.in, proto/fullnamealiasmaker.pl, proto/newaliases.in, proto/newdb.in, proto/smtpserver.conf.in, utils/makedb/makedb.c, utils/makedb/dblook.c, utils/policy-builder.sh.in: Renamed 'makendbm' to 'makedb', and 'ndbmlook' to 'dblook'. * smtpserver/policytest.c, smtpserver/policytest.h, utils/makedb/makedb.c, utils/makedb/dblook.c, router/db.c, router/libdb/search.h, router/libdb/bsdhash.c: Created 'BHASH' type database which is BSD DB_HASH * configure.in: - Fix --without-fsync option - Fix test on malloc_d - Add option: --with-ipv6-replacement-libc * include/splay.h, lib/splay.c: Keep track of elements on the splay tree for debug porposes (who is leaking memory at the router..) * libc/getaddrinfo.c, libc/getnameinfo.c: Sync with Craig Metz's versions 1.26 and 1.32 respectively. * proto/db/smtp-policy.src, smtpserver/policytest.c, smtpserver/policytest.h: Do testing on connection source IP reversal domain too. This allows us to detect certain set of dial-in users who abuse dial-in series to post spams... * smtpserver/mxverify.c: * transports/smtp/smtp.c: More armour on host connection retry, and giveup on scheduler shutdown. * utils/vacation/vacation.c: Do message production without invoking /usr/lib/sendmail Fri Aug 8 18:04:06 1997 Matti Aarnio * Makefile.in: Version 2.99.49p6 (could it be ?) * proto/db/smtp-policy.src, doc/guide/smtp-policy, include/policy.h, smtpserver/policytest.c, smtpserver/readpolicy.c, README.SPAM: Updated documentation to match code, added policy parameters: 'acceptifdns {+|-}' which accepts recipient address when it has MX or A (or AAAA) record in the DNS. Thu Aug 7 23:00:52 1997 Matti Aarnio * Makefile.in: Version 2.99.49p6a3 * SiteConfig.in, scheduler/scheduler.c, scheduler/update.c: Eugene Crosser's SYSLOGFLGs for the scheduler. * scheduler/threads.c, scheduler/update.c: A weird null-pointer derefferrence problem with gcc 2.7.2.1 at DEC UNIX 4.0 -- perhaps it is about gcc bug... * include/policy.h, proto/db/smtp-policy.src, smtpserver/Makefile.in, smtpserver/mxverify.c, smtpserver/policytest.c, smtpserver/readpolicy.c, smtpserver/policytest.h, utils/makendbm/makendbm.c: Understand new policy parameters: 'senderokwithdns {+|-}' 'acceptifmx {+|-}' of which the first (when existing) uses DNS lookup to verify that the source address is acceptable, and the second uses DNS to check that we are MX for the given target domain. '+' means 'yield SOFT (4XX) error', and '-' means 'yield HARD (5XX) error'... * smtpserver/smtpserver.c: Eugene Crosser's syslog fix * smtpserver/smtpserver.c, util/policy-builder.sh: If policy database is defined, but can't be opened, smtp-server responds "400 Policy database problem, code=NN" to every command (except 'help' and 'quit'). The codes mean: 1: Unknown db format, 2: db open failed * smtpserver/smtpcmds.c: A mistake in character counting in RFC821_822QUOTE macro. * router/rfc822.c: Detect if scheduler job-description file writing has a fault, and give up. The job will go into deferred directory, from whence it should be (via cron?) moved back into the router. * proto/Makefile.in: Removed various +XXXX targets, and touches of them.. * transports/smtp/smtp.c: When the transport-agent receives 'SIGTERM', it sees if it can die immediately, or if it should take a delayed action on the death issue. The delayed action is done when the smtp transport is doing '.' write operation, and waiting for the acknowledgement, plus doing diagnostics. * utils/vacation/vacation.c: At Alpha, storing LONG_MAX into time_t doesn't work, use INT_MAX. Mon Aug 4 17:29:26 1997 Matti Aarnio * Makefile.in: Version 2.99.49p6a2 * INSTALL, Makefile.in, proto/Makefile.in: "make install-cf" * configure.in, proto/Makefile.in: Make (for installation) proto/scheduler.conf, and its "distclean" removal.. * router/libdb/bsdtree.c: print_btree() had wrong test in db printout iteration (found while wondering other thing..) * utils/makendbm/Makefile.in: Corrected for architecture-subdir compile * utils/vacation/vacation.sh.in: Sigh... Give up, and execute the correct binary in case we called user-callable (= wrong) script. * router/Makefile.in: Architecture-subdir compile for rfc822.ssl, and its related use. * include/libsh.h, libsh/builtins.c, libsh/trap.c, router/functions.c: Altered eval() routine call parameters. Also altered ZMSH.fc header format a bit. Sat Aug 2 19:57:57 1997 Matti Aarnio * Makefile.in: Version 2.99.49p6a1 * router/libdb/ldap.c, router/db.c, router/libdb/search.h, configure.in, config.h.in, router/libdb/Makefile.in. doc/guides/ldap.doc: LDAP db driver by Lai Yiu Fai , plus integration, and autoconfig support by yours truly. * utils/makendbm/makendbm.c, utils/makendbm/Makefile.in: Massive code rearrange, now "-a" (alias) system does proper continuation line handling too. * utils/makendbm/ndbmlook.c: Concise data output dumper code * proto/newdb.in, configure.in, SiteConfig.in, proto/cf/*.cf, INSTALL, proto/zmailer.sh.in: Build default database format binary db with command: $MAILBIN/newdb $MAILVAR/db/dbase-basename also have two short-circuit commands: zmailer new-localnames zmailer new-routes * libsh/interpret.c: Found, and corrected a bug in ZMSH expression: variable=${1:-defvalue} The mechanism mistreated ARGV values * router/db.c: "btree" dbase entry was missing a call to modp_btree() routine which does exist, but just was never called.. * router/libdb/selfmatch.c, smtpserver/rfc821scn.c, transports/smtp/smtp.c IETF DRUMS smtpupd draft for IPv6 address literal says: "[IPv6" SP "]" That is, there is SPACE instead of "." after the magic prefix. Huh! Quite magic... * smtpserver/smtpcmds.c: Handle correctly zero size addresses in MAIL FROM:<>, and (heaven forbid!) RCPT TO:<>. Fri Jul 25 01:48:13 1997 Matti Aarnio * Makefile.in: Version 2.99.49p5 * All Makefile.in files, configure.in: mechanics to compile in VPATH environment, and to install with "prefix=/my/tmp/root make install" * config.h.in, configure.in, router/db.c, router/libdb/hostsfile.c: Tests for gethostent() sethostent() and endhostent() * include/ta.h, transports/libta/ctlopen.c: New "DSN" object parsing: INPTR= * libsh/interpret.c: ssift attribute string dequotation mechanics rewrite (correcting operation, avoiding core-drops..) * man/smtpserver.8, smtpserver/cfgread.c, smtpserver/smtpserver.c, smtpserver/smtpserver.h: Syncing man-pages with the reality, new PARAM ListenQueueSize nnn runtime parameter with default value of 20 000. * proto/cf/i-smtp.cf, proto/cf/map.cf * proto/scheduler.conf, scheduler/scheduler.c: channel/host selector: chan/*.{fii,foo} * proto/scheduler.conf, proto/sm.conf.in, proto/cf/standard.cf 'bitbucket' channel * scheduler/scheduler.h, scheduler/readconfig.c, scheduler/threads.c: "queueonly" flag by Alexis Yushin * smtpserver/policytest.c: Ascii-mode debug printing routines by Eugene Crosser * moved: support/vacation/ -> utils/vacation/ * transports/expirer/expirer.c, transport/expirer/Makefile.in: New transport agent with thought of it becoming a tool for the scheduler, but apparently it is also well suited for manual operation. Docs are missing.. * transports/libta/writeheaders.c, transports/hold/hold.c, transports/mailbox/mailbox.c, transports/smtp/smtp.c, transports/sm/sm.c: Moved implicite newline after header write out of the writeheaders() routine into each caller. Thus became able to do smart things with mailbox; ORCPT XTEXT decoding, for example. * transports/mailbox/mailbox.c: When writing to a pipe, don't care if the write stops due to some error, but if the pipe end-point process exists with non-zero status, THEN become upset! (Still something mysterious with pipe reports -- I see positive reports even when I don't ask for them..) * transports/smtp/smtp.c: Error reporting correction in one rare case, and general streamlining. Fri Jul 4 14:54:02 1997 Matti Aarnio * Makefile.in: Version 2.99.49p4 * router/rfc822.c: Nuke headers: "Bcc:", "Return-Path:", "X-Orcpt:", "X-Envid:", and respective "Resent-*:" variants * router/libdb/bsdbtree.c: If the db can not be opened ( open_btree() returns NULL ), don't crash on it.. * scheduler/msgerror.c: In case the report we are composing does have "NOTIFY=NEVER" in effect, but no valid recipients exist, ignore any such NOTIFY value, and return the message in full -- to the system postmaster... * transports/mailbox/mailbox.c: Found a bug in let_buffer[] reuse -- it was an "automatic" variable, that is, stack reclaimed it in between message deliveries, and of course were able to corrupt it! I also enlarged the let_buffer[] size to be of reasonable size -- 16k to 64k depending on your BUFSIZ value. * transports/mailbox/mailbox.c: Add headers "X-Orcpt:" and "X-Envid:" to the message when storing it into a file. For a pipe the data is passed in envelope variables. (I got rather fed up with spams this morning, and decided to add a bit more trace as to whom the message has been sent originally..) * transports/smtp/smtp.c: Accept non rfc-821 behaviour of the remote server by treating non-conformant reply of form: "NNN" the same as conformant reply of form: "NNN". Sun Jun 29 21:36:43 1997 Matti Aarnio * Makefile.in: Version 2.99.49p3 * configure.in, config.h.in, transports/libta/ctlopen.c, transports/mailbox/mailbox.c, transports/sm/sm.c, transports/smtp/smtp.c: Configuration option: --with-ta-mmap which uses mmap() facility to map in pages of the message data. At least at Solaris 2.x it is not advisable to use if you want to have highest possible performance, because the munmap() throws away ALL pages of the file from the system memory, even when other users have those pages in use! For that reason it is much better to rely on plain simple read()/write() behaviour that let buffer cache to exist. * configure.in: Configuration option: --without-fsync Option that turns off (globally) the use fsync() anywhere in the ZMailer. System configurers can decide what is best for their systems. It does give definite edge at buffered IO by not forceing every file modification to be commited up to the disk. * smtpserver/rfc821scn.c: Alteration at the treatment of "!" and "%" characters. Now they are special, and are analyzed in more profound manner within localpart analysis. * smtpserver/smtpchild.c: Register, and analyze properly how many SMTP sessions are currently active (non-reaped) from any given source IP address. * transports/smtp/smtp.c: Correction into notary-status-code pickup for ENHANCEDSTATUSCODES mode. * libsh/testeval.c: (+ several of proto/cf/*.cf files) New operators into the builtin 'test' function: -nt, -ot, -ef (file age comparators, and file equality comparator) * SiteConfig.in, router/rtsyslog.c, smtpserver/smtpserver.c, transports/libta/tasyslog.c: New ZENV entry: SYSLOGFLG= which can be used to turn off syslog() logging activity at the smtp-server/router/transport-agents * config.h.in, configure.in, router/router.c, scheduler/scheduler.c: Check of setgroups() routine, and at both the router, and the scheduler, use both to clean out the set of supplementary groups the processes have. * include/hostenv.h: Concise USE_ALLOCA/HAVE_ALLOCA defines * libident/identuser.c: Be a bit more tolerant at what comes back as response for the ident query. * libsh/interpret.c, router/functions.c: A bit more smarts into quote containing string processing; Especially when dequoteing string that contains only one char: ' or " ... * proto/cf/i-routes.cf: Accept incore database of form: $MAILVAR/db/routes.zmsh * router/libdb/headers.c: Allow "Reply-To:" and "Resent-Reply-To:" to have value: "<>" (sigh... It is against RFC-822, but I see such usage few times each day...) * scheduler/msgerror.c, transports/errormail/errormail.c, transports/fuzzyalias/fuzzyalias.c, transports/mailbox/mailbox.c: Stronger MIME-report boundary generator which isn't merely stocastic, but very likely unique at the creation time. * smtpserver/policytest.c, smtpserver/policytest.h: Move policy-database open out from database define routine. * smtpserver/policytest.c: - Study "rejectsource" at HELO/EHLO parameter check - For the RCPT TO:<..> study also the full addresses in addition to the domain part * smtpserver/smtpcmds.c, smtpserver/smtpserver.c: - Print out the incoming HELO/EHLO parameter as is into comment at "rcvdfrom" envelope header along with many other bits of info. - Change policy-rejection responses from: 553 5.1.8 ... to: 553 5.7.1 ... * ssl/ssl.c: Conflicting routine name within the ssl.c program: basename() ( Linux libc6 a.k.a. glibc-2.0.4 .. ) * support/vacation/vacation.sh.in: Modifications into 'vacation start' command behaviour. * transports/mailbox/mailbox.c: Found a couple off odd situations where creation of files into non-kosher location was allowed; partly corrected with the purge of supplementary groups at the scheduler, partly with a bit of new code. * transports/smtp/smtp.c: Recognize (unknown) smtp response value 571 to be a sign of policy based refusal. * utils/makendbm/ndbmlook.c: At Debian systems including causes surprises at parameters called 'errno' ... rename such parameters. Sun Jun 22 04:57:02 1997 Matti Aarnio * Makefile.in: Version 2.99.49p2 * configure.in: sprintf() return type testing fix * include/hostenv.h: Too many files need to include , thus we added it into "hostenv.h"... (for SunOS 4.1.x) * libc/getaddrinfo.c, libc/getnameinfo.c: "extern int h_errno;" added * proto/cf/SMTP+UUCP.cf.in More suitable, and more generic version * smtpserver/smtpcmds.c: A fix in ORCPT pseudo-formation in case the original sending system did not use the option.. * support/vacarion/vacation.c: Properly placed 'extern char * strerror();' * transports/libta/diagnostic.c: Again optimize int returning sprintf better, than the old char * returning... Thu Jun 19 19:41:47 1997 Matti Aarnio * Makefile.in Version 2.99.49p1 * configure.in: Reordered (and changed default) of --with-libmalloc= option parameters * lib/selfaddrs.c: Detected that SunSoft SunPro C 3.0 didn't define '__svr4__', but does '__SVR4' -- automatic Solaris support failed due to that.. * libsh/listmalloc.c: Debugged, and got it finally working.. (something for Zmailer 3.0 ...) * smtpserver/smtpcmds.c: Freed memory blocks that were not malloc()ed at all.. * transports/libta/diagnostic.c: 'notarybuf' can be a NULL pointer, prepare for it! * utils/makendbm/makendbm.c: Just some casts to please SunOS 4.1.x compilation Fri Jun 13 22:41:04 1997 Matti Aarnio * Makefile.in: Version 2.99.49 * README.PERFORMANCE: Added some notes about what performance figures we have achieved with this beast * ALL SOURCE FILES (more or less): - Threw in 'const' before 'char *', and got a LOT of changes all around - Converted lots and lots of places of 'unsigned char' into 'char' -- in very few cases also chopped off extra-high bits loading unsigned chars from strings when needed. - Added function prototypes all around, and fixed a couple real bugs in rarely used parts of the system - changed all instances of 'struct conscell' into 'conscell' (typedefed object) for a later jump in altering the conscell object handling sometime later in the developement. * lib/strlower.c, lib/strupper.c: New library routines * smtpserver/readpolicy.h, smtpserver/policytest.c, smtpserver/policytest.h, README.SPAM, man/smtpserver.8: Integrated policytest()/policytestaddr() routines with compiled in logics of earlier Gabor Kissigs interactive router subprocess based system without the need to run the router as a subprocess... * utils/makendbm/makendbm.c, utils/makendbm/ndbmlook.c: Integrated support for policy-db compilation * proto/db/smtp-policy.src, utils/policy-builder.sh: Sample policy database (header), and sample policy-db builder * scheduler/update.c, transports/libta/diagnostic.c: "ok3" status message for cases where transporter was able to relay the DSN into to the next link, thus we have no mandate to report successfull relaying, like "ok2" and "ok" would do. Fri Jun 13 22:41:04 1997 Matti Aarnio * Makefile.in: - Version 2.99.48p4 * doc/design/zmog.tex: - Corrected two typos, and were able to compile it with command "tex zmog" * lib/selfaddrs.c, libc/inet_ntop.c, libc/inet_pton.c, libident/identuser.c, router/libdb/bind.c, router/libdb/selfmatch.c, smtpserver/smtpserver.c, transports/smtp/smtp.c: Modified plain '#ifdef AF_INET6' into more complex: '#if defined(AF_INET6) && defined(INET6)' * transports/libta/diagnostic.c: Under some obscure reasons, autoconfig of the sprintf() return value did get it wrong at SunOS 4.1.4.. Coded the diagnostic without conditional optimization based on the return value type. * libc/getaddrinfo.c, libc/getnameinfo.c, include/netdb6.h: Pulled from 2.99.49beta10 -- with seriously rewritten DNS code. * transports/smtp/smtp.c: Allow talking to self host in case the destination port is a non-standard port -- not port 25, that is. Wed May 28 11:03:42 1997 Matti Aarnio * Makefile.in: - Version 2.99.48p3 * config.h.in, configure.in, configure: Pulled these from 2.99.49beta7: - timezone autoconfiguration - IPv6 in use with ONLY an explicite --with-ipv6 option! * transports/sm/sm.c, transports/smtp/smtp.c, transports/mailbox/mailbox.c: "readalready" cache handling in case the MMAP is not recognized properly (IBM AIX has problems at the auto- detection of that feature..) * smtpserver/smtpserver.c, smtpserver/smtpcmds.c, smtpserver/smtpdata.c: - Wrong parameters in one logging call at the BDAT processing, - Rearranged typeflush() calls, and pipeline status reporting. Fri May 2 16:10:28 1997 Matti Aarnio * Makefile.in: - Version 2.99.48p2 * BUGS Noticed problem at scheduler ETRN -> results in busy spin... * man/aliases.5: A new man-page... * proto/cf/fqdnaliases.c: Missing quotes from the script * proto/scheduler.conf: Sample scheduler.conf file with tags for three different channel = local -cases: - local/file* - local/pipe* - local/* The "local/*" may be run with "sm" to use procmail, or cyrus (for example) for deliveries to users. The "sm" (procmail in particular?) is unable to do pipe-feeds, nor storing to files, it seems. (Also gave an example of 'punt' on smtp..) * smtpserver/smtpcmds.c: Removed obsolete special case ESMTP token: X-TURNME * transports/errormail/errormail.c, transports/sm/sm.c, transports/mailbox/mailbox.c, transports/hold/hold.c, transports/fuzzyalias/fuzzyaliases.c, transports/libta/lockaddr.c, transports/smtp/smtp.c: "Make sure we are properly positioned at the start of the message body" (lseek() things) * transports/mailbox/mailbox.c: Save errno before doing library calls -- did yield wrong error in case the 'dotlock' was busy. Thu Apr 24 21:00:28 1997 Matti Aarnio * Makefile.in: - Version 2.99.48 - SiteConfig->zmailer.Config processing sed script modifications so that even HP-UX sed does it right.. * proto/cf/rrouter.cf: Test for a "|.+" was coded wrong, and it never matched for pipes causing stripping of possible ">" from within the pipes.. * router/router.c: - logit() routine written with different armouring against overlong input, which causes overlong syslog() output, which causes some syslogs to blow up.. * transports/libta/mime2headers.c: Properly flexible way to do DEFCHARSET definition in runtime also for "MIME-2" (e.g. headers) part. * transports/libta/mimeheaders.c: Missing "," from the canned "X-Warning" messages * transports/mailbox/dotlock.c: HP-UX has 'long gethostid()' defined in its includes. * transports/mailbox/mailbox.c: - Correction on long-term problem on PIPE running, where the pipes practically did not work at all :-/ - Expanded diagnostics (while debugging that previous bug...) * transports/smtp/smtp.c: - ssfgets() got a bit more state into SmtpState block - Most (nearly all) sprintf()s are output length limited with "%.200s" -type of format strings Wed Apr 23 17:58:19 1997 Matti Aarnio * Makefile.in: Version 2.99.48-970423 * configure.in: - Rearranged libresolv autoconfiguration - Test for possible need of -lresolv to satisfy the access to the structure: _res - Corrected the RFC822TABS (--without-rfc822-tabs) option - Do GCC optimizer resistant test on existence of external objects, e.g. char *sys_siglist[]; * include/ta.h, transports/libta/tasyslog.c: New parameter (msg) to the tasyslog() routine * include/zsyslog.h: Yield an error if the system does not have proper (TODO: Have acceptable substitutes for that case..) * libc/mail.c, smtpserver/smtpdata.c: The mail_close() routine is a wrapper for an internal routine that returns information regarding the last status of the spool file for the smtpserver to be able to report that at the message reception. * router/libdb/header.c, man/zmailer.3: Define the "rcvdfrom" to require a "user@address" type of parameter * smtpserver/smtpcmds.c: Store the "rcvdfrom" information from the IDENT information. * router/rfc822.c, router/router.c: If the system can't resolve internal uname->uid/uid->uname mappings (getpwnam() fails ?), then defer the message. If the message has a "rcvdfrom" header, and the message is not submitted by any of the "trusted" UIDs, replace the "rcvdfrom" information with a new one we fabricate * router/rfc822.c, router/rtsyslog.c: router's syslog (rtsyslog()) has a new parameter * smtpserver/smtpserver.c: - Don't syslog() anything, unless indicating an error! - Recognize IN6_IS_ADDR_V4MAPPED() type of addresses, and do reversal lookup in proper IPv4 manner. * transports/libta/diagnostics.c, transports/libta/tasyslog(): - notary_setwttip() to behave in similar manner to notary_setwtt(), but to define the IP address that the host has, and we have made a connection with. - tasyslog() takes that information as a parameter - tasyslog() has new sprintf() format strings with output width limitations * transports/libta/mimeheaders.c: Rewrote (somewhat) the text of the "X-Warning:" -lines * transports/mailbox/mailbox.c: - The recipient address is properly dequoted, and localized for feeding to the local processing (to a file, pipe, or to store at somebodys mailbox) - HP-UX 9.x oriented: 'fflush() returns an error in case the FP points to a PIPE, but lets hope it won't set ferror() unless it really is an error!' * transports/sm/sm.c: - Process bigger let_buffer[]s than before - If the system reads more than one bufferfull, always remember to zero the 'readalready'. * transports/smtp/smtp.c: - Report the notary_setwttip() along with notary_setwtt() data for the diagnostics - smtp pipelining code rearrangement completion * transports/libta/tasyslog.c: Don't be shy about telling what the return message of the transporter was when tasyslog()ing it, and the code was OK. Mon Apr 13 19:15:31 1997 Matti Aarnio * Makefile.in: Version 2.99.48 * BUGS: More TODO notes (and a few removed) * Makefile.in: More DISTCLEANFILES entries, * SiteConfig.in: New entry: DEFCHARSET=XXX * compat/sendmail/sendmail.c, include/zsyslog.h, ... Replaced all syslog() related things with our own include header, and z** -macroes. * config.h.in: IPv6 header tests, RFC822TABS definition. * configure.in: - PERL path test - Tests for several different (N|G)DB(M) formats - IPv6 related header tests - Test for the need of resolver library - Configure time option to say wether ot not to use TABs in the RFC-822 headers -- At tele.fi we have customers that CAN'T handle TABs :-( (The makers of the software with problems agree they made a mistake, but it has quite a bit of installations...) - Attempt to auto-detect TCP-Wrapper's tcpd.h, and libwrap.a from several possible locations. - Auto-generate several new files... * include/ta.h, transports/libta/lockaddr.c: New parameter: 'int mypid' * include/zmsignal.h: - SIGNAL_RELEASE() rewrite for BSD - SIGNAL_HANDLE*() with SA_NODEFER active! * lib/esyslib.c: erename() did miss ending 'return 0;' * lib/selfaddrs.c: - Better IPv4 self-address detection (IPv6 still partial..) * lib/taspoolid.c: - When presented a path, pick only the last part, not "/path".. * libc/getaddrinfo.c, libc/getnameinfo.c: - first stabs at the integration of these new ways to pick addresses.. * libident/identuser.c: - more and more adaptations for our smtpserver use (and to be generic IPv4/IPv6 ident query interface..) * smtpserver/*.c, smtpserver/Makefile.in: - Split the server in "zillion" bits from the original monolithic file - Integrated reception time character translation routines as implemented by: Eugene Crosser * router/Makefile.in: - New source module: rtsyslog.c * router/rtsyslog.c: Sendmail-like syslog messages * router/conf.c: Configure-time option to be able to control the presentation of TABs in the message headers. * router/db.c, router/libdb/selfmatch.c, router/libdb/Makefile.in, router/libdb/search.h: - New relation, which can be used to match local addresses in destination addresses of form: - The "relation" command presents lists of available database routines when it is called without arguments * router/rfc822.c, router/rtsyslog.c: - Present the addresses in properly quoted format in the transport-specification file - rtsyslog() -- sendmail-like syslog report (still partial) * scheduler/msgerror.c: New diagnostic scalar variable: time-of-last-attempt * scheduler/threads.c: Change in age-order comparison code, some systems had the qsort() to SIGSEGV.. * scheduler/transport.c: - Count '#idle' as a command. - FD_* -> _Z_FD_* macro renameing * transports/libta/Makefile.in: Missing semicolons * transports/hold/hold.c: - Process according to the HOST parameter too, if present! - Understand: NS/AAAA * transports/libta/ctlopen.c: - close-on-exec flag set on - Recoded the size estimate -- don't present the incoming message transport envelope size -- at nic.funet.fi it can be well over 100 kB, and the message itself is often mere 1kB or so.. Some systems choke on such a "large" messages.. * transports/libta/diagnostic.c: - Log the time of the diagnostics write to the TA-spec file into the diagnostics string * transports/libta/dnsgetrr.c: IPv6 things * transports/libta/mimeheaders.c: Runtime option on what the default character set is at the system (so that everybody need not to use UNKNOWN-8BIT, or ISO-8859-1, or ...) * transports/mailbox/mailbox.c: - Runtime option on default charset, when forcibly MIMEfying - Basis for properly localized "user" identity * transports/sm/sm.c: - charset system option processing * transports/smtp/smtp.c: - charset system option processing - smtp-pipelining processing with explicite flags on when the parameter is really a recipient address, and when the errors are severe, and when not.. - ssfgets() -- a timeouting version of fgets(), which when timeouting, sends a NOOP (or QUIT+close()) to the possibly open SMTP session - Various timeouts are processed with setjmp()/longjmp(), however NOT with POSIX sigsetjmp()/siglongjmp(), which in fact doesn't much help, unless you know for sure that you can accept the signal mask that the sigsetjmp() has saved for you... - Use my interface-derived name for the HELO/EHLO parameter - smtp_sync() internal logic refurnished - calling error of ranny() on the randomization of the MX- entries of same preferrence * proto/db/aliases.in, utils/autoanswer.pl.in: - Alias of 'autoanswer: "|$MAILBIN/autoanswer.pl"' - Autoconfiguring the runtime path of PERL - New routine to do automatic replying to whoever sent the original query Sat Mar 15 02:54:06 1997 Matti Aarnio * Makefile.in Version 2.99.47 * BUGS: Notes on what is missing on IPv6 front yet.. * config.h.in, configure.in: HAVE_SA_LEN -- BSD 4.4 sockaddr sa_len -field * configure.in: - Some rearrangements to try to autodetect tcpwrapper library in the system - Detect (and possibly provide replacements) of IPv6 basic API functions: inet_ntop(), inet_pton(), getaddrinfo(), getnameinfo(), gai_strerror(), freeaddrinfo() * include/libc.h, include/netdb6.h, libc/gai_strerror.c, libc/getaddrinfo.c, libc/getnameinfo.c, libc/inet_ntop.c, libc/inet_pton.c: Spare-copies of these routines in case a -linet6, or libc containing them is not in use... * compat/sendmail/sendmail.c: Ignore sendmail option -odb (as well as -obd)... * include/ta.h, include/dnsgetrr.h: Separate DNS query routine protos into their own file * libresolv/gethnamaddr.c: Modified entrypoints to provide parameter on which to return the TTL value of the queried data. * router/conf.c: Add 4th default name for possible "daemon" --> "daemons" on some newest SGIs.. * scheduler/msgerror.c: Moved repeated (big!) codesegment into separate procedure, and schrunk the error reporter a bit with that... * scheduler/transports.c: Oh my... _FD_SET et.al. are defined on some systems! Renamed them. * scheduler/update.c: Cosmetic changes actually.. (while debugging msgerror.c) * smtpserver/rfc821.c: New rfc821_v6dotnum() -routine. * smtpserver/smtpserver.c: - Bug in 2.99.46psomething, which broke ENVID= parameter parsing routine.. - Initial cut on IPv6 things * transports/hold/hold.c, transports/smtp/smtp.c, transports/libta/dnsgetrr.c: - First partial stabs at IPv6 Mon Feb 17 22:30:06 1997 Matti Aarnio * transports/mailbox/dotlock.c: A HPUX `gethostid()' routine from (David DiGiacomo) via Sven Goldt who found out why the mailbox program keeps crashing on the HPUX. Fri Feb 14 20:29:25 1997 Matti Aarnio * lib/esyslib.c: erename(), eqrename() Rewrote eqrename() cleanly in model of Andy Poling's ideas -- he observed surprising amounts of duplicate delivery while using rename(2) instead of link(2)/unlink(2) systemcalls for acquiring a lock on the messages. * transports/smtp/smtp.c: Classify "DNS inconsistency, No MX, no A" message a bit better. There are four main categories, and the "we are the best mx" could be divided still.. * libsh/interpret.c: Count on how many recursions we have done, and if exceed some limit (like 100), we are highly likely doing something seriously wrong. Abort before getting a SIGSEGV from stack overflow error. (All is rotten, but the diagnostics of the death is a lot easier...) * proto/cf/aliases.cf: Loop-cut 'expansions' testing keying was flawed, and a person with a '.forward' was able to cause a serious loop that lead to memory exhaustion in pretty short order.. Sat Feb 8 00:19:06 1997 Matti Aarnio * compat/sendmail/sendmail.c, configure.in, lib/esyslib.c, libs/Makefile.in, router/rfc822.c, smtpserver/relaytest.c, smtpserver/relaytest.h, smtpserver/smtpserver.c, utils/makendbm/makendbm.c, proto/scheduler.conf, ChangeLog: A heap of "minor" fixes to all around to improve this compilability in various systems besides my RedHat Linuxes. Thu Feb 6 19:00:24 1997 Matti Aarnio * Makefile.in: patchlevel 46 -- time for a release.. * router/functions.c, proto/aliases.cf, proto/crossbar.cf, proto/canon.cf: New router-zmsh routines: $(dequote .. ) $(condquote .. ) To provide support for having an alias entry like this: "The Space Head": st-tng (And simplify the scripts to remove special patterns tried for space-processing...) Wed Feb 5 10:02:20 1997 Matti Aarnio * utils/makendbm/makendbm.c: Runtime option (switch) -a to process alias-like input (keys are RFC-821-quoted strings) that follows format: keyname: LWSP value,value2,value3 Thus allowing feeding the output of printaliases routine to makendbm for creating some real dbase, instead of indirectly indexed ordered flat ascii file.. * libsh/sh.ssl: A new way to code white-space into '*sift' patterns by using paired quotes: 'text with " dbl quote'"' and single quote" Strings: -111111111111111111111--222222222222222222- There '-' are ignored, and chars of '1', and '2' are taken in. * proto/cf/aliases.cf, proto/cf/crossbar.cf: Recognize (with special pattern) that the local address contains spaces, and do add quotes to it! (Similarly for 'internet' rewrite rule..) * INSTALL, proto/cf/fqdnalias.cf, proto/newfqdnaliases.in, proto/db/fqdnaliases: A new routeing mechanism: fqdnalias * Overview: Added one new performance benchmark result * README.UPGRADING, scheduler/update.c, transports/libta/ctlopen.c, transport/libta/diagnostic.c: Due to a change at transports/libta/diagnostic.c, and its counterpart within scheduler/update.c, new transport agents might not work with old scheduler -- it is a matter of linking the TA with the new library. * compat/sendmail/sendmail.c, smtpserver/smtpserver.c: Do proper RFC821->RFC822 syntax quoting of possible backslashfull addresses: \rfc821@domain --> "\rfc822"@domain * config.h.in, include/syslog.h, libc/syslog.c HAVE_SYSLOG_H -- mark also if (or if not) we have in the system. * libident/identuser.c, include/identuser.h: A thorough rewamp of things -- to make sure no buffer overflows will occur due to a hostile data.. * include/libz.h, lib/rfc822scan.c, router/functions.c, router/rfc822.c, router/rfc822hdrs.c: When doing address list expansions, process each LINE of the input as a sole container of the address -- if there is some address like: "@foo.domain" in a line, previously this became catenated with the next line as: "@foo.domain:faa@fii" * include/ta.h, transports/libta/tasyslog.c: New generic tasyslog() to log out transport-agent report in similar manner to that of the sendmail(8). * proto/Makefile.in: Build/don't build some new/old dirs. Especially builds the hash-subdirs for the scheduler! * lib/Makefile.in, lib/taspoolid.c: New generic tasyslog() support routine to build unique spoolids by catenating a base-64 encoded mtime of the email contents file, and the inode number of that file. * lib/prversion.c: Year change -- Copyright 1992-1997 * proto/cf/aliases.cf: Handle all variants of list owner-, -owner, and -request address injections neatly. Does not set list outbound sender to be owner-listname, though! (See aliases-vger.cf) * proto/cf/rrouter.cf: - recognize addresses that start with a pipe character, and DON'T mess with them with the focus testings! - Handle backquote-prefix properly (i.e. strip it..) - Detect feed to files (or X.400 address..) - If the `fullnames' database, nor `newsgroup' database found anything, try handling it in aliases database. * router/libdb/headers.c: Moved two headers into "unknown" category, where we won't be checking on them: keywords, references * router/libdb/ordered.c, router/libdb/unordered.c: scan over the first token with skip821address() -- now I can do aliases with quotes on the left-side, and spaces in them! * router/rfc822.c: Bugfix: wrong pointer picked for recipient attribute causing more or less mysterious crash.. * router/rfc822hdrs.c: New form of the generated Message-Id: It is slightly shorter than the previous one, and definitely a lot more sortable.. * scheduler/msgerror.c: - Slight improvement on handling spacefull 'host' object (and 'user' object too..) - Threw away a 'dnsrecipient' information pointer, and updated error log writers/parsers for it. * scheduler/scheduler.c, scheduler/scheduler.h, scheduler/update.c: Will not keep 'contents' of the TA-spec file in memory for very long -- only up to the initial scheduling input of the message. * smtpserver/relaytest.h smtpserver/relaytest.c: Some more code -- "flesh on the bones", but not yet usable.. * smtpserver/smtpserver.c: Tried to alter the code to carry around a 'content' package so that perhaps someday the smtp-server can be threaded (or some such), and each thread runs with its own content. * transports/libta/diagnostic.c, transports/mailbox/mailbox.c, transports/smtp/smtp.c: Doing 'sendmail(8)' like syslog data preparation Sat Jan 25 15:18:08 1997 Matti Aarnio * libsh/interpret.c, libsh/tregexp.c: Trace SSIFT/TSIFT statements with clearly differentiated "trace compare/trace match" statements -- namely report "tcomparing 'pattern' and 'string'" for the TSIFT, and "scomparing 'pattern' and 'string'" for the SSIFT (and "smatched"/"tmatched" for matches likewise..) * proto/cf/rrouter.cf: Where that trace change was motivated... was something that I had implemented sloppily in a ssift statement... Namely I noticed, that the "fullname expansion traps ':include:/path/to.file' into itself, and fails". Then I had decided to implement a test for RE "^[^.:]+\.[^.]+$" (where the starting/ending limiters are implicite in ssift!) but that RE has a fault of matching only ONE dot -- so no "foo.bar.bat" is matched with it, while it correctly does not match cases where the part before that dot has a colon (from ":include:"). Now I test at first for explicite case of ":include:", and only then test for dotfull names. I am still uncertain if we should not rewrite it entirely differently -- to move also fullname/newsgroup map tests inside the $(routeruser ..) routine, and thus be able to allow sendmail -like aliases where dotfull names can be in the standard aliases as well. * smtpserver/smtpserver.c, proto/smtpserver.conf, smtpserver/Makefile.in, smtpserver/relaytest.h, smtpserver/relaytest.c: Some more coding of multi-db-format query lookups of service limitation mechanisms. * utils/makendbm/ndbmlook.c: Yet another parametrization error in the DB code.. (A non-fatal one, I presume.. affected only GDBM) Sat Jan 18 07:11:07 1997 Matti Aarnio * Makefile.in: Patchlevel 45 * Makefile.in: Command: "make install-bin" will not overwrite existing ZCONFIG file! * compat/sendmail/sendmail.c: Accept (and silently ignore) "-ob*" -options. * configure.in, config.h.in: - Have --with-yp -option - Test for res_init() instead of gethostbyname() to see if the system provided libraries are enough, or if more are needed (like -lresolv) * libsh/interpret.c: - fapply() to accept up to 30 parameters in a linked list * proto/Makefile.in: Remove the creation of "MAILBIN/bin/", which is about same as creating "MAILSHARE/bin/bin/", which is stupid... * router/libdb/ordered.c: - test mmap() error return in a bit more generic way - scan the buffer for the number of lines by executing a bit better constrained scan - than the previous one was.. * router/shliaise.c: - Got rid of svstack[] arrays in l_apply(), and s_apply() routines * scheduler/Makefile.in: - If using TCP-Wrapper stuff, have related INCLWRAP, and LIBWRAP definitions in use -- for the mailq access control Tue Dec 24 00:23:32 1996 Matti Aarnio * all transporters: Process SIGQUIT as marker for desire of termination. That is, equal to earlier SIGTERM behaviour, but with a bit less urgent death. * scheduler/scheduler.c, scheduler/threads.c, scheduler/transport.c: An attempt to create an orderly shutdown; send the scheduler signal SIGQUIT will activate this behaviour. * Makefile.in: When "make dist", at first do "chmod -R a+rX" to turn on all read-, and some x-bits. * SiteConfig.in, configure.in: Autoconfigure preferred DBTYPE, and default extension for it. From SiteConfig.in it will propagate to /etc/zmailer.conf (or where ever that one is..) * compat/rmail/Makefile.in, compat/sendmail/Makefile.in, router/Makefile.in, scheduler/Makefile.in, smtpserver/Makefile.in, transports/mailbox/Makefile.in, transports/sm/Makefile.in, transports/smtp/Makefile.in, utils/makendbm/Makefile.in, support/utils/Makefile.in: Autoconfigure MAILBIN= -entry (for installing) * utils/makendbm/makendbm.c: Corrections for BSD DB B-tree database parametrization. * proto/cf/aliases.c: Instead of using memory-hog script of (listaddresses ...), do use the builtin (listexpand ...) routine. * proto/cf/rrouter.cf: Do "ssift/in/tfiss" test on $address. * proto/smtpserver.conf, smtpserver/smtpserver.c, smtpserver/Makefile.in, smtpserver/relaytest.c, smtpserver/relaytest.h: - On STREAMs machines accept several rather mystic error states for accept() -- with a shrug.. No need to exit(1) on such things as ENOSR, or EPROTO (No STREAMS Resources, Error on STREAMS PROTOcol, ...) - Don't log to console ( openlog( ... LOG_CONS ... ) ) - Accept any input - any at all - for the HELO parameter. Be a bit more picky with EHLO... - Relay-control facility -- to whom we accept email to (must be out customer, or the sender is our customer) - Introduce new "PARAM" tokens -- several of them: - "help" - string for the HELP-response main-text, up to 20 lines - "maxsize" - max-size we accept - "reject-percent-kludge" - flag - "accept-percent-kludge" - ditto - "relaytargets" - db-file definition - "relaycustnets" - ditto the rest - "rejectnets" - "rejectsource" - "rejecttarget" * router/libdb/bsdtree.c, router/libdb/gdbm.c, router/libdb/ndbm.c, router/libdb/unordered.c: Modification test trips when the open file has link-count of ZERO -- somebody mv(1)ed a new database on top of the old one! * scheduler/msgerror.c: - An armour against WTT being NUL (upgradeing something?) - Recognizing a situation where we are sending an error report on error report -- marking that with altered subject * scheduler/scheduler.c, scheduler/threads.c: Improve the freeze-out mode, it was able to spawn kids in the freeze-out mode, though it did not feed them. * scheduler/transport.c: Alter mailq() printout fork ending to be _exit(0) * transports/errormail/errormail.c: Use the 'host' parameter if one is supplied. Always pass it to the diagnostics subroutines. * transports/mailbox/mailbox.c: - Fix processing the "lastchar" things. - append_header() without extra "\n" * transports/sm/sm.c: - append_header() without extra "\n" * transports/smtp/smtp.c: New options: -F punthost -- Send all mail to denoted host -L localname -- At multi-homed machine, bind() to named interface before doing connect() Tue Dec 3 19:17:28 1996 Matti Aarnio * smtpserver/smtpserver.c: Missing a couple of fflush()es -- moved the EXPN and VRFY routines into separate subroutines. Now have ONE fflush() after each of the calls. Mon Dec 2 18:38:09 1996 Matti Aarnio * utils/makendbm/makendbm.c: Yet another parameter error in case the system can do both NDBM and BTREE -- autch! * smtpserver/smtpserver.c: Be silent about log-files when started up as "/usr/lib/sendmail -bs" by somebody who can not write into the default log files. * Makefile.in: Version 2.99.43 * router/libdb/bind.c, transports/libta/dnsgetrr.c, transports/smtp/smtp.c, transports/hold/hold.c: Uses of "sizeof(u_long)" have been eliminated (as well as "sizeof(u_short)" in favour of numeric constants: 4 and 2 respectively. * transports/libta/mime2headers.c: Disabled the new MIME2 code. Nothing but grief... :-( * transports/libta/mimeheader.c: One missing setting of a variable. * transports/smtp/smtp.c: Missing space at DSN RET= parameter writing. Wed Nov 27 21:23:12 1996 Matti Aarnio * transports/smtp/smtp.c: Once more (SmtpState -block).. Clear it with memset(), then init those few non-zeroes. * smtpserver/smtpserver.c: Finally fixed the VRFY and EXPN behaviours; now they work as they should. Tue Nov 26 15:17:03 1996 Matti Aarnio * transports/smtp/smtp.c: SmtpState -block needed some more initializations. Missing one: verboselog Mon Nov 25 16:21:15 1996 Matti Aarnio * transports/hold/hold.c: Calling convention of getrrtype() was changed (a bit), and I forgot to update it here too. Thu Nov 21 20:37:19 1996 Matti Aarnio * Makefile.in: Version 2.99.42 * transports/smtp/smtp.c: Massive rewrite of calling conventions within the SMTP client -- passing around a state-variable containing state of the smtp session (though not the message..) Corrected one bug at PIPELINING code. * transports/mailbox/mailbox.c, transports/sm/sm.c, transports/smtp/smtp.c, include/ta.h, transports/libta/mime2headers.c: Routine headers_to_mime2() got new parameter. * proto/cf/aliases.cf: Yet couple more fixes -- when processing .forward an empty (addressless) file needs to expand to the original user, NOT to "\$user" -> $user ... * scheduler/mailq.c: Explanation on "Transport queue is empty" condition. * scheduler/readconfig.c: Now configuration file can have spaces around parts of attribute tokens: 'name=value' 'name = value' Unquoted '#' causes end of line to be comment, and thus ignored. * scheduler/scheduler.c, scheduler/threads.c, scheduler/update.c: A "-p" (procselect) option to run only selected threads, but not to log in possible error reports (debug stuff) * scheduler/scheduler.c: If a "TURNME"-request is processed, and the request does not have acompanying entry in "queue" (cfp->mid is not set), process it gracefully, and unlink(). * smtpserver/debugreport.c: Corrections on type() parameters. * smtpserver/smtpserver.c: More informative reports on message exceeding fixed limit * transports/errormail/errormail.c: Builtin error message fix * transports/libta/mimeheaders.c: Make sure the Content-Transfer-Encoding is converted from Q-P to 8BIT when it is in that value.. * transports/mailbox/mailbox.c: - Support 'dirhashes' (runtime command-line option: -D) - If the message has no 'To:' header, add sendmailish 'Apparently-To:' -header. - The sendmailish "Return-Receipt-To:" processing produces now proper and up to date MULTIPART/REPORT * transports/sm/sm.c: Produce proper DSN return values for various subprogram error messages. * utils/makendbm/makendbm.c: Correction on 'dbasename' variable specification Thu Nov 14 22:37:18 1996 Matti Aarnio * router/rfc822.c, router/libdb/headers.c, BUGS: Treat "Return-Path:" -header with respect -- and DELETE it when sending it thru the system. It is stored into the user's mailbox by the FINAL DELIVERY AGENT, it does not exist outside user's mailbox! Wed Nov 13 16:48:00 1996 Matti Aarnio * proto/cf/aliases.cf: Aargh... quotation-rules on ZMSH are "a bit" different from SH, and I keep forgetting it. Everything broke when I added some newlines here & there... Now added '\' (back-slashes) to cover them. * utils/makendbm/makendbm.c, utils/makendbm/ndbmlook.c: Got rid of the need for some referred things, uses strerror() instead. * transports/mailbox/mailbox.c, libc/mail.c, scheduler/msgerror.c: Use fsync() when finishing off the write to the mailbox/file. * transports/libta/ctlopen.c: Initialize notifyflgs properly for the case that no NOTIFY parameter is present (assume: FAILURE) * transports/errormail/errormail.c: If no reports sent, generate still OKs for each non- generated report.. * transports/sm/sm.c: Handle "channel error" at the sender.. Tue Nov 12 17:47:34 1996 Matti Aarnio * config.h.in, configure.in: Test for fsync() at the system. Mon Nov 11 18:56:21 1996 Matti Aarnio * Makefile.in: Released version 2.99.41 * SiteConfig.in: Some new default values; FORCEPUNT, SCHEDULEROPTIONS * compat/rmail/rmail.c, compat/sendmail/sendmail.c: Re-coded "to " and "todsn " -envelope generation, added "env-end" -envelope header. * libc/strsignal.c: Using AC test SYS_SIGLIST_DECLARED * include/ta.h, transports/libta/skip821address.c, scheduler/scheduler.c, transports/libta/ctlopen.c: Declare and use skip821address() -routine. Enables to have embedded spaces in 'host', and in 'user' parts of the parameter line. * lib/loginit.c, router/router.c, scheduler/scheduler.c: "Name-space pollution" -- "log" is reserved name, and cannot be used as a name of a global variable.. * libc/mail.c: Final corrections for automatic filename index array reallocing -- off-by-one, and 32/64 bit troubles.. * libc/syslog.c: Compile it always in -- somewhat difficult at SunOS 4.1.4, but it seems to work even when debugging fails.. Also uses *sNprintf -routine, that is one with explicite way to tell the max space at the buffer. * libsh/io.c: #define NO_FLOAT_CONVERSION -- don't do FP conversions, can get rid of the need to use "-lm" at the router at DEC UNIX 4.0 ... * proto/cf/TELE-FI.cf: Sample of "router.cf" we use at Telecom Finland * proto/cf/aliases.cf: Renoved aliases.cf that can do PUNTHOST AFTER checking at system aliases, but it will NEVER look at user home directories... * proto/forms/delivery: Changes at the message text (and subject). Now it is more generic to apply also for positive acks. * proto/scheduler.conf, proto/sm.conf: Sample entries for using CYRUS IMAP-server injector as local server (Tom Samplonius ) * router/db.c, router/libdb/bsdbtree.c, router/libdb/search.h: Additional DB engine: BSD LIBDB B-Tree * router/db.c: At "pathalias" -driver one can have (at $MAILSHARE/db/routes) "." to denote "match everything on this", and thus to make outgoing routes with single entry at routes db: . smtp!firewall * router/functions.c: A glitch-fix for run_expandlist() -routine generating NOTARY ORCPT data * router/libdb/ndbm.c: "extern int errno" (SunOS 4.1.4) * router/rfc822.c: Fix on detecting ERRTO pointer value change, and moving "eVerbose" envelope header detection a bit up, plus adding several "verbose-log" printouts. * scheduler/msgerror.c: - New routine decodeXtext() to decode XTEXT objects coming as DSN parameters. - Some sanity-checks added for error report address picking. (core-drops at vger.rutgers.edu) * scheduler/scheduler.c: - Lowercase the whole input string with "channel host", then later sort those lines with strcmp() instead of cistrcmp() ... - Fix parsing the NOTIFY= parameter -- wrong index.. * scheduler/update.c, scheduler/scheduler.c: Unconditionalize the syslog() * smtpserver/smtpserver.c: - Moved code around to have definitions before uses. - Fix one call-variant of "type()" definition. * transports/errormail/errormail.c: - Decode XTEXT - Produce DSN definitions on the reports with NOTIFY=NEVER.. * transports/libta/writeheaders.c: Always double-dot-quote the headers, though it should never be needed... * transports/mailbox/mailbox.c: Present POSITIVE DSN reports! * transports/sm/sm.c: - Unless using option "H", convert headers to "MIME-2" - Generate several possible variants of BSMTP, which one can thus feed via UUCP to remote system * transports/smtp/smtp.c: Write out in a bit larger chunks than 1 k to the network.. * utils/makendbm/makendbm.c: Support 3 different DB-formats: NDBM, GDBM and B-tree ALL POSSIBLY AT THE SAME TIME * transports/libta/mime2headers.c: Debugged the REAL MIME-2 header producer to produce some real MIME-2 format headers.. Mon Nov 4 22:45:37 1996 Matti Aarnio * Makefile.in: Released version 2.99.40 * compat/sendmail/sendmail.c, include/mailer.h, router/rfc822.c, router/router.c, scheduler/msgerror.c, scheduler/scheduler.c, smtpserver/smtpserver.c, support/vacation/vacation.c, transports/libta/warning.h Unconditionalized syslog() -- we have one bundled in from BSD sources. * configure.in, config.h.in, and several sprintf()s all over: We test for possible snprintf(), and will use it! * configure.in, libc/setreuid.c, libc/Makefile.in: Build setreuid() unconditionally, because at AIX the libc has one, but it is rather broken.. * transports/Mailbox.in, transports/fuzzyalias/ A fuzzy-match transporter for creating suitable error reports. Made by Thomas Knott. * include/ta.h, router/rfc822.c, router/shliaise.c, scheduler/msgerror.c, scheduler/scheduler.c, scheduler/scheduler.h, scheduler/update.c, transports/libta/ctlopen.c, transports/libta/diagnostic.c: DSN notary parameter processing in smarter ways * lib/token.c: - copyToken(): If the token size is over 50k chars, copy the first 50k of it. - printdToken(): "print" the token into limited size buffer, but will expand the buffer if needed. * libc/mail.c, libc/mail_alloc.c: Have dynamically allocated list of mail-files associated with the mailbox files being created. * libsh/io.c: Initialization of std_printf variable a bit difficult. * libsh/listutil.c: s_equal1() -routine. Matches only one conscell at any time. * libsh/tregexp.c, router/shliaise.c: Prepared to use printdToken() into limited size buffer. Will expand the buffer if needed. * router/functions.c: DSN parameter processing at run_listexpand() routine. * router/libdb/yp.c: Fix -- has to strsave() a string returning from YP. * smtpserver/loadaver.c: - "-lkvm" using systems (Solaris 2.5 on report) do not compile without it. - Usage of kvm_open() et.al. does actually need closure of /dev/kvm for not to overflow scant memory reading resources. * smtpserver/smtpserver.c: Error in type821err() routine -- forgot to report without "\r\n" at the channel output. * transports/mailbox/mailbox.c: Support "Return-Receipt-To:" -header when running mailbox with option switch "-S" (a.k.a: Sendmailism) * transports/smtp/smtp.c: - revise return status codes - recognize relaying of DSN armoured SMTP message to systems without DSN facility - recognize remote declaring ENHANCEDSTATUSCODES, and use them in our own report Tue Oct 22 18:51:07 1996 Matti Aarnio * Makefile: Version 2.99.39 * transports/libta/ctlopen.c: Parser bugfixes, and DSN data parsing * transports/mailbox/mailbox.c, transports/sm/sm.c: DSN stuff, diagnostic response fixes * transports/smtp/smtp.c: DSN stuff for sending, and for receiving. * transports/errormail/errormail.c: Use DSN parameters to the fullest we can. * transports/hold/hold.c: DSN parameters, fixes on returned diagnostics * scheduler/msgerror.c, scheduler/scheduler.c, scheduler/scheduler.h, scheduler/update.c: * router/libdb/header.c: "notaryret" envelope entry * router/shliaise.c: Generic ``build_gensym()'' routine to bring together stuff previously duplicated all over. * router/functions.c, router/prototypes.h, router/rfc822.c: The ``listexpand''-function got more DSN flags, and builtin MANDATORY ORCPT data generation. Basic rfc822() processor got more NOTARY stuff * include/mail.h.in, include/mailer.h, include/ta.h: NOTARY DSN stuff * libc/mail.c: Changed name (and params) of mail_close_scheduler() to be more generic mail_close_alternate() * compat/sendmail/sendmail.c, man/sendmail.8: More options to match with sendmail-8.8: -N -R -B -U -V (NOTARY DSN stuff) * INSTALL: A bit more verbose explanation on /etc/group, and "zmailer"-group, and purpose of "nobody" -account. * smtpserver/smtpserver.c: Fixes to: - ident processing - IETF NOTARY DSN mechanisms -- will ALWAYS create ORCPT= entry! Does an in-depth syntax analysis of DSN parameters * libident/identuser.c: Changes at the return values at various error conditions * transports/libta/ctlopen.c: Fix on DSN NOTARY parameter scanning Thu Sep 12 22:19:58 1996 Matti Aarnio * configure.in, Makefile.in, utils/makendbm/makendbm.c, utils/makendbm/ndbmlook.c, support/vacation/vacation.SH: A RedHat 3.0.3 Linux with some additions from more recent "Rembrandt" release caused several changes into the autoconfig, and makendbm system. Fri Aug 23 11:32:43 1996 Matti Aarnio * transports/mailbox/mailbox.c: Solaris mailloc() became tested (and fixed) when I was teaching utu.fi UNIX admin to handle the ZMailer.. * configure.in, configure: Forgot to move certain resolver related defines into a bit more generic place (Solaris test results) * lib/selfaddrs.c, lib/Makefile.in: Moved stachmyaddresses(), and matchmyaddresses() from smtp.c into a bit more generic location -- library, and uses ZENV variable SELFADDRESSES to list COMMA SEPARATED the names (or IP address literals: 1.2.3.4) of the local host. The goal is to get a list of all interfaces at the current host so that, SMTP knows them, and at MX selection recognizes local system. Mon Aug 19 19:02:19 1996 Matti Aarnio * (all over the place): lseek() offset parameter changed from long to off_t. This affects especially BSD/OS. * compat/sendmail/sendmail.c: Sometimes the "-v" option processing does not stop at all, because some transport agents (?) write more to the verbose logfile AFTER the scheduler has written its own "scheduler done" line so that it is not the last line read from the file. Sun Aug 18 19:10:44 1996 Matti Aarnio * INSTALL, README, README.PERFORMANCE, README.UPGRADEING, SiteConfig.in, man/*.[138]: Documentation update * Makefile.in: Version 2.99.37 for release * compat/sendmail/sendmail.c, libc/mail.c, smtpserver/smtpserver.c, router/libdb/ndbm.h, scheduler/transport.c, router/db.c, transports/libta/ctlopen.c testing check * router/functions.c, router/rfc822.c: Add "env-end\n" into the boundary of envelope, and message body. * router/rfc822.c: Add a bit more smarts (copied from the scheduler) at reading recipient addresses from canned messages. * include/ta.h, transports/libta/lockaddr.c, transports/libta/ctlopen.c, transports/libta/diagnostics.c: Parameter changes at lockaddr() * libc/whathost.c: More fixes at more systems (and less memory leakage) * libmalloc/externs.h, libmalloc/getmem.c, libmalloc/malloc.h, libsh/io.c: (back)porting to SunOS 4.1.4 with oddly configured gcc-2.6.3 It did indeed spew error messages at some files... * scheduler/mailq.c: Initial preparations for two-way interactive querying of system status * scheduler/mailq.c, scheduler/msgerror.c, scheduler/prototypes.h, scheduler/scheduler.c, scheduler/scheduler.h, scheduler/threads.c, scheduler/transport.c, scheduler/update.c: Scheduler work-spool can now be hashed into one, or two levels of indirect hashesh. See ``-H'' option at man-page. * smtpserver/smtpserver.c: ENCHANGEDSTATUSCODES -facility. * transports/mailbox/mailbox.c: - Runtime configurable selection of locking methods. - Corrected a file creation race condition * transports/smtp/smtp.c: When doing PIPELINING, use some additional smarts as to when there is any need to do system calls -- SMTP commands are sent in larger chunks, and replies are not to be expected before they are sent, nor forever after.. (Just reducing the number of syscalls when using PIPELINING.) Fri Aug 9 15:13:25 1996 Matti Aarnio * Makefile.in: Version 2.99.36 for release (pre-release installed at: Linux/ELF, DEC OSF/1, Solaris 2.5) * router/libdb/ndbm.c, utils/makendbm/makendbm.c, config.h.in, configure.in: Macro/function dbm_error() is not available at BSD DB compability package. Have to use errno in that case. * libs/Makefile.in, router/Makefile.in, router/libdb/Makefile.in: Changed the name of the router libdb to be libzmdb.c, because BSD DB library name is 'libdb.a', and linking with '-ldb' causes a name-space collision... * scheduler/qprint.c, transports/mailbox/mailbox.c: Changed the precedence of 'HAVE_UTIME' to preceed test of 'HAVE_UTIMES' -- or rather, if system has both, just using the 'HAVE_UTIME' variant. * libc/whathost.c: Completed the rewrite with minimized memory leakage. * smtpserver/fdstatfs.c: Completed the autoconfig sensing of the system * configure.in, config.h.in, router/libdb/ndbm.c: More tests at variations of different DB libraries. Especially supporting systems with BSD DB in disguise of NDBM ( RedHat Linux 3.0.3 ) Thu Aug 8 17:27:00 1996 Matti Aarnio * all over the place: #ifndef strchr extern char *strchr() ((and friends)) #endif IBM AIX (compilers) are ``different'' again... * libresolv/*.[ch]: Done 4.9.4-patch1 things to the resolver. * Makefile.in: - Version 2.99.35 - When makeing 'all', at first removed 'libs/libtag', and then proceeds with system creation, thus ensuring that libraries are remade -- ONCE -- if needed. - compiles support/vacation/vacation too * libc/whathost.c: AIX mntctl() things (modelled after GNU fileutils 3.13 code for similar use..) * libc/Makefile.in, libc/whathost-test.c: Now have easier to use test to see, if the NFS-server recognition works at various platforms. * support/vacation/*: Autoconfiguration support for the bundled vacation. Installation (make install) is not really mature. * configure.in config.h.in: - Sense 'timezone' variable with 'altzone' variable at system libc - Sense sizes of (void*) and (long) - Sense gethostbyname() at system libc without any other libraries - Corrected the sense of NDBM at -ldbm -library * lib/allocate.c: Alignments per system native parameters: - SPARCs always 8 bytes (double floats) - other 32-bit machines in 4 bytes - 64-bit machines in 8 bytes * lib/rfc822date.c: Timezone recognition when there is ``timezone'' variable, but not ``altzone'' (POSIX said some people.. Linux has this) * libmalloc/align.h: Recognize in a bit more generic way 64-bit machines (SGIs) * libsh/Makefile.in: LIBDEB fixup * libsh/builtins.c, libsh/execute.c, libsh/expand.c, libsh/interpret.c, libsh/listutils.c, libsh/regex.c, libsh/sslwalker.c, libsh/variables.c, libsh/zmsh.c, router/db.c, router/functions.c, router/libdb/bind.c, router/libdb/dbm.c, router/libdb/header.c, router/libdb/ndbm.c, router/libdb/gdbm.c, router/prototypes.h, router/rfc822.c, router/rfc822walk.c, router/router.c, scheduler/scheduler.c, scheduler/transport.c, smtpserver/rfc821scn.c, smtpserver/smtpserver.c, ssl/ssl.c: IRIX 6.1 compiler was rather picky... Minor cosmetics actually (like mismatch of enumerated variable vs. ints) * bin/mkdep.in, bin/mklibdep.in: Autoconfig stuff (CPPDEP substitution) Fri Aug 2 11:29:39 1996 Matti Aarnio * libresolv/portability.h: One test to add SVR4 sensing at Solaris 2.4 (with GCC 2.7.0) * Makefile.in: Version 2.99.34 -- compilation tested at: - Solaris 2.4 (without bundled libresolv!) - DEC OSF/1 * (many files): bzero() changed to memset() for SysVR4 port issues (Solaris 2.4) EXCEPT AT LIBRESOLV! * libc/getdtblsiz.c, libc/setreuid.c: Portability things for Solaris 2.4 without UCBLIB! * configure.in, config.h.in lib/rfc822date.c: Testing for variable 'altzone', and tm_gmtoff field at struct tm. Made also small test program into lib -directory (make rfc822date-test) Thu Aug 1 15:05:53 1996 Matti Aarnio * Makefile.in Version 2.99.33a2 (and good night folks..) * compat/sendmail/sendmail.c: Removed "Boptcount = 0", which was accidental artefact from one earlier version of "-B" -option processing. (compabilities with Alman's sendmail) Processing of SMTPOPTIONS zenv-variable needs to remove the quotes from the parameter -- but it is a bit dangerous in case there is something like: SMTPTOPTIONS="-l '/some/path/with space'" lets hope not... * configure.in: The dup2() -test appears to act as compilation environment checkup :-) Great, added comments of that effect at the error message. * smtpserver/smtpserver.c: Tested the built-in tcp-wrapper at nic.funet.fi, now it works as expected. (daemon name: smtp-input) FYI: The allow/deny rules are as follows: 1) does it match at allow ? Yes, return allow. not ? proceed to second test 2) does it match at deny ? Yes, return deny. not ? return allow. * router/rfc822.c: sequencer() transports/mailbox/mailbox.c, scheduler/msgerror.c, scheduler/scheduler.c, scheduler/scheduler.h: One particular weird mail file was able to cause 100% mortality rate at the router -- NULL pointer dereference.. (Andy Poling ) That prompted a larger go-around and checking at what happens when SMTP input arrives as MAIL FROM:<> and is destined somewhere. Fixes made at: - router - scheduler - mailbox In need of checking/fixing: - sm - errormail - hold (ok, perhaps only the 'sm') * Makefile.in Version 2.99.33a1 * config.h.in, configure.in: Created --with-tcp-wrappers=DIRPATH option, changed names of two others to: --with-sendmailpath=PATH --with-rmailpath=PATH * smtpserver/smtpserver.c, smtpserver/wantconn.c: Wietse Venema's wantconn.c for builtin tcp- wrapper behaviour -- suboptimal in efficiency, though... (but better than: inetd -> tcpd -> sendmail -> smtpserver, which is 1 fork + 3 execs..) The magic name at hosts.deny/hosts.allow files is: smtp-input * smtpserver/smtpserver.c: Fixes to file-descriptor usage so that things should work just fine with all options.. (*should*..) * router/libdb/header.c: Made "Keywords:" -header unknown per regards of the router -- it need not to touch, not comment on it. * compat/sendmail/sendmail.c, include/mailer.h, router/libdb/header.c, smtpserver/smtpserver.c: Process properly the -Bbodytype parameter, and store it into the message envelope. It is known to the router, though nothing is done with it. Also the SMTP-server does store that info into the message envelope. Also for "sendmail -bs", when invokeing smtpserver, feed it the arguments from SMTPOPTIONS environment variable ( via getzenv() ) Wed Jul 31 12:03:01 1996 Matti Aarnio * transports/smtp/smtp.c: Minor fixes at how the smtp_sync() works with the PIPELINING server -- now the reports do make sense to the ordinary message sender. (well, not exactly) * Makefile.in: Version 2.99.32 -- and releasing it * *Makefile.in (practically them all) Adjusted dependency processing, now can do 'make depend' any time, though needs GCC-based CPPDEP="gcc -MM" to work properly... * scheduler/resources.c: Autoconfigured various resource update mechanisms. Tue Jul 30 11:36:24 1996 Matti Aarnio * Makefile.in: Releasing 2.99.31 -- let others to have the experience too.. * INSTALL: Giving some additional tips at the initial configuration * lib/linebuffer.c, router/rfc822.c: Improved the error detection, and recovery so that now non-conformant input is a reason to stop the parse of headers (or envelope), and everything after it goes to the message body without looseing anything of it. The envelope header "env-end" has special meaning: It ends parse of envelope, and after it, makeLetter() routine expects to get RFC-822 headers. If it does not, the input goes to the body. * libsh/io.c, smtpserver/smtpserver.c, libc/syslog.c, transports/libta/diagnostics.c, config.h.in, configure.in, transports/smtp/smtp.c, transports/libta/mimeheaders.c: HAVE_STDARG_H, and all related stuff, in case the varargs is not supported.. Mon Jul 29 01:15:13 1996 Matti Aarnio * configure.in libs/Makefile.in (and most other)Makefile.in Made several new --with-* -configuration options: --with-system-malloc --with-bundled-resolver Neither are defaults. Test for existence of /, to do yet: converting sources to use the knowledge.. Fri Jul 26 10:06:30 1996 Matti Aarnio * Makefile.in: Version number 2.99.31 The package has now compiled smoothly at: - Solaris 2.5 with Sun unbundled cc, and gcc-2.7.2 - Sun Sunos 4.1 with cc - DEC OSF/1 v3.2D with 'cc', and 'cc -migrate' - i486 Linux 2.0.4/libc-5.4.2 with gcc 2.7.2 * libident/Makefile.in, libident/identuser.c: Canonized the parameter defines for that library module * configure.in, libresolv/Makefile.in, libresolv/inet_ntoa.c Added that routine into resolver library package, as it is not always present in the system, and we might need it -- especially in format where it is guaranteed to be structure passing convention compatible with our other code... * libresolv/Makefile.in: Sun Solaris 2.5 does not like 'ld -x -r'.. Me neither, actually. Commented it away. * libresolv/herror.c: One missing forward declaration. Thu Jul 25 02:59:13 1996 Matti Aarnio * Makefile.in, configure.in, and related autoconfig stuff: Fixes to get autoconfiguration to work at Sun Solaris, DEC OSF/1 v3.2D, Linux (kernel 2.0), and SunOS 4.1. Running this in production mode at nic.funet.fi Released version 2.99.30 for public * router/rfc822.c: Revised the run_rfc822(), makeLetter(), and sequencer() way to handling squirrel()/defer()/reject() Created a new standard error form: loopexceeded which is given when there are way too many "Received:" lines at the message (50 is our limit; router/conf.c) * libresolv/ Copied pretty much as is the bind-4.9.4-REL/res/ (resolver) subdirectory, and relevant headers into our local "include/" -tree. Then edited those C-files a bit, and got system independent resolver, which will be loaded statically, of course. Wed Jul 17 16:49:02 1996 Matti Aarnio * (all over the place..) Makefile.in: Releasing version 2.99.29 with a bit less warnings, and more usability -- it very nearly works at my Laptop now... Mon Jul 15 21:42:40 1996 Matti Aarnio * Makefile.in: Releasing version 2.99.28 with dire warnings! * configure.in, configure, Makefile.in, ... (nearly all files got touched!) Using GNU-Autoconfig system Wed Jun 19 15:18:03 1996 Ken Pizzini * smtpserver/smtpserver.c, smtpserver/rfc821scn.c, router/rfc822.c: - SMTP-server dying with SIGALRM without cleanup - Accept braindead SMTP-input, and correct it "" (trailing white-space cleanup) - bug in router/rfc822.c at recognizing the end of the envelope headers, and begin of rfc822 headers in case the line-end was CR LF instead of plain LF .. Tue Jun 11 14:34:05 1996 Jerzy Sobczyk * libsh/execute.c: Change at the ENDGRANDCHILD() -macro to clean (reap) up the intermediate child process from becoming a zombie. Mon Jun 10 11:35:41 1996 Matti Aarnio * INSTALL: Some additional documentation details as suggested by Arnt Gulbrandsen. * smtpserver/smtpserver.c: Revised a bit of message logging, and other things. There were difficulties that manifested at Sun Solaris, while same code worked at DEC OSF/1 :-( (At Solaris the server process simply died a horrible death..) Tue Jun 4 18:52:08 1996 Matti Aarnio * libsh/interpret.c: Removed one setlevel(MEM_SHCMD,...) that caused memory to be freed too agressively in some cases, and which in turn was able to corrupt memory, when it was taken into use a bit later: for x in $(/bin/cat /etc/group); do ... ; done Fri May 31 15:30:01 1996 Matti Aarnio * smtpserver/smtpserver.c: Re-implemented the "TURNME"-command. Apparently I have misplaced the previous version source.. well, this is cleaner code anyway. Added an alias "ETRN" for the "TURNME". * lib/mail.c: mail_close_scheduler() A variant of mail_close() routine, which sends the file straight to the scheduler. (For "TURNME") Tue Apr 23 14:14:55 1996 Matti Aarnio * compat/sendmail/sendmail.c: Add ignorance of "-B8BITMIME" into option processing... (We do it always, no need to tell it at the call..) (sendmail 8.7) Mon Apr 22 08:15:20 1996 Matti Aarnio * scheduler/scheduler.c: Got the ``mytime()'' -server running also with Linux, and all other systems that have at least an ability to mmap() a file. Implemented a ``_CF_TURNME'' primitive parsing. * smtpserver/smtpserver.c: Implemented a "TURNME"-command, which produces ``_CF_TURNME'' primitive for the scheduler -- and sends it straight to there without going via the router.. Fri Apr 19 16:13:34 1996 Matti Aarnio * transports/mailbox/mailbox.c: At putmail(), move the newline addition testing AFTER the appendlet() processing. That way the message is always added at the end of the file, and only 0-2 newlines are added after that at the end of the message -- if needed. At least ELM got mystified with the earlier processing, where the newlines were added at the front of the message. * scheduler/scheduler.c: (the "time server") A bit more generic approach, works at at bit more platforms, than only at OSF/1, and SunOS 4.1.x ... Tue Apr 9 11:39:12 1996 Matti Aarnio * Makefile: Released version 2.99.27 * Makefile: Add a ".PRECIOUS:" -tag for produced Makefiles, thus avoiding destruction of them at some awkward moments, when pressing "Ctrl-C" would delete one of them.. * compat/sendmail/sendmail.c: Option "-q" to take options (discarded anyway) * hostenv/AIX4.1: Add "WAITPID=", and "ANSI_TYPES=" * hostenv/Linux: Added a warning about mmap() being broken in 1.2.x -kernels (well, probably broken...) * hostenv/OSF1v3.2-Alpha: Added "STDC_HEADERS=" -- for ANSI-C-headers * hostenv/SunOS4.1: Added "BZERO=" * include/listutils.h, libsh/listutils.c: newcell() is now a routine, and not a macro.. (In preparation of LISP-like guts rewrite..) * router/rfc822.c: If the returned address reference has no protection info (a NULL-pointed object), print out "-". * scheduler/scheduler.c: (and all others) Changed all "time(&now)" to be "mytime(&now)", and created a shared-memory server on which there is a redularly running server at the other side updateing system time into given variable, and the real scheduler() can read that data. Makes a lot less syscalls at the scheduler (At SunOS 4.1 the gettimeofday() syscall does take some 80 usec...) * scheduler/transport.c, scheduler/conf.c: Implemented a "forkrate-limit", which blocks the fork from happening, if over N fork()s have been done during the same second. The count does reset at the next second for N more childs.. Default for the N is 10. Create flush_child() routine to handle some of the awkward non-completed outputs from the scheduler to the childs. * scheduler/qprint.c: "Q-mode" at the reporting -- no reports in the traditional way. Also avoid a null-pointer deferrence. * scheduler/scheduler.c: Parse scheduler 'Q'-mode argument, handle gracefully the main scheduler inner loop exit. * transports/mailbox/mailbox.c: Move the mailbox-file time-stamp reset into after the write of the time. (Linux has peculiar read-time stamping due to unified-buffer-cache system.) * transports/libta/lockaddr.c: FCNTL-locking -- for systems starved by the IO-capacity.. (INCOMPLETE!) * transports/smtp/smtp.c Added "-P"-option to turn off the PIPELINING use. Do unusual, and overwrite also the environment variables, when updateing info-area at the memory image for the benefit of BSD-derived systems where such data update is possible for reporting program state outside. Detect "#idle", and report it outside. Working implementation of PIPELINING full-duplex mode for the SMTP-protocol. * utils/makendbm/makendbm.c: Support also the GDBM, if it is available, and no NDBM is. Fri Feb 9 02:32:01 1996 Matti Aarnio * Makefile: Releasing 2.99.26 * README.UPGRADING: Noted change in canned error message headers * smtpserver/rfc821scn.c: Give a bit more meaningfull report for case of "", which is "spurious dot at the end of the address".. Tue Feb 6 00:54:37 1996 Matti Aarnio * proto/forms/*, scheduler/msgerror.c, transports/errormail/errormail.c: Added third header tag: "ADR" which expects to find recipient address (to be placed into "to " envelope line) in there * smtpserver/smtpserver.c, smtpserver/rfc821scn.c: If the input to HELO/EHLO was syntactically incorrect, always trap it! Although don't always cause full abort.. Actually giving meaningfull report at each input, and storing "Bad.HELO.Input" into "rcvdfrom" entry.. * include/hostenv.h: Macro FILENO(x) was revised a bit to work properly on SunOS 4.1, which has the problem, need to revise it for other systems, if need be.. (IF the STDIO file descriptor is stored on a signed char, the default propagation of an fd over 127 makes it NEGATIVE, which causes some real trouble..) * lib/rfc822date.c: (Tom Samplonius) - gettimeofday() does not return timezone information on most BSD systems as "...time zone information does not belong in the kernel..". Also, this call is redundant, since timezone information is the already retrieved struct tm - I removed the "if (prettyname) sprintf(cp+5, " (%s)", ts->tm_zone);" statement because it is contained in the else part of a "if (prettyname) .." block and can never be executed. - I removed the NO_TM_ZONE stuff in the USE_BSDTIMEZONE section. This seems like a bad thing, however it seems to be required by AIX 3.2 though. I don't know whether AIX 3.2 is actually still used, or whether newer versions of Zmailer actually build on it. AIX 3.2 may be able to use the tzname[] global variable to get the name instead. * include/mailer.h, router/libdb/header.c, smtpserver/smtpserver.c, scheduler/msgerror.c, transports/errormail/errormail.c: Added new HeaderClass object: env-eof / end-end (two spellings) which is intended to be the last entry of the transport envelope headers. * include/ta.h, transports/libta/lockaddr.c, transports/libta/ctlopen.c, scheduler/msgerror.c: Routine lockaddr() uses now MMAP()ed memory on systems where it is available. Less syscalls on low-power machines.. * include/zmsignal.h: Ah well... truly portable signal processing appears to be rather elusive target.. Now Solaris, and some versions of BSD work again. * libc/__fopen.c, libc/siglist.c: Couple FreeBSD related patches from Tom Samplonius * libsh/builtins.c: Don't remember what system it was, but got a report of failing signal processing/wait()ing. Altered feature testing into hostenv/* -feature USE_UNIONWAIT .. * router/db.c, router/libdb/dbm.c, router/libdb/gdbm.c, router/libdb/header.c, router/libdb/core.c, router/libdb/ndbm.c, router/libdb/ordered.c, router/libdb/search.h, router/libdb/unordered.c, router/libdb/yp.c: "db -c database-file" --> outputs the number of elements in the database. Also incorporated a bunch of Nicholas Briggs patches (NIS -- USE_YP) for his environment. * router/functions.c: Rewrote (minimized) the rd_stability() related data amount used at sorting.. * scheduler/msgerror.c: Small changes to get it better online with IETF NOTARY-WG's report format.. * scheduler/update.c, scheduler/scheduler.c: Understands "#resync .." reqests, drops the info pertained to the jobspec, and returns to the problem sometime later (via normal job scan..) * transports/smtp/smtp.c, smtpserver/smtpserver.c: Implemented RFC XXXX of PIPELINING, which gives us radical speed-increase on case the sourcing system can do PIPELINING, and we have a need to transfer a large number of recipients in envelope. Tue Jan 23 11:49:58 1996 Matti Aarnio * transports/libta/lockaddr.c: Reduce the number of systemcalls by one -- write whole lock info in one go.. (instead of two) Mon Jan 22 09:34:35 1996 Matti Aarnio * scheduler/transport.c: Moved a piece of code to ``nonblocking.c'' * transports/libta/nonblocking.c: New file containing fcntl()s to turn file blocking on and off.. ( fd_nonblockingmode() / fd_blockingmode() ) * smtpserver/smtpserver.c: - Added EHLO response "PIPELINING" (server-side was easy..) - Made sure that if no RCPT TO:<> -addresses are given, will issue an "550"-error at the "DATA"-phase. * transports/libta/diagnostic.c: Minimize the "ok"-case report size, as filling pipe of the report-back channel does not help at all to keep the system running at full blast.. (average report size has been 110-120 chars, and the pipe-size is usually 4096.. It doesn't fit as many reports as it fits jobs.) * include/zmsignal.h: SIGACTION does not always have the luxury of SA_INTERRUPT defined at the system includes (odd at that.. but Solaris is odd anyway..) Fri Jan 19 19:42:16 1996 Matti Aarnio * Makefile: Release 2.99.25 * include/zmsignal.h: Changed the order of tested alternates, now: - BSDish SIGVEC stuff - POSIXy SIGACTION stuff - ... when all else fails.. (SVR3) (BSD and POSIX interchanged -- because OSF/1 3.2 didn't like SIGACTION stuff with SA_INTERRUPT, while Linux is happy -- and SunOS 4.1 needs SA_INTERRUPTs..) * router/functions.c: rd_doit() Made a mistake at pid-renaming scheme, which caused errorneous results to flood systems, and even to kill the scheduler, when filename length became obscenely large -- overflowling a stack-allocated buffer, and ... * scheduler/update.c, scheduler/threads.c: Use, and bookkeeping of feed_child() had a ``one off'' error in them causing occasional scheduler/ta deadlocks. (both waiting each other to tell more..) * scheduler/transport.c: feed_child() Feed the child only when the "proc->fed == 0", that is, as we set the flag, (and thread_start() clears it), we won't feed same message twice, and cause potentially excessive "#DBGdiag " loggings.. (Messages are not lost, it just disturbs postmaster's peace of mind..) * hostenv/SunOS4.1, hostenv/OSF1v3.2-Alpha: Turn back on the "MMAP=" settings, lack of them does hurt more, than they did themselves.. Thu Jan 18 11:25:50 1996 Matti Aarnio * include/zmsignal.h: SunOS 4.1 uses sigaction() in posixy style (sometimes at least), and thus the programs jammed in weird ways. Autch... Added SA_INTERRUPT into sa_flags.. * Releasing 2.99.24 to the public * libsh/interpret.c: findfreefd() It did `fstat()' in a loop, which is a bit heavy to my liking.. changed it to use fcntl(fd, F_GETFL), which at most platforms is quite light-weight indeed.. * scheduler/transport.c: stashprocess() Clear the procinfo structure when preparing to store process info to it. Had forgotten to zero one new variable in it, and caused mysterious halt gridlocks at scheduling.. * scheduler/transport.c: mux() Because the select-loop can take ages at slower machines, added queryipccheck() calls after each readfrom() call.. Now the responsivity to the management interface should improve.. * transports/mailbox/mailbox.c: When delivering to a file, errors at the storing (filesystem full, for example!) did ZERO-TRUNCATED the target file.. Oops.. (Existing file-size was looked up at lseek(fd,0,SEEK_SET) position, which happened to be at zero, of course..) * transports/libta/warning.c, transports/libta/lockfile.c, transports/libta/ctlopen.c: Changes at error reporting to get rid of "Misformed diagnostics" entries at the scheduler log.. * transports/smtp/smtp.c: Altered the log-file writing a bit, now each line gets prefix of format: ("%05d%c%05d",getpid(),randchar,seqnum++) and the log should be sort(1)able to see all sessions in happening order -- and sessions with reused pids should have differing randchar anyway --> no two SMTP-sessions should be intermixed within each other. * scheduler/scheduler.c: The new ``-S'' option causes startup to be all synchronous, and will not start scheduling before all of the queue is read in. * scheduler/qprint.c, scheduler/threads.c, scheduler/mailq.c: New queue report -- old one is suppressed per default. ----------- sample ----------- bash# mailq -sQQ 0 entries in router queue: idle 70 messages in transport queue: working smtp/*.de/0 Threads: 6 Procs: 19/6 Plim: 19 Flim: 150 Rcpts: 47 Idle: 13 smtp/*.net/0 Threads: 1 Procs: 1 Plim: 10 Flim: 150 Rcpts: 1 Idle: 0 hold/*/0 Threads: 1 Procs: 1 Plim: 20 Flim: 150 Rcpts: 2 Idle: 0 ... Kids: 64 Idle: 48 Msgs: 67 Thrds: 20 Rcpnts: 128 Uptime: 1m16s -------- end sample ---------- Tue Jan 16 13:04:08 1996 Matti Aarnio * transports/smtp/smtp.c: getmxrr() Wrong analysis of MXes causing occasional core, and deliveries to wrong machines.. * router/rfc822.c: makeLetter() After a ``getline()'' do see if the input has a terminating CRLF character pair. If it has, convert it to LF-only.. This makes system to process peacefully messages which have CRLF at their line change, which certainly is not UNIX style.. * scheduler/msgerror.c: Minor case improvement into ``delivery expired without diagnostics having been returned to the scheduler'' (caused most likely by an overfed transport-agent..) * scheduler/*: Still more performance related tuning, and bug hunting at vger.rutgers.edu -- now that small machine appears to be able to push some 200 000 messages in a day (1..n recipients of same message at same remote host count as one message), and linux-related lists are flowing out at furious rate.. :-) Profiling does tell weird things :-) Most of the time was spent at thread_linkin(), which took 2.3 ms/call, but was called some 200 000 times.. Small change there speeded it up to 0.3 ms/call, and it made cistrcmp() to top.. * router/functions.c: Altered multi-router locking mechanism a bit. Now it will use format: inode-pid which should be unique within one machine. Thu Jan 11 13:52:06 1996 Matti Aarnio * Makefile: Release 2.99.24 -- into testing at vger at first.. * hostenv/SunOS4.1, hostenv/OSF1v3.2: Commented OFF the "MMAP" feature -- the systems can do it just fine, but we need to see, if systems perform better, when not using it... (low-memory machine problems) * transports/hold/hold.c, transports/errormail/errormail.c, transports/smtp/smtp.c, transports/mailbox/mailbox.c, transports/sm/sm.c: Accept input of queued up multiple job descriptors, which MIGHT meet an fatal input error so that the last entry does not end with a '\n'. Reject such an entry silently. * transports/smtp/smtp.c: When getting a bunch of job-descriptors with host selector on the input, and if the last reported diagnostic was a "retryat" failure, do silently drop them, until the given retry-time is reached. * transports/hold/hold.c: Understand hold-definition: "ns:host.name.there/any" right.. * scheduler/readconfig.c, scheduler/scheduler.c, scheduler/mailq.c, scheduler/qprint.c, scheduler/scheduler.h, scheduler/threads.c, scheduler/transport.c, scheduler/update.c Radical changes at the scheduling algorithms.. - "ageorder" config flag: makes vertices within the thread to be kept, and processed in the order of spool-file ctime. - "overfeed=NN" config flag: tells how many entries are fed to the transport client in one go so that it may "eat from the pipe" for the job definitions. * scheduler/msgerror.c: Make it to tolerate old (pre 2.99.22) errors, and thus allow smooth upgrade from old system to the new one. * lib/prversion.c: Year has changed, updateing copyright statements a bit. * include/hostenv.h, lib/loginit.c, libc/mail.c, libsh/interpret.c, libsh/io.c, libsh/trap.c, router/functions.c, router/libdb/ordered.c, router/libdb/unordered.c, router/rfc822.c, router/rfc822hdrs.c, router/shliaise.c, scheduler/mailq.c, smtpserver/smtpserver.c, transports/mailbox/mailbox.c, transports/smtp/smtp.c: On some machines exceeding 128 FDs on the scheduler (or elsewere) causes some rather fatal things to occur, namely the variable storing the fd in the STDIO descriptor is signed char, while it should be at least ``unsigned char'', or preferrably ``short'' ( SunOS 4.x .. ) Wed Jan 3 10:24:59 1996 Matti Aarnio * proto/cf/rrouter.cf, proto/cf/canon.cf: Got rid of ``UUCPACTION'' and ``BITNETACTION'', let them be resolved via $MAILSHARE/db/routes: .bitnet smtp!some.old.relict .uucp smtp!uunet.uu.net (And if you REALLY have BITNET routing database, it can handle just fine..) Also commented away the IN% -matching, which after all was my local hack to handle VMS users coming to UNIX.. .. but as we are killing the VMS, there is no need for that hack anymore .. * proto/cf/server.cf: The quadprint() -routine needed a bit of an addition to handle non-found user->fullname mappings -- using same method as the mailbox uses when delivering: If given "UsEr" does not match, lowercase it, and try again. (If lowercased version doesn't match, yield an error..) * scheduler/update.c: Hunting up and down for accesses to invalidated memory blocks, found one at the end of the u_retryat().. * scheduler/msgerror.c: Plugged memory leakage (strdup()ed string) Sat Dec 30 02:33:23 EET 1995 Matti Aarnio * Makefile: Releasing 2.99.23 * router/router.c: Added "nosyslog" variable, and "-S" option, which together disable (per default) router from syslog()ing everything to syslog in addition to some logfile used per default... * transports/smtp/smtp.c: Rearranged diagnostics reporting, now it will correctly report the status of the connection problems in addition to the interactive transaction problems. * scheduler/threads.c, scheduler/transport.c: When feeding a task, clear "vtx->ce.pending" -flag, thus having the queue-reports not to show spurious "thread-/channelwaits"... * transports/libta/mime2headers.c, transports/libta/mimeheaders.c Split it into two, and took friend's MIME-2 encoder into it... not yet truly in use. * routers/, and scheduler/ Several fixes to get it to compile again at OSF/1 (well, I didn't compile EVERYTHING before releasing 2.99.22 ..) * Makefile: Releasing 2.99.22 * Large changes to scheduler -- trying to make it a bit smarter at the retry (whole thread retries) * Some IETF-NOTARY-changes to the scheduler, and transporters * router, scheduler, and smtpserver will now abort, if they fail to write any content to their respective ``pidfile''. Fri Dec 15 15:54:44 1995 Matti Aarnio * transports/mailbox/mailbox.c: Add a "setpwent()" just before looking up named user, at least on SunOS4.1 (vis NIS) this apparently makes difference ?? Sat Dec 2 03:05:36 1995 Matti Aarnio * transports/smtp/smtp.c: Fixed the treatment of maxpref picking at the MX handling. (Darren Kinley ) * router/shliases.c, proto/cf/aliases.cf, proto/cf/standard.cf: Looked into ways of making the duplicate removal once again setting a global XXXX Sat Nov 25 18:39:56 1995 Matti Aarnio * lib/rfc822scan.c, router/rfc822.ssl, router/rfc822walk.c: Teached the syntaxes a new special token: "::" like exists in the DECNET addresses.. Thu Nov 23 16:34:50 1995 Matti Aarnio * scheduler/msgerror.c, scheduler/update.c, scheduler/scheduler.c, scheduler/scheduler.h, include/ta.h include/mail.h, all TAs, transports/libta/diagnostic.c: Altered error messages (and a couple other things) to match more evenly with IETF NOTARY-WG's error report format, including status codes, etc. If old errors are fed to the scheduler, it WILL crash! See README.UPGRADING about 2.99.22 ... Fri Nov 17 13:22:18 1995 Matti Aarnio * router/rfc822.c: Changed rules on when to add "To:"/"Resent-To:" headers at all... If any "To:" or "Resent-To:" header exists, DO NOT add "(Resent-)To:" -header. Makes list expansion a much more forgiving to humans.. * transports/smtp/smtp.c, transports/libta/dnsgetrr.c, lib/hostent.c: - Randomize the order of MXes with same preferrence - Randomize the order of A-records before connecting to anywhere (multiple addresses on same host) The goal is to distribute the load to all servers of the target domain, when it lists multiple servers.. Fri Nov 10 12:54:30 1995 Matti Aarnio * Makefile: Release 2.99.21 * Updated: README.UPGRADE, INSTALL, Overview, README, README.solaris, Config.osf1-funet, Config.solaris * Switched: scheduler -> scheduler-old, scheduler-new -> scheduler * hostenv/FreeBSD-2.0: Some minor tweaks by Tom Samplonius, we shall see how they blend in.. * include/hostenv.h: Unconditionally include , to be used elsewere to detect some system dependent things. * Config, Config.osf1-funet: Using best C-compiler on AXP platform with hardest optimization switches... It makes compiler apparently to halt sometimes, but after 3-4 CPU MINUTES, those larger modules are compiled -- the result is truly magnificent :-) (That compiler does need substantial amounts of memory to run, at least 200M to "ulimit -m" -- max memory size!) A bit later: Decreased optimization demands by small amounts -> smaller code (less inlining), and faster compile.. A LOT faster.. * lib/rfc822scan.c: Tolerate a case where "\" ends the line on which we are scanning tokens. (PINE originates such sometimes?) * scheduler-new/threads.c: Last (?) of the thread bugs fixed -- this one occurred rarely at UTU.FI SPARC server. Some sort of timing thing, which turned out to happen very infrequently.. Mon Nov 6 12:09:28 1995 Matti Aarnio * smtpserver/smtpserver.c: - Wrong code on "HELP" responses: 241, changed to 214 - EXPN and VRFY changed to demand HELO/EHLO before being usable * hostenv/SunOS5.*: Added the forgotten "SVR4MNTENT=" -entry. * scheduler-new/ Still more hacking on all fronts (eh, source modules) * transports/libta/ctlopen.c: Reset the ctlsticky's internal state in the begin of its use.. * transports/mailbox/mailbox.c: Do something with the SIGALRM -- don't just SIG_DFLT it (which means program aborting itself), rather use it to break infinite wait on where you were (NFS lockf() ) Also reset the [re]uid to zero (root) before opening a control file -- things work better that way.. Moved BIFF into the delivery loop, we shall see... * transports/smtp/smtp.c: Still leaked SMTP connection file-descriptors, last one plugged now ? (on target host change..) Mon Oct 30 11:52:34 1995 Matti Aarnio * transports/libta/mimeheaders.c: Small change to MIME-2 processing (it is imperfect, actually "broken", but lets not be too purists..) Now a quoted token will not contain enclosing parents, that is, "QUOTABLE)" -> "=?xxxx?Q?QUOTABLE?=)", and not like it was: "=?xxxx?Q?QUOTABLE)?=", which breaks the headers really badly.. * scheduler-new/transports.c: Put pipes_shutdown_child(tofd) into suitable places instead of close(tofd), because socketpair() "pipe" does break at normal close()... Tue Oct 24 22:30:50 1995 Matti Aarnio * Released 2.99.20 * include/ta.h, transports/*/*.c Changed the way the diagnostic() call works, now the SMTP transporter can send a "retryat +NNN " request. It is used for rescheduling entire thread for a bit later time -- now fixed at 60 seconds. * scheduler-new/ Revised somewhat the internal logics, introduced active use of "retryat" comminication primitive from the transport agents to the scheduler. Fixed (?) the timeout problem reported from Hongkong, also fixed another "PROC NOT IN IDLE CHAIN", which was caused by childs dying prematurely for some reason. Wed Oct 18 09:10:07 1995 Matti Aarnio * Released 2.99.19 * hostenv/*, scheduler-new/pipes.c, scheduler-new/transport.c, scheduler-new/threads.c, scheduler-new/Makefile.in: If the system can do a bi-directional pipe, at least a socketpair(), we need only ONE fd per child. Enable by hostenv define SOCKETPAIR= (TODO: SysV has STREAMS-pipes, add support for them into the scheduler-new/pipes.c, then all three existing methods are covered... ) ( scheduler-new/pipes.c is a new module.. ) * scheduler-new/threads.c, scheduler-new/transports.c: Finally found (?) the reason, why running childs are lost, and mis-calculated occasionally. (It turned out to occur most frequent with error-channel, but had nothing to do with any channels per se..) Anyway, sometimes a child is killed, and the recovery was mis-treated, but the reporting-fd still had data! Thus every now and then, while a process was gone, but fd (and thus a pid-slot) was still in use, and the process was thought to be valid for "feeding a child", for example.. * transports/smtp/smtp.c: When doing close() due to change of host, or a final close, don't just close(), do "QUIT" + close()! Fri Oct 13 17:21:27 1995 Matti Aarnio * hostenv/IRIX53, router/prototypes.h, scheduler*/mailq.c: IRIX-5.3 diffs from Darryl Miles * scheduler-new/ - Still more hammering at it with DEC ATOM-tools Third Degree :) Now only memory leakages around are in the system libc! (Brr...) [ well, new leaks introduced later, but that is another story.. ] Thu Oct 12 20:23:26 1995 Matti Aarnio * smtpserver/smtpserver.c: Fixed an error in SMTP EXPN, and VRFY. * scheduler-new/ - fixed expiry processing (when multiple recipients exist for the message) - fixed process counters -- idle_cleanup() leaked pids.. Mon Oct 9 11:07:25 1995 Matti Aarnio * scheduler{,-new}/transport.c: Improved (?) support for IBM AIX systems (select() system call.) * include/hostenv.h, hostenv/*: STDLIB= -> USE_STDLIB -> include in hostenv.h UNISTD= -> USE_UNISTD -> include in hostenv.h * router/rfc822.c, router/rfc822hdr.c: Hide away (effectively removed) "resent"-processing, so that if there are headers in there, no new "Resent-" headers are added just because some "Resent-Message-Id:" -exists, but no "Resent-To:"... Wed Oct 4 15:15:03 1995 Matti Aarnio * Released 2.99.18 * transports/smtp/smtp.c: Two fixes to a case where system doesn't have working mmap() One change so that report() routine tells more meaningfull info. * router/router.c: Rid the USE_BSDGETPGRP, and use USE_BSDSETPGRP instead. It is too much hassle to change all hostenv files to handle this othervice.. * scheduler-new/: Hacking continues at all fronts, still there is some discrepancy in between data-structure counters, and values counted, when chains are traversed! Now the scheduler has an option to create performace monitor log: -l logfile.name The file looks like this: 812819196 167223-3 0 12 ok smtp/ugcs.caltech.edu 812819197 167227-1 0 11 ok smtp/udcf.gla.ac.uk 812819212 167231-3 0 7 ok smtp/sci.fi Columns are: - spool-file creation time (ctime) - spool-id (filename) - time difference from creation of spool file to route file creation (secs) - time difference from creation of route file to delivery of the message (secs) - what happened (ok/error/expiry) - recipient channel/host * Makefile: Better integration into the whole, I hope. Someday soon the "scheduler-new" will be changed to be "scheduler", and current "scheduler" becomes "scheduler-old" ... Sun Oct 1 15:57:39 1995 Matti Aarnio (mea@oh1mqk) * Dumped 2.99.17 -- and a bit later revised scheduler-new.. * libsh/Makefile{.in,}: (Michael Thompson ) Fix to an awk script embedded into the Makefile. * scheduler-new/ It works ! Old configuration information is out of date. There are still open issues, see associated commentary files.. Fri Sep 29 09:22:50 1995 Matti Aarnio * transports/smtp/smtp.c: Odd, SMTP response "552 permanent resource limitation" did trigger ONLY an EX_TEMPFAIL, which caused retry.. Does now EX_UNAVAIL, which causes snappy reject. Wed Sep 27 22:00:28 1995 Matti Aarnio * Makefile: Changed subdirectory makefile re-generation method slightly. Now if the production fails, it won't (I hope) scramble the original file... * compat/rmail/Makefile, compat/sendmail/Makefile: Install "rmail", and "sendmail" into $MAILBIN ! Let sysadm to add symlinks from where-ever the original locations are to the $MAILBIN/ .. * proto/zmailer.sh: The script will check that $MAILBOX points to a directory so that the ZMailer can't be started on a mis-configured system... (.. at least of that part.) * transports/libta/mimeheaders.c: Recognize "KOI8*" -fonts as alias to US-ASCII, when content is 7-bit only.. Recognize also when HEADERS need MIME-2, that is, coding 8-bit chars in headers (subject et.al.) into MIME-2 format.. * transports/libta/writeheaders.c: If converted headers exist, write them ALWAYS! * transports/mailbox/mailbox.c, transports/sm/sm.c, transports/smtp/smtp.c: Use strdup() instead of strsave(). Makes difference, when it occurs on a transporter waiting for next work.. * transports/smtp/Makefile*: For "install"-target, make sure also "mprobe" exists. * transports/smtp/smtp.c: If headers have 8-bit chars, we may need to convert them to MIME-2 (RFC-xxxx) "coded words" If EHLO+HELO fails, open same machine again, and do HELO only, if THAT too fails, try next MX... Report (into log-file) about connection making, and possible errors therein.. * scheduler/ Lots of small changes - prototypes for functions. * scheduler/mailq.c, scheduler/qprint.c, scheduler-new/mailq.c, scheduler-new/qprint.c, scheduler/scheduler.h, scheduler-new/scheduler.h: Noticed that qprint on Linux didn't like at all the reported vertices from AXP machine, where vertex storage address was well ABOVE 4 G (max value for unsigned 32-bit long at i486..) Now the vertex structure has a field where qprint() can store incrementing SMALL integer unique one for each vertex. The vertex identifiers start now from 1 each time the mailq is run... * scheduler-new/ The face-lift project continues, it CAN do most of the things already, but it is somewhat sloppy at process scheduling, and tracking. Tue Sep 19 11:00:52 1995 Matti Aarnio * transports/smtp/smtp.c: A set of options that change their values during the runtime are reset before smtpconn(). If "EHLO"+"HELO"-pair crashes (like it does with some SMTP-servers), open a new connection, and do "HELO" only. Mon Sep 18 10:42:37 1995 Matti Aarnio * Makefile: Dumped version 2.99.16 after public demand.. * smtpserver/smtpserver.c: Extended the trappings of errno's from accept(). On Linux we have seen (among others) ECONNRESET, which earlier did cause swift exit(1), because only EINTR was tolerated... ( .. and incidentally the EINTR didn't happen at all, because the signals are of syscall-restarting nature..) * transports/smtp/smtp.c: When writing to a log-file, write also info regarding connection attempts so we can look when remote systems are unreachable... Fix the '-8' -flag when the message content isn't MIME, but we want to be 8-bit transparent.. Sat Sep 9 01:13:20 1995 Matti Aarnio * scheduler-new/ : The scheduler will have complete ``face-lift'', before this is over... Now made separate work copy, and left 2.99.15 version with old name.. Sat Sep 9 01:04:55 1995 Eugene Crosser * transports/mailbox/mailbox.c: More generic (?) approach of UID/GID handling, we hope. * router/libdb/search.h: Changed one "void *" to "char *", because Sun Solaris C-compiler (SparcWorks ?) didn't like to do pointer arithmetics with void pointer.. Fri Sep 1 11:37:09 1995 Matti Aarnio * transports/smtp/smtp.c: Thomas Knott spotted (and fixed) a bug on email processing when sending 8-bit mail over an ESMTP connection, and the system does not support MMAP service.. * include/mailer.h: Added macroes: STREQ() and STREQN() (case SENSITIVE string equal) * router/rfc822.c: The ``FindHeader()''-macro uses now case SENSITIVE lookup of the header lines. Thus "From "-lines will not match those of "from ".. (Nothing breaks, I hope..) * libsh/io.c: PUTC(), vsiodoprnt() Two subtle one-off errors which DEC OSF/1 ATOM-tool "Third Degree" detected. These caused actually buffer overflows :-( (by one byte), which in turn caused far later errors, which looked rather weird.. * scheduler/ - all files :) Made prototype functions for all inter-twined routines used in the scheduler, cleaned out several unused ones. (Not yet doing things quite like I want, but better..) * hostenv/FreeBSD-2.0: Contributed by Tom Samplonious. Fri Aug 25 13:24:30 1995 Matti Aarnio * Release of code version 2.99.15 A big pile of changes all around, summary below: * Incorporated (finally!) Lehigh's AIX-port into the main source. I don't know how clean it is, but... * hostenv/Linux, hostenv/OSF1v3.0-Alpha, hostenv/OSF1v3.2-Alpha: Added "MMAP=", and "STRERRNO=" for Alphas, "STRERRNO=" only for Linux (on which the mmap() is a bit troublesome..) Added also "RESOLV_LIB=-lresolv" to Alphas. * doc/zmdirs.{fig,ps}, doc/zmsched1.{fig,ps}: Couple pictures telling about directory, and file relations within the ZMailer. * include/sysprotos.h: DEBUG_FOPEN -things for debugging FILE * leakage.. * libauth/authuser.c: Fixed variable declarations. * libc/__fopen.c: Compile the library always, just the USE of it is left to user applications compilation. * libc/strerror.c: Verify that error number is in valid range -- detected a case of errors where the returned errno was 8800+, and got most peculiar error results... * libsh/execute.c: Small changes to ensure compilability with multiple compilers; apparently somehow the optimizers (gcc-2.7.0, and DEC's "cc -migrate" both on Alpha) generated skewed code, which produced wrong result. * libsh/tregexp.c: The "trace regexp", and "trace matched" (I think) regexp TRACE routines produced parts of their output to wrong place (to stdout via putchar() -routine) thus causing trouble when trying to trace large-scale processing. Now all output from this file goes to some "char *" -buffers, or to stderr. * proto/cf/aliases.cf: Had one split line, and other minor changes. * proto/cf/aliases3.cf: Yet another slightly edited version of the default "aliases.cf" (untested!) * proto/zmailer.sh: The prototype of the "zmailer" -command got rid of vestiges related to "$POSTOFFICE/scheduler/" -directory. * router/functions.c run_listexpand(): New option "-p", tried to use "router()" -entrypoint, but apparently it wasn't very successfull. Hmm.. * router/libdb/bind.c: Problems related to 64 bit machine, where "u_long" is rather much larger, than expected "u_int"'s size... * router/rfc822.c: Changed directory scheme to match that of the scheduler by getting rid of "$POSTOFFICE/scheduler/" -directory * scheduler/agenda.c: Fixes on time-stamp processing, adding up-to-date code into time handling -- setting all variables that are later read from. * scheduler/mailq.c: Got rid of "$POSTOFFICE/scheduler/", and to some extend replaced it with "$POSTOFFICE/transport/". * scheduler/msgerror.c: Use new directory/file-paths. * scheduler/scheduler.c: Use new directory/file-paths Read files from the directory into time-stampted queue (with file's st_ctime), and accept into it only those that are not already in processing. Parse _CF_SENDER -entries from the scheduled job, and use that address (well, last one of them) as an address for sending error replies to. * scheduler/scheduler.h: Replaced "int rptwait" with "int hungry", though neither are in use... * scheduler/transport.c: Use new directory/file-paths. Prepare for using bi-directional communication in between scheduler, and transporters -- so that scheduler starts a transporter, and gives it synchronously one job at the time, and waits for "#hungry\n" -message before feeding another... * scheduler/update.c: Use new directories/file-paths. Support "#hungry\n" -messages * smtpserver/rfc821scn.c: Accept domain name components. Increase the listen() queue size. * smtpserver/smtpserver.c: Voice complaint, when user tries to feed us: MAIL FROM: someone@somewhere that is, when the input is against the RFC-821 syntax.. * transports/errormail/errormail.c, transports/hold/hold.c, transports/mailbox/mailbox.c, transports/sm/sm.c, transports/smtp/smtp.c Support the new "#hungry\n" -style "feed me" messages. Fixed a couple "if (foo = faa)" vs. "if (foo == faa)" errors in the source. Tue Jun 6 23:35:34 1995 Matti Aarnio * transports/sm/sm.c, transports/smtp/smtp.c, transports/hold/hold.c, transports/mailbox/mailbox.c, transports/errormail/errormail.c: Amended acceptable command-entry format to be: spool/file/name [ \t host.data ] \n This in preparation to alter scheduler to keep TWO pipes open for each running transporter, and to feed them ONE JOB at the time. To achieve syncronization the report-back format is amended with a blank line after the last response from entries in a file. (That is, an extra "\n".) Sat Apr 15 02:05:18 1995 Matti Aarnio * transports/sm/sm.c: Fixes on MIME QP-decode conversion mode. Now QP-decode works right. (Somewhat earlier..) Added also couple lseek()s to that the message is now processed with correct headers.. Sun Mar 26 20:56:22 1995 Matti Aarnio * (almoast all files..) "A bit" more of 64-bit stuff. A change on splay-tree symbol format caused severe repercussions to occur, namely a widely-held assumtion that "sizeof(void*) == sizeof(int)" caused serious trouble :-( Thu Mar 23 00:56:05 1995 Matti Aarnio * transports/libta/ctlopen.c ctladdr(): A bit more generic way to scan for control addresses... a way which does not miss the beat when there is 8-bit chars on tokens... * router/functions.c: Re-arranged the usage/setting of errors_to -variable. Now it should not cause spurious errors on free()-time! * transports/mailbox/mailbox.c putmail(): Parametrize putmail() with "w" (for pipes) and "a+" (for files). OSF/1 3.2 had problems when pipe was fdopen()ed with "a+"... :-/ Instead of using getservbyname("biff","udp"), do hard-wire the BIFF-port to the client -- sometimes (SunOS 4.1.3 w/ NIS) that library call is extremely SLOW to return... A SCO-user reported, that MMDF-format is: C-A C-A C-A C-A message1 C-A C-A C-A C-A C-A C-A C-A C-A message2 C-A C-A C-A C-A ... Instead of (like it was coded here): C-A C-A C-A C-A message1 C-A C-A C-A C-A message2 C-A C-A C-A C-A ... Oops about that... Thu Feb 9 20:28:22 1995 Matti E. Aarnio (mea@oj287.astro.utu.fi) * Dump 2.99.13 * router/rfc822.c makeLetter(): Process all headers, don't just SQUIRREL() away when there is mal-formed envelope information. This way the messages will not just disappear into $POSTOFFICE/postman -directory. One particular bogon format is perhaps detectable.. ">From ..." as the FIRST line. Such case is possible, when a faultly configured sendmail/smail feeds our system via SMTP, but is configured to feed UUCP.. * scheduler/transport.c: Changed all references to number of open files in the system to use common reference platform from library call resources_query_nofiles(). This mattered especially when the cpids[]-array was allocated by using resources_query_nofiles(), and other references to it were limited by FD_SETSIZE (for example), but THAT happened to be larger (like 2 or 4 times) than the allocation size... Wed Feb 8 00:51:59 1995 Matti E. Aarnio (mea@oj287.astro.utu.fi) * transports/smtp/smtp.c: Yet one more loose thread on the SMTP connection closure survivability. (Detected by an SIGSEGV on NULL ptr..) * transports/mailbox/mailbox.c main(): New option: "-h 'hostname'" Assists the success with scheduler configuration containing "byhost", when scheduling this piece of email. (Such scheduling may be used to run "Zillions" of parallel mailboxes -- "maxchan=20, maxhost=1, byhost", and thus have each email recipient processed individually. This REQUIRES that the router stores something personal to each recipients "host"-field, not just "-"...) When this option is not used, it defaults to the original behaviour, where "mailbox" delivers ALL "local/*" -channel emails of the file during one invocation. * transports/libta/ctlopen.c ctlopen(): Revised a bit the "Cannot open control file..." -message by removing a potential `printf("%s",NULL);' -case.. * libc/myhostname.c: An extra indirection causing getmyhostname() to yield wrong data in place of FQDN hostname... Hmm.. Wrestling it to be more commonly available, (various platforms) was a bit difficult -- even inbetween SunOS and Solaris-2.. * transports/smtp/smtp.c: Rewamped the connection tear-down code. Now it SHOULD happily use same connection for multiple emails. Tue Feb 7 00:47:26 1995 Matti E. Aarnio (mea@oj287.astro.utu.fi) * Dump 2.99.12 * scheduler/scheduler.h, scheduler/scheduler.c, scheduler/agenda.c, scheduler/transport.c: Actual long-standing (ever since 1.0 ?) problem was the usage of wrong filepath in the transport.c:tscan(), but without some external help, the right one was not available.. (Scrambled especially the "-s" option!) For fixing it was necessary to realize, that 1) Each file is scheduled only once for each thread 2) thus can be created a pointer into CTLFILE-block which points to the CURRENT thread-vertex (a companion for "mark" used for tscan() ) 3) the current vertex pointer can now contain an entry to the path where the file is linked to With these pointers in places, transport.c:tscan() and after it, the client scheduling, can finally be done reliably. * scheduler/scheduler.c, scheduler/transport.c: Once more, re-wamped the SIGCHLD processing, as 2.99.11 generated scheduler zombies on SunOS 4.1.x Mon Feb 6 03:33:03 1995 Matti E. Aarnio (mea@oj287.astro.utu.fi) * Dump 2.99.11 * scheduler/transport.c: The waitandclose() is called when the file-descriptor is closed, or the read causes an error. (EINTR is no error..) * libc/getopt.c: Wrong system-wide definition file -- must be "hostenv.h" instead of "libsupport.h".. Sun Feb 5 03:12:02 1995 Matti E. Aarnio (mea@oj287.astro.utu.fi) * scheduler/resources.c: sysconf()-calls appear to be problematic -- on Solaris it yields wrong data :-/ Wow, must check if scheduler uses stdio anywhere, if not, it can run (On Solaris) 1024 fds.. * scheduler/{transport.c,scheduler.c,scheduler.h}: Defined "USE_SIGREAPER" in the scheduler.h, and when it is defined (no reason to take it off ?) uses wait() immediately on the childs, and if it matches to actual transporter child, marks it as reaped on cpids[] -array. End of "Scheduler Zombies" ? Sat Feb 4 17:34:14 1995 Matti E. Aarnio (mea@oj287.astro.utu.fi) * hostenv/00README, include/libz.h, include/mailer.h, lib/pwdgrp.c: Inverted the test, and changed the variable name.. Now if we ever DO need to use our own version of getpw{nam,uid}() and getgrnam(), we can define ZGETPWNAM on the host environment. I doubt we ever do. (This change removed one COMMON problem on porting to other platforms..) * transports/smtp/smtp.c: Revised the timeouts. Now output will abort if it can't write at the rate of 1kB/5 minutes, rearm after every kilobyte. If commands don't respond within 5 minutes (earlier was 10 minutes) it also timeouts -- except with "." to "250 OK", which is 10 mins. ( RFC 1123 ) * smtpserver/smtpserver.c: Timeout parameters, now it dies out pretty fast if the remote end does not feed data in fast enough... 20 minutes for in between SMTP commands 10 minutes for a kilobyte of SMTP DATA. * smtpserver/smtpserver.c mgets(): Something funny with response buffer realloc(); It did scramble the pointer to the chars in the buffer (GCC 2.5.8 on SPARC), and eventually caused a SIGSEGV.. Amazingly small change fixed it :-/ * A LOT changes right and left, NATIVE port to OSF/1v3.0 on Alpha. Files affected: Config.osf1, hostenv/OSF1v3.0-Alpha, lib/rfc822date.c, compat/sendmail/sendmail.c, libc/getopt.c, include/authuser.h, nclude/limits.h, include/mailer.h, libauth/authuser.c, libauth/authuser.h, libmalloc/assert.h, livmalloc/externs.h, libsh/execute.c, libsh/interpret.c, libsh/sslwalter.c, libsh/trap.c, router/libdb/bind.c, router/db.c, router/functions.c, router/prototypes.h, router/rfc822.c, router/rfc822hdrs.c, router/router.c, scheduler/mailq.c, scheduler/msgerror.c, scheduler/update.c, scheduler/scheduler.c, scheduler/transport.c, smtpserver/smtpserver.c, transports/mailbox/mailbox.c, transports/smtp/smtp.c * (no file): Noticed that Rayan Zachariasen no longer has an account on U of Toronto. Thu Jan 19 02:34:02 1995 Matti E. Aarnio (mea@oj287.astro.utu.fi) * Makefile: Patchlevel=10, DUMP * transports/smtp/smtp.c main(): Fixup of parameters of the getopt() call. Wed Jan 18 13:29:50 1995 Matti E. Aarnio (mea@oj287.astro.utu.fi) * router/functions.c run_listexpand(): When "rrouter" evaluation meets a duplicate removal, it returns a list of "()" -- so, an object without contents. (Again SIGSEGV..) * libc/mail.c mail_close(): When ZENV-variable ROUTERDIRS does have invalid elements, use the last valid element (or the default of "router") as the directory name. ( "sendmail", and "smtpserver" are affected by this! ) ( Sure, nobody enters invalid elts into the dirlist ? Heh.. ) * transports/smtp/smtp.c writemimeline(): Did wrong thing, when outputing QP coding, had a one-off with "column" variable, when it was a question about line prefixing "."... (Auch!) Tue Jan 17 10:53:02 1995 Matti E. Aarnio (mea@oj287.astro.utu.fi) * hostenv/Linux, hostenv/SunOS5.3: "GETPWNAM=" -- MUST BE DEFINED, as the otherwise default of iterating with getpwent() is BAD MEDICINE! This has been the reason for several odd problems on POSIX, and alike, systems! ( "include/mailer.h" has "#define getpwname zgetpwnam" to map the normal library call to ZMailer wrapper.. ) * scheduler/scheduler.c: A missing "int wrkcnt;" variable declaration.. (How on earth it did slip into the dump ?) Fri Jan 13 13:59:07 1995 Matti E. Aarnio (mea@oj287.astro.utu.fi) * Makefile: PATCHLEVEL=9, dump for public tests. Thu Jan 12 12:31:02 1995 Matti E. Aarnio (mea@oj287.astro.utu.fi) * router/functions.c: When the $ROUTERDIRS is defined, but not all dirs in it do exist, skip the nonexistent ones. (It was FAST core-drop time at nic.funet.fi ..) * Makefile: When cleaning, DO NOT delete libmalloc/version.c ! That file is separate from all other version.c's ! * router/router.c main(): Altered XMEM-compile-option related usage of "fileno()" macro. Actually got rid of it... Wed Jan 11 14:22:14 1995 Matti E. Aarnio (mea@oj287.astro.utu.fi) * router/rfc822.c run_rfc822(): Open the mail-file as O_RDONLY ( fopen(file,"r") ) instead of "r+" (O_RDWR), thus under no circumstances allow our internal bugs to cause leakage of logs (!) into the mail-body ! (Original code claimed the "r+" is needed for locks to succeed, however link()/rename() -"lock" works quite well without...) * scheduler/scheduler.c, scheduler/transport.c, man/scheduler.8: A runtime option "-N ##" to set how many file-handles are to be left open when a client is started. On most systems that number is low -- under 10 or so, however on Solaris it must be far higher.. Default value (32) works for me (on Solaris), however I have had comments that it isn't enough (on Solaris) :-/ * router/functions.c run_listexpand(): Changed somewhat the address parser -- finally (?) got it right! * router/rfc822hdrs.c pureAddressBuf(): (NEW) Outputs pure address into a buffer for usage as router argument. * router/libdb/ordered.c, router/libdb/unordered.c, router/db.c: - Moved readchunk() to the end of the .../unordered.c from router/db.c - Teached "db" to flag a case of using "Indirect" post- processor for things other than ordered/unordered relations. * libmalloc/, (libmalloc-o/) Mark Moraes' malloc-1.17. The previous one is in libmalloc-o, until deleted.. * transports/smtp/smtp.c, transports/sm/sm.c: When translating to QP, always translate the last SPACE/TAB on the line. * transports/libta/mimeheaders.c check_conv_prohibit(): Added another KLUDGE meaning to the "Content-conversion:" header... "forced-qp", which forces QP output. (Applied only into SMTP so far..) Tue Jan 10 13:46:33 1995 Matti E. Aarnio (mea@oj287.astro.utu.fi) * smtpserver/smtpserver.c: Analyze headers a bit more. Now complain immediately, if subjected to 8-bit characters in any header aside of "Subject:". (Because for example PP will choke on us, when we forward such a message and there is no MIME-HDRS encoding on them..) We can't (easily/portably) do the MIME-HDRS encoding for the user, because we don't know his/her character-set. (We can assume, of course, but it would not be universally usable.. Hmm.. UNKNOWN-8BIT ?) * transports/sm/sm.c: New option: -Q to create "narrow-QP-encoding", that is, to encode TABs and SPACEs in QP. Without this the QP-encoding is created leaving them untouched.. Mon Jan 9 16:16:36 1995 Matti E. Aarnio (mea@oj287.astro.utu.fi) * transporter/mailbox/mailbox.c: If the host does not have /var/spool/rwho/ -directory, DO NOT complain about it (with exit()!) Failing to RBIFF is NO fault! * scheduler/transport.c: Before bind()ing querysocket, do set SO_REUSEADDR, so it won't complain just because of old lingering connection trails... Sun Jan 8 22:26:35 1995 Matti E. Aarnio (mea@oj287.astro.utu.fi) * router/functions.c, router/rfc822.c: Restore my old (and cumbersome) "errors_to" trick on list expansion. Even though it causes the error address for ALL messages cascading from one source to expand into the LAST expansion's owner, it is better, than no trap at all.. (the rfc822.c:sequencer() needs a serious rewrite with those attributes..) Fri Jan 6 01:59:09 1995 Matti E. Aarnio (mea@oj287.astro.utu.fi) * Makefile: Patchlevel: 8 * router/functions.c: rd_{in,}stability(), run_daemon(): Revised a bit when it became evident, that the multi-directory picking did not work anymore. * smtpserver/smtpserver.c: Process the incoming message for headers. Set "mail_priority" according to what you find on the (possibly) existing "Precedence:" -header: - (nothing) 0 "bulk" 1 "junk" 9 * doc/draft-*: Pulled in new IETF SMTP related drafts (and some other) * README.UPGRADING: Noted a change on 2.98 when the proto/cf/*.cf -files were altered to have a new way of processing the "thishost" relation from $MAILVAR/db/localnames -file. * smtpserver/smtpserver.c: "-M ###" -option: Absolute maximum (size) ever accepted Code reorganize (a bit), load-aver blocked-access code "421" instead of "400". * hostenv/*, libsh/execute.c, scheduler/transport.c, smtpserver/smtpserver.c: WAITPID coding for POSIX.1 waitpid() with WNOHANG .. Thu Jan 5 00:28:45 1995 Matti E. Aarnio (mea@oj287.astro.utu.fi) * libc/getzenv.c, libc/mail.c, libc/whathost.c, router/functions.c, router/rfc822.c, router/router.c, scheduler/mailq.c, scheduler/msgerror.c, scheduler/update.c, smtpserver/smtpserver.c, utils/listexpand.c, compat/rmail/rmail.c, compat/sendmail/sendmail.c, transports/errormail/errormail.c, transports/libta/ctlopen.c, transports/smtp/smtp.c, transports/libta/diagnostic.c, transports/libta/lockaddr.c, transports/libta/markoff.c, transports/mailbox/mailbox.c, transports/sm/sm.c, transports/hold/hold.c: Converted to use #include "mail.h" instead of the previous #include * smtpserver/smtpserver.c, lib/detach.c, lib/loginit.c, libauth/authuser.c, libsh/execute.c, libsh/trap.c, libsh/zmsh.c, router/functionc.c, scheduler/scheduler.c, compat/sendmail/sendmail.c, transports/hold/hold.c, transports/errormail/errormail.c, transports/sm/sm.c, transports/mailbox/mailbox.c: Converted to use the new "zmsignal.h" * include/zmsignal.h: NEW FILE Created a file with all the signal handling that various programs appear to need: SIGNAL_HANDLE(SIG,HANDLER) SIGNAL_HANDLESAVE(SIG,HANDLER,SAVEVAR) SIGNAL_HOLD(SIG) SIGNAL_RELEASE(SIG) SIGNAL_IGNORE(SIG) (.. and "#define SIGNAL_TYPE void" if it is not defined when using this header file -- libauth/authuser.c ! ) * transports/smtp/smtp.c appendlet(): When decoding MIME QP on flight we MUST prepare to meet pathologic input, namely a last line with trailing "=" (so, no trailing "\n" ...) When that occurs, we must inject the "\n".. * router/functions.c: run_listaddrs(), run_listexpand(): Changed (a bit) of the error message those two can mail out. "From: Error Channel ", and "Precedence: junk". * scheduler/msgerror.c scnotaryreport(): Changed the function name -- clashed with libta.a .. Made sure it becomes called only when there is real NOTARY-DATA available, not in a case when the NOTARY- DATA is "" ... * lib/rfc822scan.c, router/shliase.c: Been wrestling with proper procedures for handling one particular errorneous address ("From:" -line) which used to cause router coredump.. * transports/mailbox/mailbox.c: Got a report that the new MAILVAR-misconfig-reporter misses a case of valid MAILVAR when the user does not (yet) have a mailbox file in the mail-spool. Hmm.. (Have not been able to reproduce this SunOS 4.1.3U1 reported error on my Solaris, have to try SunOS yet..) * libsh/interpret.c fapply(): When analyzing argument list (lisp-list), used wrong data source for the for-loop continuation.. (Affected "runas" facility running "filepriv"...) Mon Jan 2 00:03:33 1995 Matti E. Aarnio (mea@oj287.astro.utu.fi) * transports/hold/hold.c: From way back in history a "hold/home" -condition, though I am not sure it really is that usefull. * Makefile: Patchlevel: 7 * README.UPGRADING: (new file) Made some notes on what changes the ZMailer has had during its evolution, and thus what things users should be aware when upgrading from the previous versions. * router/libdb/ordered.c: Removed the "seq_remap()" from here too. Fri Dec 30 23:36:03 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * Makefile: Patchlevel: 6 * Dependencies: Changed the order of *dbm -entries. The lattest GDBM (1.7.3) has compability functions with DBM and NDBM. Thus their linkage-order can be used to chooce which *DBM is really used. (And especially if there is a wish to use genuine NDBM in parallel with GDBM!) * router/libdb/gdbm.c, man/router.8: Updated it truly to GDBM 1.7.3, and sent wishfull thoughts to the bug-gnu-utils@prep.ai.mit.edu for a real gdbm_filefno() function... The present one is, well, err... Added a few words of advice to the man-page of the router. Using GDBM from a non-standard include/library location is a bit painfull, having following kind defines on hostenv/XXX may help: GDBM_INCL= -I/opt/gnu/include GDBM_LIB= -L/opt/gnu/lib -lgdbm (The example assumes your GNU-things are on /opt/gnu/ ..) * hostenv/Ultrix*, libmalloc/getmem.c: Defined a NO_MADVICE cpp-define to have USE_MMAP usable on Ultrix 4.3A (well, I am guessing, I don't have the machine). On a case I got a report about, the system does not have madvice(2) while has defined MADV_RANDOM, why ??? Some odd library to be included ? I DO NOT KNOW IF Ultrix3.* HAS MMAP() AT ALL! It is historic stuff anyway, which nobody uses anymore, right ? * router/libdb/unordered.c, man/router.8: Disabled/removed seq_remap() and its associated fstat() call. It is better to use "-m" on the relation definition. (Documented it, as if it would be enough to teach people..) * libsh/trap.c: Changed the message of pre-compiled FC-file magic mismatch into something which is clearer - I hope. It reports about automatically recompiling the CF files to FC files, nothing else. * smtpserver/smtpserver.c, smtpserver/Makefile, smtpserver/fdstatfs.c: Localized all *stat*fs() versions into one file, which now returns valid (?) data for the most systems that are around. (Prompted by "Yet Another Special Case" -- Ultrix..) Thu Dec 29 08:31:42 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * Makefile: Patchlevel: 5 * smtpserver/smtpserver.c: Gives "a bit" mystic result-codes when the EHLO response generation procedure is unable to create a mail-spool-file -- it was used earlier to determine available free storage, but when I became informed that using it in the "SIZE" code would mean: "this server will never accept more than this size.." which for such a dynamic quantity is clearly wrong.. * transports/smtp/smtp.c: Leaked file-descriptors -- the SMTP channel connections.. ( smtpopen() leaked them, to be precise.. ) * Makefile: Patchlevel: 4 * transports/smtp/smtp.c, transports/sm/sm.c, transports/mailbox/mailbox.c, router/libdb/unordered.c: The case of NOT having USE_MMAP had things missing (being non-tested, but one Ultrix version was done with it..) Wed Dec 28 01:51:49 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * router/functions.c run_listexpand(): Fixed (again) an uninitialized "al" variable. Odd that I THOUGHT of having it fixed at the master source... (I thought this was a bug in ssift/tsift, but no.) * Makefile: Set version to be: 2.99.3 * transports/mailbox/Makefile: Remove the reference to dotlock.o -- it is INCLUDED into the mailbox.c, thus no separate compilation! * router/rfc822.c, router/rfc822hdrs.c: Gave a deep look to the analysis of "Sender:" -header. In a case when the provided "Sender:" is junk, like: to demo From: demo2 Sender: demo3-[12] it used to pick up the mal-formed "Sender:" header, which is stamped as BadHeader ! .. and used that junk on a call to the router() -- amazingly the router() didn't crash on it.. Now it is a bit more persistent on picking up only valid headers (so headers which are not stamped as "BadHeader"), and use them. If no valid sender header is found, one mkSender() was added for creating a valid sender-header as a last-ditch measure.. * scheduler/transports.c: Upped the number of FDs for a child-processes from 10 to 32. Solaris childs were unable to mmap() all the necessary dynamic libraries :-/ * libsh/builtins.c: sh_elements() Replaced it with the new Toronto (22e4) version. * libsh/{interpret.c,sh.ssl,sslwalker.c,optimizer.c,zmsh.c}: Found out that tsift DID NOT work :-/ A lot of wrestling and spotting of tiny details got it finally running... Sat Dec 24 00:46:15 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * Makefile: Did set version number to be 2.99mea * Overview, INSTALL, README, README.solaris, README.HP-UX, SiteConfig*, Config*, hostenv/00README, doc/guides/*, man/*: Updates, but no real DOCUMENTING! (that is, no new ZMOG...) * lib/nobody.c: A bit of SVR4 features -- will automagically be able to use correct value for "nobody" uid, if no ZENV variable NOBODY is set -- or if it is set to "NOBODY=nobody" Fri Dec 23 08:50:14 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * Dependencies: Added "kvm" to the "smtpserver", and "gdbm" to several others. ALSO DOCUMENTED THE FILE! * smtpserver/loadaver.c: Made some TRUE load-average extracters: - __linux__ - SunOS 4.x - SunOS 5.x * scheduler/resources.c: Made some TRUE resource codes for various platforms: - _POSIX_SOURCE - __linux__ - SUNs (SunOS & Solaris) * transports/smtp/smtp.c smtpopen(): Trevor Paquette spotted a missing execution path on case when '-E' parameter is used for the smtp transporter, and there is no "ESMTP" on the remote host initial banner. (But my fix is different with less code duplication..) * libsh/execute.c, libsh/interpret.c, libsh/io.c, scheduler/msgerror.c, smtpserver/smtpserver.c, transports/sm/sm.c, transports/smtp/smtp.c, router/functions.c, compat/sendmail/Makefile.in Several changes by Byron Rakitzis to suit BSDI (and probably most of the current BSD 4.4) Added also new variable for the Config: SENDMAILLIB=/usr/lib becase for BSDI, BSD4.4, etc. it is desired to be: SENDMAILLIB=/usr/sbin Thu Dec 22 10:38:26 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * transports/smtp/smtp.c, transports/mailbox/mailbox.c: Pulled some more bits from Edwin Allum's Zmailer-2.2e4 * router/libdb/bind.c: From Edwin Allum's Zmailer-2.2e4 a resolved retry tuneup -- to speed the resolver to give up when there is no data coming up.. Thu Dec 15 23:29:59 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * scheduler/mailq.c: Altered a bit the style of "nonlocal" settings. Now the thing should not jam on "localhost" being unavailable, because DNS says so... * hostenv/*: Copied in Edwin Allum's configs for IRIXes, and Guy Middleton's AIX3.2 -- and added DOTLOCK scheme ( + some comments ) to SunOS4.1 * sendmail/sendmail.c: fchown() call was missing GROUP parameter. Added there "-1" * transports/mailbox/mailbox.c: Created probably_x400() -function, and altered error reports on anonymous delivery to a file ("/...") to reflect possible X.400 addresses, when such is likely. ("/X=XXXX/Y=YYYY/...") Merged in Ken Lalonde's/Edwin Allum's DOTLOCK facility. * proto/zmailer.sh: Small change on the parameters of the router startup: was: ... router -dkn$NROUTERS new: ... router -dkn $NROUTERS Now it should start more than 9 routers... (On some platforms there is a behaviour where former would not start 10 routers, only one, while the later would start all 10.. -- On Solaris 2.3 it works with both ways, SunOS 4.1.3 (I think) had problems. ) * libsh/interpreter.c, libsh/sh.ssl, .... Altered sift/in/tfis construction to do STRINGWISE regexpr matches with GNU regex engine, and converted the old behaviour to name: tsift/in/tfist -- a bit later altered the names to be: sift/in/tfis -- Tokens - compability alias tsift/in/tfist -- Tokens (new PRIMARY name) ssift/in/tfiss -- Strings (new name) Wed Dec 14 09:12:16 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * transports/smtp/smtp.c: When doing verbose trace to a host without any MX-records, Solaris 2.3 crashed the smtp transport -- printf() doesn't like NULLs as string pointers... * transports/libta/lockaddr.c, transports/libta/ctlopen.c, include/mail.h, scheduler/scheduler.c, scheduler/mailq.c, router/rfc822.c: In attempt to create a state-preserving safe multi-invocation transaction locking: A PID based locking -- THIS ALTERS THE SCHEDULER FILE FORMAT AND THUS OLDER TRANSPORTERS MUST NOT BE USED IF THE ROUTER IS NEW! (The scheduler must also match!) Recipient-definition line (lock location) was altered: Old: "r~...." New: "r~PPPPPP...." where "~" is lock-flag, and "PPPPPP" is space for storing transport-client PID -- or 6 spaces, when there is no lock. From now-on, if there is a lock-state on some recipient, the scheduler tests to see if there is a process (pid) processing it, and continues with next job, dropping the locked one for a moment. All NEW jobs arrive both to the $POSTOFFICE/scheduler/, and the $POSTOFFICE/transport/, collecting them happens from the $POSTOFFICE/scheduler (so the current queue need not to be touched), however while there are locked files running, the scheduler will re-scan the $POSTOFFICE/transport/ directory every 5 minutes. If a file from that scan has its i-node number in spt_mesh[L_CTLFILE] -database, it is skipped (it is an active file, not some previously locked one). * transports/smtp/smtp.c: deliver() Oops on parameter processing, the forced 8-bitness didn't happen.. * router/libdb/bind.c, transports/smtp/smtp.c, transports/hold/hold.c: resolver usage got a little change -- to avoid screwing up request size parameter when doing retry.. (size storage variable was also used as a return value storage -> query screwed up on the retry..) * proto/cf/*.cf: Many files, minor changes... (And a couple major ones) Tue Dec 13 13:26:05 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * smtpserver/loadaver.c: NEW A stub for getting the system load-average * smtpserver/smtpserver.c: Respond with a "400 Sorry, system is too loaded for email reception at the moment" when the system load-average (from loadavg_current()) is over a given threshold value (XX: that message is most likely wrong.. ) (Yes, correct one: "421") * router/functions.c: run_listexpand() The pseudo-lisp is tough stuff to get right... Finally it does any arbitary expansion with very little memory consumption :-) => Rendered the util/listexpand.c obsolete... Mon Dec 12 20:42:59 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * router/libdb/ordered.c, router/libdb/unordered.c, router/libdb/search.h: Use USE_MMAP capability and map the textual database file into memory for later use -- will need less syscalls for the lookup... * transports/smtp/smtp.c: writemimeline() Counter each byte twice on the output when producing Q-P encoded output -> output was circa 35 chars wide... Oops.. Fri Dec 9 08:15:15 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * transporters/smtp/smtp.c: One missing "break;" on getopt() processing. Oops.. Thanks to Terry Combs for the fix (and I guess he uses the "-E" option..) * scheduler/ The Scheduler got some revampings into the mux()/ipcpause() routines, and their invocations so that inter-reaper delay should shorten -- should do it via signal handler ? * router/libdb/header.c: Re-defined the envelope environment entry "authinfo" to be of nilUserSemantics -> it will accept any input * router/rfc822hdrs.c: hdr_print() Changed the printout of DateTime -header so that when a previously existing header is available, use it as is, but skip the leading white space -> no more EXTRA SPACE/TAB every time a message goes thru the Zmailer. * util/listexpand.c: Add envelope environment entry "via listexpand" to the produced (new) file, thus the output will be like this: Received: by nic.funet.fi via listexpand id <92706-1>; Fri, 9 Dec 1994 08:01:41 +0200 Received: by nic.funet.fi id <92690-3>; Fri, 9 Dec 1994 08:01:37 +0200 * transports/sm/sm.c: * transports/smtp/smtp.c: "USE_MMAP"-version had problems at scanning for/advancing/ pointing to proper location in the input file. The net result was a loop in sm, and wrong contents in smtp. * smtpserver/smtpserver.c: *Sigh* Solaris signal handling is more complex than "just" "signal(SIGCHLD,reaper)" in the BSD... (a generic problem) Luckily it is enough to handle it with signal(SIGCHLD,SIG_IGN) which works pretty much everywhere. Fri Nov 25 09:41:24 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * transports/smtp/smtp.c: One NULL passage to the sprintf() caused crash on Solaris, same one prints "(null)" on SunOS... * lib/cfgets.c: Forgot to keep an eye on counted length, and thus were able to overflow the buffer... Uargh.. ("--n;" missing..) Wed Nov 16 22:57:22 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * router/functions.c: run_daemon() Use rewinddir() instead of seekdir(). The rewinddir() works better on BSDI systems very least, and apparently works on a major part of the systems (I have yet to come by a system where it doesn't work..) Fri Nov 11 11:44:28 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * hostenv/README -- some documenting things (barely begun..) * transports/libta/ctlopen.c, include/ta.h, transports/smtp/smtp.c, transports/mailbox/mailbox.c, transports/sm/sm.c: Made a case of using mmap() to map into memory the sharable (R/O) parts of the mail-file.. * router/rfc822hdrs.c: mkMessageId() Altered the created message-id format a bit, previous timezone name "(ett)" will now be presented as "ett", and the file will get a unique sequential number on it as well (unique within each instance of the routing process.) Thu Nov 10 10:01:55 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * Implemented "Content-Conversion: prohibited" -header to the transporters: transports/sm/sm.c transports/smtp/smtp.c transports/libta/mimeheaders.c transports/mailbox/mailbox.c (follow the suite of the Sendmail+Emil..) * Eradicted bcopy() from the sources. We can use memcpy() (possibly inline, possibly optimized library stuff) easily. Files: lib/linebuffer.c lib/token.c libc/bcopy.c libc/whathost.c libsh/expand.c libsh/io.c libsh/zsh.c router/functions.c router/rfc822hdrs.c scheduler/transport.c scheduler/readconfig.c scheduler/mailq.c router/libdb/bind.c transports/smtp/smtp.c transports/mailbox/mailbox.c support/nfslock/nfslock.c Didn't eradict bcopy() from support/vacation/vacation.c ! It is a bit sensitive issue in there... * Incorporating patches for OSF1v2.0 by Franz Fischer lib/allocate.c, libc/getdtblsiz.c libsh/io.c, router/libdb/bind.c router/prototypes.h, router/router.c scheduler/msgerror.c, smtpserver/smtpserver.c transports/smtp/smtp.c, utils/makendbm/makendbm.c utils/makendbm/ndbmlook.c Those were some ALPHA related problems on the DNS handling, and related declarations... Also a couple "beauty points" at the makendbm tools. * router/libdb/gdbm.c, hostenv/Linux Finally re-tried compiling Zmailer on Linux, and caught problems... GDBM has changed since my previous touch with it. Mon Oct 10 01:07:24 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * router/functions.c run_expandlist(): New "lisp-like" function to combine the functions of the listaddresses, and zsh-script routine maprrouter(). With this the router core expansion should minimize. Fri Oct 7 16:21:53 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * lib/allocate.c: Replaced one reference to bcopy() with memcpy().. * transports/sm/sm.c: Edited in the same routine as is in the transports/smtp/smtp.c to handle MIME conversions -- some details do vary.. * transports/libta/mimeheaders.c, include/ta.h, transports/smtp/smtp.c: Still more restructuring, now clean (?) way to make all the TEXT/PLAIN conversions on transfer: -- 8BIT -> QP -- 7BIT -- no change -- QP -> 8BIT (if the target is "force_8bit" ...) Thu Oct 6 09:10:09 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * transports/smtp/smtp.c: Restructured header pre-processing, now it can downgrade "TEXT/PLAIN; CHARSET=ISO-8859-*"+"C-T-E: 8BIT" into "TEXT/PLAIN; CHARSET=US-ASCII"+"C-T-E: 7BIT", when it notices that message body is all in 7-bits.. * transports/libta/mimeheaders.c downgrade_charset(): A new procedure * man/* Some updates to man-pages.. * SiteConfig*, Makefile{,.in}, libc/Makefile{,.in}: "make install" so that it installs also libzmailer.a into user-defined LIBRARYDIR, and zmailer.h into user-defined INCLUDEDIR.. Thu Sep 29 09:55:40 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * transports/mailbox/mailbox.c: Mailbox access-time preservation test was the wrong way :-( Wed Sep 28 16:22:40 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * transports/mailbox/mailbox.c: Nasty rewrite bug, which nobody had found for ages.. (Nobody but myself uses the MIME-rewritters ?) Mon Sep 26 09:09:15 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * Dump 2.97mea * transports/mailbox/mailbox.c: - MMDF-style mailbox separators with '-M' option! - MIME-processing is now line oriented to facilitate easier improving of it into handling beasts like "message/alternate" with multiple bodies.. Thu Sep 22 18:55:47 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * transports/smtp/smtp.c, transports/sm/sm.c, transports/hold/hold.c, transports/mailbox/mailbox.c, transports/errormail/errormail.c, scheduler/msgerror.c, scheduler/update.c, scheduler/transport.c, scheduler/scheduler.c, smtpserver/smtpserver.c, libc/whathost.c, lib/detach.c, transports/libta/warning.c, libmalloc/malloc.c, libmalloc/getsize.c, libc/getdtblsiz.c, libsh/regexp.c, libsh/execute.c, libsh/interpret.c, lib/rfc822date.c, router/db.c, router/functions.c, transports/libta/ctlopen.c, hostenv/Linux: Did port to Linux (Various POSIX.1 problems.. Usually solved by "#include NDIR_H" ..) * transports/libta/buildbndry.c: Added sequence "=_" into the boundary, thus rendering it such which most likely will not be generated by any MIME-compliant system in their message texts. * Makefile, libsh/Makefile, proto/Makefile: Got rid of /bin/ed on file editing. Instead uses AWK and SED. (Portability problems to Linux caused this..) * Config, SiteConfig, bin/mklibsupport, ... Configuring changed somewhat: - Compiler/compilation dependent things are on file Config - Site policy things (location of directories) are on file SiteConfig * transports/mailbox/mailbox.c ( putmail() ): When CREATING a (mailbox) file, create it with ATIME of 0, that is, "NEVER READ"... Fri Sep 16 12:04:25 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * router/functions.c ( run_filepriv() ): Added option -M maxfilepriv to the $(filepriv XXX) to get rid of a need to have system-wide compiled value for it. Now I think we can create a system with 644 for ~user/.forward and 664 for $MAILSHARE/lists/... My users did ask for it.. Thu Sep 15 17:03:58 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * transports/mailbox/mailbox.c ( putmail() ): Added ONE space to the FROM_ line: From SPC from@addr SPC SPC date ^^^ NEW (To be compatible with competition -- err.. sendmail..) Sat Sep 10 02:37:31 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * After a few days of testing at MAILHOST.UTU.FI, dumped 2.96.. * smtpserver/smtpserver.c: More fixes to the complex SMTP-option handling.. * scheduler/readconfig.c: SIGUSR1 -- rereadconfig() -- did crash on memory corruption due to double-freeing a buffer. * scheduler/*.c: * transports/libta/ctlopen.c: More fixes on the DSN, and related topics. * transports/mailbox/mailbox.c: A message from "channel error" (SMTP MAIL FROM:<>, or alike) got bad UNIX mailbox header. Now the "From_"-address is always "postmaster" under such circumstances. * transports/smtp/smtp.c: Do recognize the "DSN" (and "XDSN") magic tokens of NOTARY-DSN capability on the remote system. (Missing: actual DSN interaction..) * router/rfc822.c: ( sequencer() ) * router/libdb/header.c: Will accept DSN parameters, and carry them as attributes to the router tasks, and in the end will output them to the channels. The scripts need to address problems on the DSN data alteration, when it comes to list-expansions. (Specs are developing..) Thought: Internal list-expansion needs to be expanded into a full-scale C-code which invokes router for the recipients, and feeds them all the necessary attributes. That way it will (perhaps) be SMALLER with its memory usage as compared to the present maprrouter() script function. * router/rfc822hdrs.c, router/rfc822.c: rfc822date() returns delivered UNIX-time in RFC822 format INCLUDING TIMEZONE CORRECTION, but we actually want to create it only when it doesn't already exist, otherwise it shall be left alone. Trouble on translating to the local TZ is on several mismatches in the time-zones, and frankly it is too painfull to be really usefull. When the full header- string is there already, off we go and use it. * router/rfc822.ssl, router/rfc822walk.c, router/rfc822hdrs.c: Parse and "understand" "Received: ... convert XXX;" -token! Fri Sep 2 11:35:53 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * scheduler/scheduler.c: * scheduler/scheduler.h: Initial code for parsing DSN related transporter header entries. * scheduler/msgerror.c: * scheduler/scheduler.h: Print out "Original-Envelope-Id:", and "Original-Rcpt:", if data is available (if not, be quiet about them.) * smtpserver/smtpserver.c: Accepts (and syntax-checks!) IETF-NOTARY-DRPT parameters, and checks upon illegal characters on the input line, like feed of control/8-bit chars on the SMTP protocol line.. (ASCII 0:es are also flagged!) * include/mail.h: _CF_RCPTNOTARY-tag -- "DSN parameters for preceeding recipient" _CF_DSNENVID-tag -- "DSN ENVID data" * include/mailer.h: enum HeaderClass got new elts: eToDSN, eEnvid * router/libdb/header.c: Learned to recognize new transport envelope headers: todsn, envid * include/ta.h: * transports/libta/ctlopen.c: "struct ctldesc" has now "envid" field, and it gets filled in on the ctlopen() parser.. There is also "dsnflags", which also gets filled.. Thu Sep 1 11:13:03 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * libsh/optimizer.c: Bad interaction with USE_ALLOCA() -- the ncode[] array MUST be returnable to the callers scope, thus it can't be alloca() buffer... Mon Aug 29 22:45:09 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * Dump 2.95mea, and put it running at NIC. Public release tomorrow ? * transports/smtp/smtp.c: EHLO-processing had parameter bug on smtpwrite()... * scheduler/msgerror.c: * scheduler/scheduler.c: Still more wrestling with NOTARY error reporting.. Now it prints REWRITTEN headers also from within scheduler's error reporter... Sun Aug 28 22:26:26 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * transports/mailbox/mailbox.c: * transports/smtp/smtp.c: * transports/libta/diagnostics.c: More wrestling with NOTARY error reporting.. * transports/errormail/errormail.c: More wrestling with NOTARY error reporting.. Will write out REWRITTEN headers -- so that there is local "Received:" entry as well! * scheduler/msgerror.c: Still more wrestling with NOTARY error reporting.. Wed Aug 24 02:00:17 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * Dumped 2.94mea * smtpserver/rfc821scn.c: Address like: is to be accepted, even though it is illegal, if you have a strict look at the standards.. * transports/mailbox/mailbox.c: * transports/smtp/smpt.c: Testing NOTARY message production. * transports/libta/diagnostics.c: Produce the NOTARY diagnostics ( notaryreport() ) * scheduler/transport.c: Had a peek as to why some machines (SunOS 4.1.3 most notably) can't run more than about 50 (!) channel programs at a time! (Marco Hernandez @ CREN.ORG reported this problem..) At the moment: NO IDEA :-/ (20 hours later: Each scheduler file is kept open when it is in active processing, each active transporter has a diagnostics pipe back to the scheduler, and some systems have SIGNED CHARs as FD storage in FILE- structures.. If transporters run with GANG-SCHEDULER, that is multiple files are spooled to same transporter, all those files must be kept open by the scheduler until the transporter acks them back.. ) (1-Jan-95: Most likely it is Open-Files resource limitation..) * scheduler/msgerror.c, scheduler/update.c, transports/libta/diagnostic.c: Revised report-back format, as well as error reporting procedure fairly completely. Now the transporters CAN report IETF-NOTARY data, and the scheduler will then use it to report on the return email. (Exact format evolves still, but one thing is sure: it shall not contain \n's, nor \t's!) Transporter main error reporter is scheduler/msgerror.c, NOT the transports/errormsg/errormsg.c ! The later can be used for some trivialities, though.. (How the heck it can get NOTARY data ? -- solved three days later..) Tue Aug 23 21:52:18 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * transports/smtp/smtp.c: An error in EHLO response analysis did omit the final response line, which in one case did have effect: 250-XXXXX 250 8BITMIME Thus SMTP didn't learn that the receiver is 8-bit capable.. Added an option to smtp: -E -- try EHLO only if the remote system has "EHLO" on one of its initial message lines... If not, don't do it. Default case is to do EHLO ALWAYS, and if it fails ("500 Unknown command", or hangup.., use HELO..) * transports/errormail/errormail.c: * scheduler/msgerror.c: Skip over the Zmailer envelope lines in the original email so that the file to be written out does not contain anything but RFC-822 (+MIME) headers.. Wed Aug 17 14:12:06 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * utils/listexpand.c: Wrote this to test how much 650 email recipients will really expand the memory of the router. Instead of running a built-in expansion within the aliases.cf, this is to be called externally to produce similar remapping and resubmitting the file back to the router. (Via a trip thru scheduler, that is) Internal expansion: 32 MB External expansion: 3.5 MB (yes, only about 10 %!) * libsh/trap.c (eval): Carelessness on referring to the "savefile" buffer. It can be NULL ptr under some occasions. ( -> newaliases crashed ) Tue Aug 16 17:04:43 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * Freezing and dumping 2.93mea * *.cf -files: Processing of addresses which turn to be LOCAL to the machine will with sample files (proto/cf/) produce results like: (local nobody /dev/null 123) (local postmaster root 0) (local somealias "|/path/to/program args" 1234) instead of the old style: (local - /dev/null 123) (local - root 0) (local - "|/path/to/program args" 1234) (These DO NOT contain my earlier quote additions, which were for helping on expansions of esoteric addresses. Oh well.. 'Wrong' set of source scripts..) * many files.. FINALLY the mystic crashes on "router recipient" got explained. Stdio uses its own file descriptors, and it takes only so many (resource: filehandles) leaked (non closed) stdio buffers to use them all up, and even though there are lotsa free fd's, there is no free stdiobuf.. Mainly router/functions.c: run_listaddrs() Now the beast is written in portable manner, which only assumes UNIX fd semantics (and that dup() and dup2() work) Mystic libc/__fopen.c -file becomes used, if you define -DDEBUG_FOPEN, Then it traces the usage of stdio files.. * router/functions.c: run_daemon() Fixed the memory allocation/freeup. Memory that gets freed explicitely is to be picked up from MEM_MALLOC space, not from anywhere else (did cause some cores..) * libs/ -- all libraries, most Makefile's, .. All library files (*.a) get collected into one directory, the "libs/"-directory. * transports/libta/mimeheaders.c: Occasionally a transported file had: Content-Type: text which caused carelessly written code to crash when subtype (as in "text/plain") was not present at all.. * hostenv/SunOS4.1: Define GETDTABLESIZE so that libc/getdtblsiz.c won't get compiled into a code and thus override system library routine.. * hostenv/SunOS5.2: Use only "MAILLOCK" on Solaris, not the LOCKF at all! * transports/mailbox/mailbox.c: On Solaris the file locking shall not be used on "/dev/null".. Incorporated Edwin Allum's Solaris patches into the code -- maillock() et.al. * transports/smtp/smtp.c: Some minor tuneups on detecting when to downgrade headers, and when not to. * libsh/interpret.c: Under some conditions under-read varstack. PURIFY test use did help me to detect that at all.. ( * At some non-defined point in history did dump zmailer-2.92 --- 940802 it appears to be. ) Wed Jul 27 02:31:28 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * Freeze and dump zmailer-2.91mea-940727.tar.gz * many files.. Converting emalloc(strlen(buf)+1) + strcpy() to strsave() did cause amazing amount of problems. They aren't all gone (don't know for sure, though), and some others may have surfaced. * transports/errormail/errormail.c, scheduler/msgerror.c: Alter error reporting towards that of the IETF NOTARY working group draft -- not yet there by million miles, I am afraid.. THIS ALTERED ALSO THE SYSTEM ERROR MESSAGE FILES! * */Makefile{,.in}: Lotsa things.. Mainly on getting things to compile ALL the related libraries before linking them in. * transports/libta/ctlopen.c, transports/libta/mimeheaders.c, transports/libta/writeheaders.c, ... Improve email header conversion facilities. Proper email BODY conversion is still missing.. (text/plain gets processed, but other formats call for improvements..) Mon Jun 27 07:24:44 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * transports/libta/mimeheaders.c : Improved processing of "Content-Type:", and added "Received: .... convert rfc822-to-XXXX;..." clauses. ( to "quoted-printable" and to "8bit") * BUGS, ChangeLog, README, README.solaris, INSTALL : Updated documentation, worklist, .... Sun Jun 26 01:31:54 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * Freeze and dump zmailer-2.90-940626.tar.gz * router/functions.c (run_listaddrs): An empty "include" file (empty .forward, as was the case) caused SIGSEGV due to a NULL-ptr dereference. Created smarter error handling. * router/rfc822.ssl: Teached "AMailboxList" to understand '<>' as a valid address ("channel error" source..) as is often case with error-messages from mailers.. Now it accepts: From: <> That unearthed a lot of problems elsewere with NULL-pointer dereferrence and now don't even remember where everywhere... ( s_copy_tree() at least ) Sat Jun 25 01:31:54 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * libsh/builtins.c, libsh/path.c, libsh/vcall.c, router/functions.c, router/rfc822.c: USE_ALLOCA -- use alloca() for some local buffer allocations which don't need to be preserved over the scope of stack-frame existance. * hostenv/ : Don't select on own getopt() vs. system supplied one. Use the own one always. * libsh/interpret.c: Improve interaction with the getopt() * libmalloc/ : Improved the integration with the Zmailer -- this is a new version of the "Toronto Malloc", and quite a headache to integrate.. Tue Jun 21 09:52:21 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * router/rfc822.c (run_rfc822): Argument processing bombed -- possibly due to usage of the GCC, possibly due to something else. Tue Jun 21 09:52:21 1994 Matti E. Aarnio (mea@oj287.astro.utu.fi) * Created this file to track changes Building Zmailer 2.90xxx -- that is, alphas/betas of 3.0 * Pre-ChangeLog transcript follows: Various features/fixes done at Univ of Turku by Matti Aarnio Compiling with GCC 2.3.3, and 2.4.5 with -Wall -- generated a lot of warnings, which one after another have been eradicted.. Found also some real bugs with that. (BUT NOT ALL!) I am compiling it on SunOS 4.1.3 with GCC 2.5.6, and "-traditional" option.. Referred is IDA's big-brother (IDA-sendmail's MIME processing toolbox) Multilevel incoming mail priorizing into router: - zmailer(3) has mail_priority variable, which must be set to desired value before mail_close() is called. Value 0 means normal behaviour (and is the default). Any positive value means picking further down into ROUTERDIRS Z-environment variable. (See zmailer(3), and router(8)) - compat/sendmail/sendmail accepts -Pnn, which is feed to mail_priority (see above) As an alternate, it also honours environment variable MAILPRIORITY which has numeric value. - Router processes at first all files from the POSTOFFICE/router/ directory, and once it is empty, it looks into alternates. If alternate had anything to work upon, directory scanning returns to the first level dir. As an end result, the further messages are into the alternate directories, that less frequently they get processed. router/functions.c: router/conf.c: router/dateparse.c: router/rfc822.c: router/router.c: - Implemented first real multi-router system with working locks. - Extended it with ROUTERDIRS - multiple priority levels. - Added some {}'s into dateparse.c - Added configurability of $(filepriv ...) into conf.c - Accepted and inserted a patch on RFC header processing which blew on "header"-line like: ": some text" (Some programs inserted such junk into HEADERS - tin among them.) Source: "Michael S. Shappe" - Added option -E into $(listaddresses ...) creating an "Errors-To:" header, and making sure the SMTP 'MAIL FROM:<>' contains that address. router/libdb/bind.c: - added query-type 'any' to be used in place of '-t cname' on DNS canonization. (In manner of BSD sendmail.) smtpserver/smtpserver.c: - SMTP-server does check syntax of RFC821 inputs, specifically it checks "MAIL FROM:< .... >" and "RCPT TO:< .... >" syntaxes, and is adamant with them. -- many misconfigured PC-mail programs have been found with it... -- rfc821scn.c does most of it. - Uses Remote Authentication Service (RFC 1413) to (try to) see who is it, who contacts us. (See also: "libauth/*") - SMTP-server supports RFC 1425-1428 EHLO greeting (in place of HELO command), and related 8BITMIME, and SIZE extensions. Summa summarum: Now we advertice that we can accept 8-bit SMTP, not just blindly accept it.. - SMTP-server fully uses RFC 1427 SIZE feature (if user defines it), and thus can report "insufficient storage space at the moment", if sender tries to send email which is far too big.. transports/libta/routermxes.c (new) - router can feed target "mx" information, same as it got when analyzing routing, to the actual transporter program. In place of a hostname, it codes: ((mxer)(mxer mxer)(mxer)) (This is into the router results!) transports/libta/ctlopen.c transports/libta/mimeheaders.c (new) - new way to read in the headers, and to handle them when rewriting them. Original was to get them in as single multiline string. This splits them to individual lines and keeps track of them in that way.. transports/smtp/smtp.c: - Better support for local sender verbose trace - If connection setup fails (channel open timeout, or remote closes the channel upon our face -- whatever before HELO), we try to talk to next assigned MX target. - Support of RFC 1425-1428 8BITMIME transport - Uses RFC 1427 SIZE=, if receiving system knows it (EHLO protocol) - Downgrades message to "Content-Transfer-Encoding: QUOTED-PRINTABLE", 1) "Content-Transfer-Encoding: 8BIT" is present in the primary headers, 2) remote system has no 8BITMIME capability, 3) message contains 8-bit characters. - Downgrades message header to "Content-Transfer-Encoding: 7BIT", if message data is 7-bit, and said header is present. - If "C-T-E:" was turned to "7BIT", and "Content-Type:" was "text/plain; charset=ISO-*", then update "C-T:" to have "charset=US-ASCII" -- simplifies things (?) - Doesn't do full job of analyzing things like "Content-Type:", rather assumes that "C-T-E:" is present on "C-T: text/plain" messages, and any complex things are built properly. - setsockopt() arguments corrected. transports/sm/sm.c: - Support for local sender verbose trace - Downgrades messages in same manner as smtp does, if "sm" doesn't have "-8" in its options when running some subchannel. transports/mailbox/mailbox.c: - Allow local addresses to be user@domain, and use that kind of addresses on storing mails to mailboxes. - Use proper NFS locking ( lockf() ) - Partial MIME converter: A text/plain; Quoted-Printable message will be converted into 8BIT format, IF the "mailbox" program is called with argument: "-8"! (If that argument is not present, nothing new is done.) libc/libc.a: - mail_open() does a whathost() to figure out NFS-mounted postoffices. - using free()d storage -- Matthias Urlichs, et.al. in mail.c - whathost.c generalizes NFS mountpoint figuring system. - mail_close() is affected by mail_priority variable, as well as ROUTERDIRS Z-environment. See zmailer(3) libauth/*: - RFC 1413 using identd protocol using library. (Host configuration option USE_AUTH) compat/sendmail/sendmail: - Accept (and ignore) options: -oem, -J, -em - Accept option -Pnn, see above about mail_priority (zmailer(3)) - Making sure "to
" -lines contain only valid characters -- well, don't contain newlines! - Accept "From " (FROM-SPACE) input (sendmailism), and drop such line away, if next line looks like a real header. - Do binary (-t option) processing buffer at a time, instead of a character at a time.. compat/rmail/rmail: - Bill Wisner's rmail patch applied. lib/rfc822scan.c: - Nested comments fix -- don't remember who - Kicked out bit-stripping -- headers are 7-bit stuff! lib/linebuffer.c: - Explicit exits with ferror() amd feof() conditions! Suggested by haa@cs.hut.fi - Made a serious mistake, and two weeks later found it.. - accepted a patch from Ken Lalonde , finally did it correctly ! ssl/ssl.c: - array space reservation one-off, fix by Matthias Urlichs utils/makendbm/ - wrote whole package utils/zmstats.perl - wrote this small piece utils/fullnamealiasmaker.perl - wrote this (larger) piece support/vacation/ - extended to behave somewhat like SunOS vacation(1), only better :) scheduler/mailq.c: - Moved function getmntpt() to (new file) libc/whathost.c