The Time-stamp: line indicates when this file was last modified.
Time-stamp: <1998-10-20 18:48:44 james>
List of fixed bugs
Naturally there are many CSSC bugs that have been fixed but are not
listed here. Bugs are only listed here if they have ever been listed
in the BUGS file. Bugs that are discovered by the maintainers and
which are fixed before the next release are also probably not listed
(even if the big ones are mentioned in the NEWS file). This doesn't
mean that the unmentioned bug is not in any released version; it may
have been present in all previous versions, there just wasn't a
release made while this bug was KNOWN. The ``real'' criterion is, if
it's easier to just fix the bug than list it in BUGS _and_ I do it
promptly, it does not get listed. If you don't like this (lack of?)
policy, let me know. All bugfixes are listed in ChangeLog, in some
fashion.
Thu Oct 8 1998 12:11:43: Dick Streefland <dick_streefland@tasking.com>
In version 0.09alpha.pl0. Fixed in version 0.09alpha.pl1.
Core dumps if no SCCS file is specified on the "get" command line.
This was fixed by a patch submitted by Mr. Streefland.
Thu Oct 8 1998 12:11:43: Dick Streefland <dick_streefland@tasking.com>
In version 0.09alpha.pl0. Fixed in version 0.09alpha.pl1.
We omit to return a failure exit status of 1 from get and prs when
the output file suffers an out-of-space condition.
This was fixed by a patch submitted by Mr. Streefland.
Tue Jun 16 22:12:46 1998: Achim Hoffmann <hoffmann@tebis.de>
In version 0.07-pl3 [and most earlier]. Fixed in version 0.08-pl0.
On systems lacking the <string> header file, CSSC cannot open files
whose name as specified on the command line contains more than one
slash. This bug was also reported by Mr. Anthony C Cole.
Thu Feb 19 21:50:49 1998: Peter Kjellerstedt <peter.Kjellerstedt@axis.com>
In version 0.06-pl0. Fixed in version 0.06-pl6.
The -t option is missing from get. BSD's sccs.c uses it.
Wed Mar 11 18:01:54 1998: Peter Kjellerstedt <peter.Kjellerstedt@axis.com>
In version 0.06-pl4. Fixed in version 0.06-pl5.
If you use rmdel, your s.-file will end up totally corrupted.
This is due to the fact that all the lines which specify
what should be included/excluded for each SID are removed...
Tue Mar 10 23:00:45 1998: Wolfhard Straehle <straehle@ee.hnt.heller-machines.com>
In version 0.06-pl3. Fixed in version 0.06-pl4.
Absolute pathnames don't work for "admin -n" because it looks for
an "s." after the first slash in the filename, when it should check
after the *last* one.
Thu Feb 19 21:50:39 1998: Peter Kjellerstedt <peter.Kjellerstedt@axis.com>
In version 0.06-pl0. Fixed in version 0.06-pl3.
Trying to run tests/admin/i-option.sh with the CSSC files on
the sun machine [Solaris; SunOS 5.5] causes the test I3 to fail with
../../admin: fgetpos() failed!
29 - Illegal seek
- The fix is that we now no longer *ever* seek on the file designated
by the "-i" option, no matter what kind of file it is. If we need to
reread the initial data from it, we just read that back out of the
partially written x-file.
Thu Feb 19 21:50:35 1998: Peter Kjellerstedt <peter.Kjellerstedt@axis.com>
In version 0.06-pl0. Fixed in version 0.06-pl1
The option order seems to differ with SCCS and CSSC. We have
a number of tools which issue the command 'get file -Gfile'
which works correctly with SCCS, but for CSSC I need to issue
the command 'get -Gfile file' instead...
Thu Feb 19 21:50:54 1998: Dave Bodenstab <imdave@mcs.net>
In version 0.06-pl0. Fixed in version 0.06-pl1
The operation of the "admin" command for s.foo will silently destroy
any files named x.foo. The manual does not mention this.
[file is now renamed to x.foo.bak, message sent to stderr. Warning
given if the old file could not be renamed]
Thu Feb 19 21:50:44 1998: Peter Kjellerstedt <Peter.Kjellerstedt@axis.com>
In version 0.06-pl0. Fixed in version 0.06-pl1
The %I% %G% etc tags aren't substituted correctly (they are
substituted with the version and date of the file at the time the
tags were first checked in). I have provided a patch
to the tests/get/subst.sh file below. Also included is a replacement
for the keywords.uue file to go along with the
patch. [bug fixed, single test added, patch not yet applied -- jay]
Thu Feb 19 18:11:08 1998: Peter Kjellerstedt <peter.Kjellerstedt@axis.com>
In version 0.06-pl0. Fixed in version 0.06-pl1
prt -d:PN: shoulf prvide fully qualified pathname for s-file.
It does not. "prs -d:PN: s.1" produces just "s.1".
Mon Feb 18 1998: Richard Polton, the Rubber Buccaneer <richardp@scopic.com>
In version 0.06-pl1. Fixed in version 0.06-pl2
On Irix, we get a warning from GCC: "sfwrite.cc: comparison is always
1 due to limited range of data type" [jay: this is version 1.13 of
sf-write.cc, the code is "while ( EOF != (ch=getc(f)) )" and ch is a
char]. ch should be an int, not a char. [jay: if only LCLint worked
for C++!]
[...]
I think I can safely say that something went wrong becaise after a
brief pause in the output to the console I took a look at the
tests/cdc directory and found a 17MB file called x.testfile!
"s.testfile" did not exist. The last.command is
"admin -i testfile s.testfile". I get the impression that it did
not work as expected!
[jay: this is a lightly edited version: rap did not relalise that
he had already given the solution to the problem he was reporting.
What a hero!]
Mon Feb 9 23:47:08 1998: Marty Leisner <leisner@sdsp.mc.xerox.com>
In version 0.05-devel-pl0. Fixed in 0.05-devel-pl1.
SunOS 4.1.4 makes tests/common/test-common emit the "Unusual
system" warning regarding the options accepted by "echo", and
so the escape codes in the strings in the test suites don't work.
So we have to roll out own echo, I think. Very slow, no doubt.
-
Our own echo is in testutils/ekko.c, taken from GNU sh-utils.
It seems to work fine. tests/common/test-common has been updated
to use this if the system echo(1) is unsatisfactory.
Sat Jan 24 16:58:50 1998: James Youngman <jay@gnu.org>
In version 0.05pl0.
prs -d:FL: s.foo on an empty s-file produces
MR Validation:
Module Name:
when neither of these flags is set.
(This is now fixed.)
Sun May 4 13:12:57 1997:
./get -r2 -e s.foo doesn't create a new release, it just gets the most
recent delta on the trunk, if release 2 doesn't exist.
Sun May 4 18:03:20 1997: (fixed in version 1.4 of sf-get.cc)
Selection of the correct SID to get is still wrong in get. For
example if you specify -r1 you should get the last 1.x release, but
you don't.
Thu Nov 20 22:29:33 1997: Richard Polton <rap@maths.soton.ac.uk>
CSSC's "admin" adds users to the SCCS in the reverse order
compared to SCCS.
Mon Nov 17 18:01:05 1997: Richard Polton <rap@maths.soton.ac.uk>
In version 0.03-pl2. Unfixed.
If you create an SCCS file using "-i" from a file with (say)
500000 lines, real SCCS does it, but puts "99999" in the
summary line for that delta. Currently, CSSC puts the "correct"
number of lines, and then fails when reading the file back in
for some other purpose, giving the error "Number too big".
The correct behaviour would be to follow what SCCS does.
Local Variables:
Mode: text
End: