$Id: BUGS,v 1.8 2001/03/21 00:05:03 shane Exp shane $

- On Solaris, new POSIX threads cannot be created from a chroot() jail!
  pthread_create() uses mmap() on /dev/mem (or something like that) to
  create the stack for new threads.  However, in a chroot() jail, no
  /dev files are visible (hopefully).  It would still be possible to
  support Solaris by using malloc() and specifying the stack explicitly,
  but then there's the hassle of cleaning up this memory and so on.
  Yuck.  So for now, no Solaris.  In 0.5.x I expect to support Solaris
  again, as I'll only need a very small number of threads.

- The pthread_cancel() does not actually meet POSIX compliance.  It is 
  forbidden to use PTHREAD_CANCEL_ASYNCHRONOUS on anything that makes
  system calls.  This seems to work fine on Linux, probably due to its
  bizarre thread model.  This will also be fixed on 0.3.x, as I'll pay
  careful attention to cancel points.

- Probable memory leak when pthread_cancel() operates on a thread that
  has allocated memory (in glob() for instance).  Need to disable cancel 
  for appropriate regions of code.

- Doesn't handle permissions properly.  If a file is world-readable but 
  not group-readable and someone in the group tries to read it they
  should not be allowed, apparently.  :(

- strerror() may not be thread safe (depends on implementation).  Write
  a thread safe version.