new syscall stub support for ia64 libc

[David Mosberger]

>>>>> On Thu, 13 Nov 2003 01:17:18 -0800, Ulrich Drepper <drepper@redhat.com> said:

  >> OK, I believe tst-cancel6 may be failing because cancellating the
  >> victim thread fails to release the lock on the file that it's trying
  >> to read from.  Is this test known to work for ia64 at all?

  Uli> All tests work on ia64.

OK.  Which compiler?  I'm using gcc-3.3.2 (provided by Debian).

  >> Is there a good way to trace what's going on during cancellation?

  Uli> There isn't any although rth was going to look into adding something
  Uli> like this.

I can build libunwind with debugging enabled, but I don't think it's
gonna do me any good, since all fprintf calls will get cancelled, once
cancellation has begun.  Would it perhaps be possible for glibc to
provide a non-cancellable version of fprintf?

Where can I find the definitely list of cancellable routines?  I'm a
bit worried about libunwind.  It may call mmap (for memory allocation)
and pthread_mutex_{lock,unlock} (when global caching is in effect).

	--david