fix ia64 longjmp() to work from alternate signal-stack

[David Mosberger]

>>>>> On Mon, 13 Sep 2004 11:39:34 -0700, Ulrich Drepper <drepper@redhat.com> said:

  Uli> I've applied the patch after minor changes.  There is no need
  Uli> to require the sigaltstack function being exported from
  Uli> libc.so.  It's a simple syscall, just inline it.

OK, that's certainly fine by me.

  Uli> The result seems to work for me but then, I never saw the
  Uli> problem this is supposed to fix in the first place.

I re-checked and the test-suite (in particular tst-cancel20) still
passes, so everything seems fine.  (As a reminder, you don't see the
problem with a GCC that's using the built-in unwinder since the
built-in unwinder maskes the problem due to some excessive copying;
the problem only shows when using a GCC that has libunwind enabled.)

Thanks,

	--david