PR 23868 & gcc.dg/torture/stackalign/builtin-apply-4.c

[Steve Ellcey <sje at cup dot hp dot com>]

gcc.dg/builtin-apply4.c has failed on IA64 (HP-UX and Linux) for a long
time due to PR 23868.  Joern Rennecke submitted a patch to fix it on 
some platforms and I submitted another patch back in 2006
(http://gcc.gnu.org/ml/gcc-patches/2006-08/msg00765.html) that would fix
the problem on IA64 but neither patch was ever approved or checked in.

There was also talk at one time of fixing libobjc to not use
__builtin_apply and __builtin_apply_args and getting rid of them entirely
but that didn't happen either.

I am bringing this up now because rth's generic stack alignment changes
have caused IA64 to start running the gcc.dg/torture/stackalign suite and
gcc.dg/torture/stackalign/builtin-apply-4.c, which is based on
gcc.dg/builtin-apply4.c, fails on IA64 for the same reason.

Since it is in torture I am getting 8 failures of that test plus the original
builtin-apply4.c failure  and on HP-UX I get twice that many because I run in
32 and 64 bit modes.

I would like to fix these failures, the options I see are:

My 2006 patch, Joern's patch (extended to include apply_args_mode as well
as apply_result_mode), remove __builtin_apply and __builtin_apply_args,
or just XFAIL the tests.

Are there any opinions on the best way to fix this problem?

Steve Ellcey
sje@cup.hp.com