This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: PR 23868 & gcc.dg/torture/stackalign/builtin-apply-4.c
- From: Richard Henderson <rth at redhat dot com>
- To: sje at cup dot hp dot com
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Fri, 15 Oct 2010 15:18:59 -0700
- Subject: Re: PR 23868 & gcc.dg/torture/stackalign/builtin-apply-4.c
- References: <201010152106.o9FL6K213852@lucas.cup.hp.com>
On 10/15/2010 02:06 PM, Steve Ellcey wrote:
> 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?
I guess we should fix __builtin_apply. I had really hoped that
objc could be converted to libffi, but that never seems to have
happened.
In http://gcc.gnu.org/ml/gcc-patches/2006-08/msg00765.html
Joern is quite right that you're not finding the right mode.
Consider
_Complex long double f() { return 2 + 2i; }
which returns its result in both f8 + f9. Thus it would
seem that the mode you must use for "f8" is XCmode.
Thus the "minimal" fix appears to be Joern's first patch.
Although I'm certainly not sure what's going on with some
of the mode-switching code...
r~