This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Someone broke bootstrap
- From: Roger Sayle <roger at eyesopen dot com>
- To: John David Anglin <dave at hiauly1 dot hia dot nrc dot ca>, Eric Christopher <echritso at redhat dot com>
- Cc: Zack Weinberg <zack at codesourcery dot com>, <gcc at gcc dot gnu dot org>
- Date: Wed, 8 Oct 2003 20:23:28 -0600 (MDT)
- Subject: Re: Someone broke bootstrap
Hi David and Eric,
> > I'm seeing this on frv too. I blame Zack (at least in the little
> > bit of looking I did...), but that's ok. I'm sure he'll fix it.
>
> Yes, it looks as if this is caused by one of Zack's recent changes.
Mea culpa. Zack is completely innocent, the testsuite regressions
are my fault. Zack asked that I modify my patch to optabs.c's
prepare_float_lib_cmp to call emit_libcall_block directly, but I
screwed up the reorganization, not realizing that "equiv" mustn't
be a NULL_RTX, otherwise very bad things happen...
I've already posted a fix at:
http://gcc.gnu.org/ml/gcc-patches/2003-10/msg00635.html
Hopefully, it will be reviewed in the very near future.
The problem shouldn't affect bootstrap as it only shows up with
-O0 when using software floating point, but it is likely to be
the cause of FRV's/HPPA64's regressions.
Sorry again for any inconvenience. The problem doesn't show up
on x86, and even the -msoft-float tests are never run at -O0,
so it wasn't found during my retesting. The original patch to
prepare_float_lib_cmp was more widely tested without any problems,
but I underestimated the "obvious" tidy-up.
Roger
--