This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Patch to decl2.c
- From: Mark Mitchell <mark at codesourcery dot com>
- To: David Edelsohn <dje at watson dot ibm dot com>
- Cc: Geoffrey Keating <geoffk at apple dot com>, Richard Henderson <rth at redhat dot com>, gcc at gcc dot gnu dot org
- Date: 05 Sep 2003 13:48:58 -0700
- Subject: Re: Patch to decl2.c
- References: <200309052010.QAA23664@makai.watson.ibm.com>
On Fri, 2003-09-05 at 13:10, David Edelsohn wrote:
> Mark,
>
> Geoff's patch also fixed bootstrap on a small memory AIX system.
> Richard's patch had no effect.
>
> If you want to change the patch, fine, but it is unacceptable to
> have the regression testers fail for such a long time. In fact, we now
> have a large number of new failures that occurred during the period when
> GCC could not bootstrap on these systems and no easy way of pinpointing
> the cause. These are regressions that will have to be fixed before GCC
> 3.4 is released.
It's certainly good to have fixed the bootstrap.
I think this conversation is redundant in that Richard's new patch
(being tested now) makes sense to me, and will not have the downsides
that I complained about. So, I think we're making progress in a
sensible way.
However, if Richard's patch doesn't work for some reason, I would not
simply revert Geoff's patch. As you say, that would break the
bootstrap. Instead, I would revert both Geoff's patch and the previous
patches that caused the failures; Geoff's patch *itself* introduces
regressions, albeit of a more mild form.
--
Mark Mitchell
CodeSourcery, LLC
mark@codesourcery.com