This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] proposed fix for PR40987 if-conversion bug on HWI32 hosts
- From: Mikael Pettersson <mikpe at it dot uu dot se>
- To: Eric Botcazou <ebotcazou at adacore dot com>
- Cc: Mikael Pettersson <mikpe at it dot uu dot se>, gcc-patches at gcc dot gnu dot org
- Date: Mon, 5 Oct 2009 18:11:54 +0200
- Subject: Re: [PATCH] proposed fix for PR40987 if-conversion bug on HWI32 hosts
- References: <19130.34705.539856.885099@pilspetsen.it.uu.se> <200909260757.42681.ebotcazou@adacore.com> <19133.62834.70059.548359@pilspetsen.it.uu.se> <200909261318.16262.ebotcazou@adacore.com>
Eric Botcazou writes:
> > It's not tight enough, otherwise it would have prevented the bug
> > in the PR because this test dominates the actual transformation
> > parts of this function.
>
> I meant, is it tight enough for the problem it is supposed to avoid? Maybe
> the problem for PR40987 is different.
I think that test is Ok for the thing it's testing, except for
the signed overflow I think it can trigger. But PR40987 is not
caused by this one test being wrong, ALL of the code in there
is fundamentally unaware of larger-than-HWI constants.
I don't have time to do a detailed analysis of that code right now,
but I'll keep the brute-force fix locally as a reminder to revisit
it later.