This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
RE: [Patch ARM] Fix PR45070 - Miscompiled c++ class with packed attribute on ARM with -Os optimizations
- From: "Joseph S. Myers" <joseph at codesourcery dot com>
- To: Ian Bolton <Ian dot Bolton at arm dot com>
- Cc: gcc-patches at gcc dot gnu dot org, 'Mikael Pettersson' <mikpe at it dot uu dot se>
- Date: Wed, 11 Aug 2010 16:34:01 +0000 (UTC)
- Subject: RE: [Patch ARM] Fix PR45070 - Miscompiled c++ class with packed attribute on ARM with -Os optimizations
- References: <000001cb3951$455dcae0$d01960a0$@Bolton@arm.com> <19554.40929.481335.91702@pilspetsen.it.uu.se> <000401cb396d$e2510490$a6f30db0$@Bolton@arm.com>
On Wed, 11 Aug 2010, Ian Bolton wrote:
> Mikael Pettersson writes:
> > The test case options appear to be incorrect.
> >
> > 1. You'd end up trying to execute cortex-a8 specific object
> > code on non cortex-a8 compatible CPUs.
> > 2. The bug and the test case are not specific to cortex-a8.
> > In the PR trail you see at least two people confirming the
> > bug on plain old armv5te.
> >
> > The test shouldn't need any option except "-Os".
> >
>
> Thanks for spotting this. I have fixed up the patch and
> reattached. OK for trunk?
I don't see anything ARM-specific about this testcase at all. Why isn't
it going in gcc.c-torture/execute, with no special dg-* markings, where it
will be run for all targets with a range of options that includes -Os?
gcc.target is only for tests that are target-specific in some way or where
special target-specific options are needed to exercise the bug in
question, not for target-independent tests that show the bug with options
that are among those c-torture uses.
--
Joseph S. Myers
joseph@codesourcery.com