PR 8146

[Eric Botcazou]

> But in my experience, HJ's patch fixes PR 7591 and 2002307-1.c on the
> branch for -march=i686 (and it looks like it fixes PR 8404 too). At least 2
> "distros" I know of are using HJ's patch (Gentoo and LinuxFromScratch) so
> some wider testing has been achieved. And I can confirm no testsuite
> regressions.

PR opt/7591 (which has nothing to do with optimization btw) is a regression 
from gcc 3.0.x, so should really be marked as high-priority and taken care of 
on the branch. It's a i686 specific problem (doesn't occur on i586).

> None of this has likely anything to do with subject PR 8146 but it would
> be nice if HJ's patch was applied on the branch to stop the blatant
> miscompiles.

PR 7591 appears not to be related to PR 8146, since the mainline (on which it 
is fixed) can't bootstrap gcc 2.95.3 either. And I just verified that the 
mis-compiled stage2/f/target.o is identical for branch and mainline.

On the other hand, this patch:

2002-08-17  H.J. Lu <hjl@gnu.org>

	* calls.c (PUSH_ARGS_REVERSED): Define only if not defined.
	* expr.c (PUSH_ARGS_REVERSED): Likewise.

2002-07-26  H.J. Lu <hjl@gnu.org>

	* config/i386/i386.h (PUSH_ARGS_REVERSED): Set to 1.

fixes PR 8146, both on the trunk and the branch.

-- 
Eric Botcazou