Re: gcc-4.4 ICE
On Sun, 24 Oct 2010, Thorsten Glaser wrote:
> Andreas Schwab dixit:
>
> >Thorsten Glaser <tg@mirbsd.de> writes:
> >
> >> jdsample.c:140: internal compiler error: in
> >> reload_cse_simplify_operands, at postreload.c:396
> >
> >http://gcc.gnu.org/PR37053
>
> Thanks, that’s fast. I’m glad so many people who work in, for example,
> gcc, participate on this list. So, Comment 19 backport?
I tested it, and yes, the PR37053 patch is sufficient to fix this. Maxim
Kuvyrkov regression tested it:
http://gcc.gnu.org/ml/gcc-patches/2009-08/msg00375.html
Despite that, there may be complications (?)
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41064#c8
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40414#c15
It would appear that we could safely apply the PR37053 fix for m68k only,
but if it is applied unconditionally then we may see a regression on cris,
unless we also cherry-pick the fix from PR41064 ... but I could be
completely wrong, as I'm having a hard time figuring out what is going on
in these bug reports and whether perhaps they are not related at all!
The PR37053 fix also fixed PR40414 in my test, which contradicts the
contradiction in PR41064:
> Hans-Peter Nilsson 2010-05-11 17:06:25 UTC (In reply to comment #6)
> > Causes PR40414 on GCC 4.4.x, reopening to make the other PR depend on
> > this.
>
> Looks like the above statement is wrong: the committed patch for this PR
> is just (one of the two) required for PR40414. I'm closing this as
> there's no reason to hold *this* PR open regardless of the state of
> PR40414.
Since this makes no sense to me, I'd ask Matthias to apply the fix but
keep it to m68k only (unless someone can run regression tests on other
architectures...)
Finn
Reply to: