http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54540
Steven Bosscher <steven at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |steven at gcc dot gnu.org
--- Comment #7 from Steven Bosscher <steven at gcc dot gnu.org> 2012-11-20
11:09:12 UTC ---
(In reply to comment #6)
> (In reply to comment #5)
> > Can this be closed now?
>
> Well the comment 4 is still relevant, I suspect that there are still latent
> issues in postreload.
Are we going to keep a bug report open because of a _suspected_ bug, with
no test case to reproduce the problem with the current trunk? Any new bug
report without a test case would be almost immediately closed as INVALID,
and I see no good reason to do things differently for an existing PR with
a fix applied and no test case remaining to reproduce the issue.
Have you at least looked at postreload (using a compiler without your fix
for reload) to see where things go bad? I suspect the bad insn is created
in reload_combine_recognize_pattern. You could try to rule out move2add
using debug counters (-fdbg-cnt=cse2_move2add:0) at least.