[Bug rtl-optimization/57425] [4.8/4.9 regression] RTL alias analysis unprepared to handle stack slot sharing

[amylaar at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=57425

--- Comment #7 from Jorn Wolfgang Rennecke <amylaar at gcc dot gnu.org> ---
Created attachment 30310
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=30310&action=edit
patch for alias problem during cse pass

(In reply to Richard Biener from comment #4)
> Note that I only recently fixed a wrong alias query from using
> true_dependence
> to anti_dependence.  Similar issues may exist elsewhere.

Indeed, looking at the .172r.cse1 dump more closely, I saw it
actually contained actually two dumps, first looking OK, the
second with the deleterious changes in place.

First, cse is applied to the asms.  I'm not sure if this should
be called a bug or a feature.
At any rate, cse does have an alias problem as you suspected.
The attached patch addresses this issue, but doesn't fix the testcase;
something goes again wrong between the reload and postreload dump.