[PATCH] D40790: DAGCombiner bugfix in MergeStoresOfConstantsOrVecElts()

[Sanjay Patel via Phabricator via llvm-commits]

spatel added a comment.

In https://reviews.llvm.org/D40790#946382, @jonpa wrote:

> In https://reviews.llvm.org/D40790#945008, @spatel wrote:
>
> > In https://reviews.llvm.org/D40790#944803, @jonpa wrote:
> >
> > > > Did this bug start after r319036?
> > >
> > > This problem disappeared at least on the test case when I reverted that patch, so yes, this seems to be the cause (It also explains why I all of a sudden got some 10 csmith wrong-code results relating to this recently...)
> >
> >
> > Sorry I hadn't seen it until now, but https://reviews.llvm.org/D40701 is a superset of this fix. I think we can abandon this (though we might want to include a SystemZ test for extra coverage).
>
>
> I don't see this fix yet in the tree... What is holding it back?


It was approved and committed, but there were other failures, so it was reverted and being investigated.

> OK, I trust you then to commit the other fix and if you want also include the test case here for SystemZ.

Not sure if this was to me. I'm not the author of the patch; @niravd is working on it from what I can see (and should see this discussion too).


https://reviews.llvm.org/D40790