[llvm] MemorySSA patch for 3.9

[Hans Wennborg via llvm-commits]

Sounds good to me.

George, go ahead and commit to the branch, or let me know if you'd let
me to do it.

Thanks,
Hans

On Wed, Jul 20, 2016 at 8:45 PM, Daniel Berlin <dberlin at dberlin.org> wrote:
> Yeah, it was actually an assert i was using to make sure the code wasn't
> doing something broken on "normal" code while debugging the walker (it would
> occasionally infinite loop or do silly things when i first started writing
> it, and this was the easiest way to bugpoint things :P)
> It  should not be there for real, as it is definitely possible to have cases
> where it is valid to visit > 1000 times.
>
>
> On Wed, Jul 20, 2016 at 5:05 PM, George Burgess IV via llvm-commits
> <llvm-commits at lists.llvm.org> wrote:
>>
>> Hi Hans,
>>
>> Can we please apply the attached patch to the 3.9 branch? It fixes
>> PR28632, which is a release blocker.
>>
>> It doesn't apply to trunk, because we entirely replaced the piece of code
>> that this patch "fixes" in r275940.
>>
>> (I realize that fixing an assertion failure by removing the assertion
>> looks kind of bad, but said assertion was really more of a warning that
>> we've hit a pathological-ish case than an error. ;) )
>>
>> Thank you!
>> George