llvm dev - Nov 2018 - [MergeFuncs] Help request for broken test

Hey list,

I've made a small change to some enum values
in include/llvm/IR/Instruction.def and it seems to be breaking a fragile
test of the MergeFunctions pass. This seemingly small change is exaggerated
by the hashing function used by MergeFunctions. Since the hashes are used
to sort internal MergeFunctions data structures, the output of this pass is
different enough to break the test.

Is there an expert in the MergeFunctions pass that could help me sort this
out? I've attached a small patch to elucidate the problem. Also, I'd be
happy to share more of my findings if desired.

Thanks in advance,
Cameron
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20181108/5f0a694e/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: merge.patch
Type: application/octet-stream
Size: 2238 bytes
Desc: not available
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20181108/5f0a694e/attachment.obj>

On Thu, Nov 8, 2018 at 2:53 PM Cameron McInally <cameron.mcinally at
nyu.edu>
wrote:
> Hey list,
>
> I've made a small change to some enum values
> in include/llvm/IR/Instruction.def and it seems to be breaking a fragile
> test of the MergeFunctions pass. This seemingly small change is exaggerated
> by the hashing function used by MergeFunctions. Since the hashes are used
> to sort internal MergeFunctions data structures, the output of this pass is
> different enough to break the test.
>
A little digging found that the test in question, namely
test/Transforms/MergeFunc/call-and-invoke-with-ranges.ll, was modified the
last time the enum values in include/llvm/IR/Instruction.def were changed.
This was revision r255522.

It seems that it would be safe to make a similar change for my work on D53877.
Are there any objections to this?
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20181108/b7ab8a6d/attachment.html>