[PATCH] D49297: expose debugify as an LLVM option in clang

[Greg Bedwell via Phabricator via llvm-commits]

gbedwell added a comment.

In https://reviews.llvm.org/D49297#1165863, @vsk wrote:

> First off, @tyb0807, I really appreciate the work you've been doing on LLVM.
>
> With that said, I think this is trying to solve the wrong problem, and I don't think it makes sense to proceed.


That's fair, although I somewhat disagree and I'll say that for our workflow it's a bit more than marginal in terms of how much easier it is, but one thing I'd really like to get to the bottom of before rejecting this approach entirely are the following failures above:

  445 CheckFunctionDebugify [Merge disjoint stack slots]: FAIL
  727 CheckFunctionDebugify [CodeGen Prepare]: FAIL

I've had a background process running on my PC hitting some preprocessed game code files pretty hard for a couple of weeks now.  It's produced small bug reports for a number of passes, but the above failures don't get spotted with the clang piped into opt approach.  I've unfortunately been temporarily waylaid by another project for the last couple of months or so (otherwise I'd already be contributing fixes for the bugs I've raised) so I've not had a chance to investigate whether these are genuine failures or false positives.  The big question for me, is does this increase the potential coverage?


Repository:
  rL LLVM

https://reviews.llvm.org/D49297