[PATCH] D33357: Avoid calling report_fatal_error in the destructor of raw_fd_ostream when saving a module timestamp file

[Bruno Cardoso Lopes via Phabricator via cfe-commits]

bruno added a comment.

Any idea why we're hitting this issue in the first place? The error that gets cleaned up is reported at some point before? Seems to me that we're going to fail to update the timestamp but continue as nothing happened, I wonder how many other issues this might trigger...

> I couldn't think of a way to test this, do you think it's possible to have a test for this?

If you can come up with a testcase it would be awesome, but for this to trigger you'd have to reproduce the concurrent scenario in the testcase, which I don't see how.


Repository:
  rL LLVM

https://reviews.llvm.org/D33357