[libcxx-commits] [PATCH] D108080: [libcxx] simplifies nodiscard by permanently enabling the attribute...

[Arthur O'Dwyer via Phabricator via libcxx-commits]

Quuxplusone added a comment.

> Since I'm in no hurry to merge this patch, how about I test it on various Google codebases and see what breaks? That's at least half a billion lines of code (probably more) that we can get data from. Doing that takes a bit of time, so don't expect an answer before the end of the week. Since I'm heavily biased against anything other than _LIBCPP_NODISCARD, I don't think I should be the one to set what we should consider statistically significant, but I'm going to offer 0.001% of the codebase as a starting threshold

If you find zero false positives (that is, if you find nothing; or if you find only things that I agree are bugs) then personally I'll accept the sample size of "2 codebases, 0 harm" as good enough for the "just ship it and wait for customer complaints" strategy.  However, if you find //something// — that is, if applying this patch //does// break Google's build — then personally I'd consider that evidence in favor of the more cautious strategy. I don't think there's any need to do math here.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D108080/new/

https://reviews.llvm.org/D108080