[PATCH] D53729: [llvm-cov] Don't remap existing paths

Max Moroz via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Fri Oct 26 12:41:05 PDT 2018


Dor1s added a comment.

In https://reviews.llvm.org/D53729#1277534, @vsk wrote:

> Even if there's a file at the path we're supposed to remap, shouldn't we still do the remap? Couldn't the file we're intended to read be at the remapped path?


Yeah, I'm on the same side as @vsk here. There might be different workflows, but the most obvious behavior of path remapping would be to always remap paths. It would be nice to maintain it rather than change.

I've also seen cases where binaries compile system headers (e.g. from `/usr/include`) with coverage instrumentation, and if you need to generate a coverage report for those on another machine, you'd better copy all of them rather than hope to have the same headers. For instance, i've seen this problem with wireshark and irssi projects: https://github.com/google/oss-fuzz/issues/1743#issuecomment-416265306


Repository:
  rL LLVM

https://reviews.llvm.org/D53729





More information about the llvm-commits mailing list