[PATCH] D49890: Clang-Tidy Export Problem

[Ng Zhi An via Phabricator via cfe-commits]

ngzhian added a comment.

I found this patch while searching for a similar problem I faced while tryin to use clang-tidy to fix some warnings in V8 (Chrome's JavaScript engine).
The setup for our compilation is as such:

- basedir
  - src/
  - out/ -arch1/
      - .o files, generated files
      - compile_commands.json
    - arch2/

We use ninja to build, so usually with a command like `ninja -C out/arch1`, this means (roughly) cd into out/arch1 and build.

Given this, I have an alternative suggestion for a fix. In run-clang-tidy.py apply_fixes, we call the clang-apply-replacement with cwd set to build_path, like so:

  diff --git a/clang-tools-extra/clang-tidy/tool/run-clang-tidy.py b/clang-tools-extra/clang-tidy/tool/run-clang-tidy.py
  index 1eb13529575..cde9e3671af 100755
  --- a/clang-tools-extra/clang-tidy/tool/run-clang-tidy.py
  +++ b/clang-tools-extra/clang-tidy/tool/run-clang-tidy.py
  @@ -150,7 +150,7 @@ def apply_fixes(args, tmpdir):
     if args.style:
       invocation.append('-style=' + args.style)
     invocation.append(tmpdir)
  -  subprocess.call(invocation)
  +  subprocess.call(invocation, cwd=args.build_path)

def run_tidy(args, tmpdir, build_path, queue, lock, failed_files):

I don't know the internals of clang-tidy enough to do this, but it works for me. And it seems reasonable, since the Replacement FilePath in the yaml file consumed by clang-apply-replacement was generated as a path relative to the build_path anyway. For now this is a sufficient hack for my use case (tried with with just 1 data point as a test), and it will be good to get feedback if this fix is correct. Thanks!


Repository:
  rCTE Clang Tools Extra

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D49890/new/

https://reviews.llvm.org/D49890