This reduced the download size for bug 1328454 by an order of magnitude.

Comment on attachment 8824528 [details] [diff] [review] Only package the clang-tidy binaries for the clang-tidy builds Review of attachment 8824528 [details] [diff] [review]: ----------------------------------------------------------------- ::: build/build-clang/build-clang.py @@ +133,5 @@ > + shutil.rmtree(path) > + # exists() returns false for broken symbolic links, so we need to allow > + # them explicitly. > + elif os.path.exists(path) or os.path.islink(path): > + os.unlink(path) I'm pretty sure that the "pythonic" way to write this would be something like: try: os.unlink(path) except: pass @@ +290,5 @@ > except which.WhichError: > raise ValueError("%s not found on PATH" % f) > > + > +def prune_final_dir_for_clang_tidy(final_dir): Can we have a doc comment on this function? I understand the goals of this function but I imagine that just the name may not be enough to make it clear in the future. @@ +591,5 @@ > llvm_source_dir, stage3_dir, build_libcxx, osx_cross_compile, > build_type, assertions, python_path, gcc_dir) > > + package_name = "clang" > + if import_clang_tidy: I feel like "import_clang_tidy" is a poor name now that the artifacts which are produced don't actually contain "clang + mozilla clang tidy extensions" but contain _only clang tidy_. Perhaps build_clang_tidy would be better?

Pushed by eakhgari@mozilla.com: https://hg.mozilla.org/integration/mozilla-inbound/rev/c05634d71a73 Only package the clang-tidy binaries for the clang-tidy builds; r=mystor