[clang-include-fixer] Skip .rc files when finding symbols

[rnk]

Summary

For some reason CMake includes entries for .rc files, but
find-all-symbols handles them improperly.

See PR43993

[sammccall]

This seems fine as a workaround, but will obviously only affect include-fixer. I'm wary of embedding it deeper without understanding the problem though.

Can you paste the *.rc entry from compile_commands.json?
I'm curious how CMake wants to send it to clang, and why it works with clang but not other tools.

[rnk]

This is a single .rc entry:

{
  "directory": "C:/src/llvm-project/build",
  "command": "C:\\PROGRA~2\\WI3CF2~1\\10\\bin\\100183~1.0\\x64\\rc.exe -DGTEST_HAS_RTTI=0 -DUNICODE -D_CRT_NONSTDC_NO_DEPRECATE -D_CRT_NONSTDC_NO_WARNINGS -D_CRT_SECURE_NO_
DEPRECATE -D_CRT_SECURE_NO_WARNINGS -D_HAS_EXCEPTIONS=0 -D_SCL_SECURE_NO_DEPRECATE -D_SCL_SECURE_NO_WARNINGS -D_UNICODE -D__STDC_CONSTANT_MACROS -D__STDC_FORMAT_MACROS -D__
STDC_LIMIT_MACROS -DRC_FILE_VERSION=\\\"10.0.0git\\\" -DRC_INTERNAL_NAME=\\\"llvm-tblgen\\\" -DRC_PRODUCT_NAME=\\\"LLVM\\\" -DRC_PRODUCT_VERSION=\\\"10.0.0git\\\" -DRC_VERS
ION_FIELD_1=10 -DRC_VERSION_FIELD_2=0 -DRC_VERSION_FIELD_3=0 -DRC_VERSION_FIELD_4=0 -IC:\\src\\llvm-project\\build\\utils\\TableGen -IC:\\src\\llvm-project\\llvm\\utils\\Ta
bleGen -IC:\\src\\llvm-project\\build\\include -IC:\\src\\llvm-project\\llvm\\include  /DWIN32   -UNDEBUG /nologo /foutils\\TableGen\\CMakeFiles\\llvm-tblgen.dir\\__\\__\\r
esources\\windows_version_resource.rc.res C:\\src\\llvm-project\\llvm\\resources\\windows_version_resource.rc",
  "file": "C:/src/llvm-project/llvm/resources/windows_version_resource.rc"
},

There is one for every binary. I think .rc files can include headers to get things like struct sizes and macros. My understanding is that it's an old fork of the Visual C compiler, so it is in some sense a "compilation". I suppose if you wanted to write a global renaming tool for a macro, you'd need to know about uses in .rc files.

[kimgr]

I wonder if it would be better to look at the compile-command than the source file?

It's not unthinkable that someone would use another extension for an rc file, or use a .rc extension for a C or C++ source file. But if the compiler is rc.exe, the source file must be in rc format and should be ignored, right?

[rnk]

We already have llvm-rc.exe which one could plausibly use to compile rc files, and I didn't like binary.lower().endswith('rc.exe') as a test.

[kimgr]

Oh, I see. I thought rc.exe was the one and only implementation. Fair enough, this is probably the lesser of two evils :)

clang-tools-extra/clang-include-fixer/find-all-symbols/tool/run-find-all-symbols.py

else:
     build_path = find_compilation_database(db_path)
 
   tmpdir = tempfile.mkdtemp()
 
   # Load the database and extract all files.
   database = json.load(open(os.path.join(build_path, db_path)))
   files = [entry['file'] for entry in database]
 
+  # Filter out .rc files on Windows. CMake includes them for some reason.
+  files = [f for f in files if not f.endswith('.rc')]
+
   max_task = args.j
   if max_task == 0:
     max_task = multiprocessing.cpu_count()
 
   try:
     # Spin up a bunch of tidy-launching threads.
     queue = Queue.Queue(max_task)
     for _ in range(max_task):