MinGW is GCC-like and does not support clang-cl, so clean up the lit cfg
to reflect that. Also mark all Windows test cases using clang-cl as
UNSUPPORTED for now. We should probably adapt most of these test cases
to use the GCC-like clang driver, but that will be for later.
Details
Diff Detail
- Repository
- rG LLVM Github Monorepo
Event Timeline
Looks reasonable overall I think, nothing to object to from my point of view, just a few minor comments.. The patch is a bit unwieldy with the number of tests marked as unsupported, but I guess it's as good as it will be. If we're only touching tests that use specifically clang_cl_* to build/run, the mass editing probably is ok.
compiler-rt/test/asan/lit.cfg.py | ||
---|---|---|
109 | Side note; since Clang 16.0.0 (fda44bedd64dbcde1b54432d99ad2f0fca0ad204) there's a new option in the GCC-style driver that does the same as /MD - -fms-runtime-lib=dll. So if/when we can assume Clang >= 16 here, we could simplify this line. | |
150 | The RE match expression gets repeated quite a lot - is there any existing pattern for storing such things in a variable within asan (maybe for tests for other targets)? |
compiler-rt/test/asan/lit.cfg.py | ||
---|---|---|
150 | If you mean only in lit.cfg.py and lit.common.cfg.py, I guess it wouldn't hurt to use a local variable. By the way I am considering changing this check to detect msvc target instead of "not gnu". If you mean for the UNSUPPORTED: tag in tests, I think the intended way to check for target is with the regex judging from 38e9660eaa3478b003cbc48cb4ed0adcf22e053a. |
compiler-rt/test/asan/lit.cfg.py | ||
---|---|---|
150 | Yeah, I meant in the lit python scripts. Changing it to positively check for msvc instead of excluding gnu probably makes the logic clearer, so that sounds good! |
Side note; since Clang 16.0.0 (fda44bedd64dbcde1b54432d99ad2f0fca0ad204) there's a new option in the GCC-style driver that does the same as /MD - -fms-runtime-lib=dll. So if/when we can assume Clang >= 16 here, we could simplify this line.