Hello Antonio and Mikael. I've tried reproducing this also after removal of the redundant test (D28943), and I still see the fail for CHECK10.
The problem is that we (both me an Mikael) are running this test from inside our llvm repo (with build&test artifacts ending up in a "build-all" directory inside the llvm clone). So if I run the test with -debug I can see that it is searching for the format file like this: Trying /repo/llvm/build-all/tools/clang/test/Format/Output/.clang-format... Trying /repo/llvm/build-all/tools/clang/test/Format/Output/_clang-format... Trying /repo/llvm/build-all/tools/clang/test/Format/.clang-format... Trying /repo/llvm/build-all/tools/clang/test/Format/_clang-format... Trying /repo/llvm/build-all/tools/clang/test/.clang-format... Trying /repo/llvm/build-all/tools/clang/test/_clang-format... Trying /repo/llvm/build-all/tools/clang/.clang-format... Trying /repo/llvm/build-all/tools/clang/_clang-format... Trying /repo/llvm/build-all/tools/.clang-format... Trying /repo/llvm/build-all/tools/_clang-format... Trying /repo/llvm/build-all/.clang-format... Trying /repo/llvm/build-all/_clang-format... Trying /repo/llvm/.clang-format... Using configuration file /repo/llvm/.clang-format So the test case will find the format file that exists in the llvm repository and it uses that one (and not the one defined as a fallback). I'm not sure that there is a rule against having the build&test artifacts inside the repo. I think there are both pros and cons to have such a setup. And it is something that has been working well for us so far. Regards /Björn
_______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
