rnk wrote: @llvm/clang-vendors -- We're looking for additional input on the best way to customize the clang crash diagnostic. The question to downstream vendors is, is this interface (a cmake defined macro) sufficient flexibility, or is there a better way to enable this customization, or are our needs so different that we should all manage our own local downstream patches?
In particular, at Google, where we make heavy use of clang header files, collecting a single crash reproducer, even with PCM files, is not a good use of customer time. When we need to reduce a compiler crash, typically our first step is to rule out modules by replaying the build with modules disabled, which is doable, since we use clang header modules, not C++20 modules. If we can get single-file pre-processed source, that is preferred, since cvise works on it, and will go faster than it does working over a multi-module reproduction script. CVise does have support for module reducers thanks to @emaxx-google 's efforts, but even still, ruling out modules is generally a good first step. https://github.com/llvm/llvm-project/pull/157536 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
