MaskRay wrote: > > Having a `-Wdeprecated-include-gch` looks good to me. > > But do we intend to remove it later? That will break the drop-in ability of > Clang, and probably a lot of people's builds. > > > The inconvenience probing .gch happens when GCC and Clang .pch files are > > mismatched > > To solve this, couldn't the driver peek at the magic at the start of the file > or something like that?
This is feasible but not elegant, and likely unnecessary. I have tried many keywords to figure out `-o .*\.gch` users and the number is likely extremely small. Among them, there are many that place the `-include` header and `.gch` file in the same directory, and these users won't be affected at all (a lot of distributions do not use a separate build directory) https://sourcegraph.com/search?q=context:global+-o%5B+%5D.*%5C.gch&patternType=regexp&sm=1&groupBy=repo Many modern build systems (Bazel, Buck) require the full set of output files. As an opt-in feature, PCH is typically not modeled at all, at least for the majority of packages. https://github.com/llvm/llvm-project/pull/67084 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
