mstorsjo wrote: > This seems to have had an unexpected effect. In a build where I don't use the > new path style, I used to get the old path style returned like this: > > ``` > $ clang -target x86_64-w64-mingw32 -print-runtime-dir > /home/martin/clang-nightly/lib/clang/19/lib/windows > ``` > > However after this change, now I'm getting the new style path, even if it > doesn't exist, and if the old one actually did exist: > > ``` > $ clang -target x86_64-w64-mingw32 -print-runtime-dir > /home/martin/clang-nignhtly/lib/clang/19/lib/x86_64-w64-windows-gnu > ``` > > I'm ok with changing the default if the old path style doesn't exist - but if > it does exist, we should still return that. (I haven't dig into it to see why > this is, yet.)
The reason for this seems to lie here: https://github.com/llvm/llvm-project/blob/ccdebbae4d77d3efc236af92c22941de5d437e01/clang/lib/Driver/Driver.cpp#L2213-L2219 Previously, this picked the `TC.getCompilerRTPath()` case, while this now always ends up going with `TC.getRuntimePath()` as it's never empty. https://github.com/llvm/llvm-project/pull/87866 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits