nikic wrote: > Thanks! Would you mind splitting those changes off into a separate PR? They > look good to me.
Hm, I think this issue is not fully fixed yet. I *was* able to get a dylib compile-time-tracker run out of this (see https://llvm-compile-time-tracker.com/compare.php?from=798856a172f89ffdab47b14089cc34bab7bcc767&to=e019b5ffc43c12735e15ed36559022ea7672ff24&stat=wall-time for results, this has about 3-4% wall-time increase at `-O0` with CIR built but not used.) But when trying to reproduce something closer to our production configuration I'm still seeing issues. This is the same cmake as previously, but using `-DLLVM_ENABLE_PROJECTS="clang;mlir"`, so it will link against libMLIR.so instead of including MLIR subsets in libclang-cpp.so. (I'm also using `-DLLVM_ENABLE_ASSERTIONS=false` now, but I don't believe that should matter.) With this I get the following errors when running clang on a dummy test.c with `int main() { return 0; }`: https://gist.github.com/nikic/70b7755db938a9c56bb8c24a0a4e1dcb https://github.com/llvm/llvm-project/pull/221291 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
