ChuanqiXu9 wrote: > Looks like this change has some compile-time impact even if modules are not > used: > https://llvm-compile-time-tracker.com/compare.php?from=edc02351dd11cc4a39b7c541b26b71c6f36c8e55&to=7201cae106260aeb3e9bbbb7d5291ff30f05076a&stat=instructions:u > It seems to add about 0.5% to C++ compilations.
Then it looks it relates to https://github.com/llvm/llvm-project/pull/122887/files#diff-e32652ec4ba35b724c68c1c00c7f522d33f0f9beab16ac437140adc0f2843c01R1861-R1874. It is not expected to introduce the regression if modules are not used. I'll revert this and try to send another PR. > > cc @AaronBallman on the process here. I find it a bit concerning that big > changes get landed without approval to make it into the LLVM 20 release. This > is not how things are supposed to work. I am the maintainer for modules&serialization. I think I can land PR directly. And also there are lacking reviewers on the topics of C++20 modules. There are multiple experiences that a patch waiting for approval for months without rare actual reviews. And after I am the maintainer, I'll try to land patches that only affecting C++20 modules if I am confident. And the outcoming looks good for these years. https://github.com/llvm/llvm-project/pull/122887 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
