zequanwu added a comment. In D138846#4657195 <https://reviews.llvm.org/D138846#4657195>, @zequanwu wrote:
> In D138846#4657194 <https://reviews.llvm.org/D138846#4657194>, @alanphipps > wrote: > >> In D138846#4657193 <https://reviews.llvm.org/D138846#4657193>, @zequanwu >> wrote: >> >>> In D138846#4657175 <https://reviews.llvm.org/D138846#4657175>, @hans wrote: >>> >>>> We're seeing crashes in `initializeValueProfRuntimeRecord` that bisects to >>>> this commit. I think Zequan is investigating: >>>> https://bugs.chromium.org/p/chromium/issues/detail?id=1503919 >>> >>> It turns out to be that our rust code with old version of rustc without >>> this change, so mixture of raw profile versions are used, causing segment >>> fault. >> >> Thank you for the update! Is this also the case of the issue @glandium >> reports above as well? I think both issues manifested as a ValueProf issue. > > I don't know, but I'd suggest to check if all sources compiled with profile > runtime are using the llvm version that contains this change. Otherwise, > mixed raw profiles versions will cause unexpected behaviour. I just saw @glandium's earlier comment: > Code built with older versions of LLVM (e.g. rust) and running with the > updated runtime crash at startup with this change. This is the exact same issue we encountered. Because there is a profile format change, it's expected to update both clang and rustc to use the same version of llvm in order for it to work. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D138846/new/ https://reviews.llvm.org/D138846 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
