MaskRay added a comment. In D149162#4517500 <https://reviews.llvm.org/D149162#4517500>, @MaskRay wrote:
> `registerTargetGlobalVariable` relies on the iteration order of StringMap, > which is not guaranteed. The bug is caught by D155789 > <https://reviews.llvm.org/D155789> > > curl -L 'https://reviews.llvm.org/D155789?download=1' | patch -p1 > cmake ... -DLLVM_ENABLE_REVERSE_ITERATION=on > ninja -C ... check-llvm-unit # `LLVM-Unit :: > Frontend/./LLVMFrontendTests/OpenMPIRBuilderTest/registerTargetGlobalVariable` > fails Looks like the issue might be introduced by 03f270c900e1f8563419fdd302683a9503e98722 in the iteration order of `OffloadEntriesDeviceGlobalVarTy OffloadEntriesDeviceGlobalVar` (a StringMap). Unfortunately, changing it to `MapVector<StringRef, OffloadEntryInfoDeviceGlobalVar>` will break other tests like `clang/test/OpenMP/declare_target_codegen.cpp`, which suggests that there are other weird iteration order dependent behavior. Hopefully someone familiar with OpenMP can investigate :) Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D149162/new/ https://reviews.llvm.org/D149162 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
