aaron.ballman added a comment. In D101790#2748474 <https://reviews.llvm.org/D101790#2748474>, @NoQ wrote:
> I completely agree with that as well, these are 100% true positives, which is > why I left them as fixmes. The reason why I think we can't have them is that > we don't have an appropriate alias analysis implemented. Namely, we have two > analyses, one inside the checker that leaves out false positives like the one > in `negative_by_val()`, the other is the universal/reusable analysis in Utils > that i'm patching up that doesn't support happens-before relation. > > I believe the right way forward is to unify these analysis by making the > analysis in Utils account for happens-before relation. Ideally this means > re-implementing it over the CFG which probably isn't even that hard, it's > just a simple graph traversal problem. I may look into it in June or so. I think that sounds like the right way forward. Thank you for looking into the unification! CHANGES SINCE LAST ACTION https://reviews.llvm.org/D101790/new/ https://reviews.llvm.org/D101790 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
