ymandel requested changes to this revision. ymandel added a comment. This revision now requires changes to proceed.
It turns out that this change wasn't necessary. I've deleted the relevant FIXME in https://reviews.llvm.org/rGabc16c7a5b0a63d14172262153608b3d24de957f. AFAICT, the reason that the new test `JoinBackedge` fails at HEAD is because we discard the results of analyzing the function body, so the update `Foo = false` is simply ignored. This hypothesis rests on observing that each block is visited only once during analysis of the test. It's also consistent with the known (broken) behavior of our comparison operator. Interestingly, this is a nicely stark illustration of the unsoundness of our current approach. :( So, this test will be useful in the future when add proper widening. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D130270/new/ https://reviews.llvm.org/D130270 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
