ziqingluo-90 wrote: > > The closure runs after the pointer-flow DFS has converged and does not feed > > its own output back in, so a flow edge out of a newly discovered EPL is > > still > > missed. FamilyClosureDoesNotRerunDFS pins that gap. > > I think the alternative is to weave the override-relation into the graph. > Your call that it is level-preserving is correct. Hence, the pointer-level > expansion I did in the DFS blocks you from doing so. The good news is that > the blocker will be gone (see #218209) and the DFS will be simple again. > > After #218209, an edge `p -> q` in the graph simply means that if the EPL`p` > is unsafe, so is `q`. You could encode the override-relation 'x and y are in > the same family' into two graph edges `x -> y` and `y -> x`.
you could re-stack this PR to follow #218209. https://github.com/llvm/llvm-project/pull/213319 _______________________________________________ llvm-branch-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
