https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89693
--- Comment #4 from Jan Hubicka <hubicka at gcc dot gnu.org> --- The verifier is checking that all redirections look sane (as one made by our IPA optimizers). As number of IPA tranforms grows this gets harder to track. Perhaps we can just drop the check now. I do not remember it noticing real bug recently (it was useful to check somewhat subtle way tree-inlie updates callgraph edges after cloning or inlining, but that code should be pretty stable now). Alternatively we can of course add a flag "redirected" and disable all the checks for edges which we know that was updated (as we already do for icf_merged), but I am not sure this is worth the effort.
