https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125857
--- Comment #7 from Andrew Macleod <amacleod at redhat dot com> --- Created attachment 64781 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=64781&action=edit possible fix WE can bypass the issue for now with the attached patch. It looks at the value returned from ipa_return_value_range, and if it is not gimple_min_invariant in this CU, it clears the points to field. Do you want me to test and commit this until someone in IPA land decides what to do here? That seems reasonable to me.
