https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89693
--- Comment #3 from Martin Jambor <jamborm at gcc dot gnu.org> --- True. the node that IPA-CP clones is itself a former thunk, expanded at symbol_table::finalize_compilation_unit time (which I admit is something I did not anticipate), all the thunk information is copied to the clone node, so the clone is now a former node itself too. As a consequence, the first while loop in clone_of_p iterates an extra time and encounters a node which is not thunk.this_adjusting and returns false. Even if it sis not check this_adjusting, it would attempt to find a callee which does not exist and segfault. At the moment I do not have an idea for a quick fix, unfortunately, I will keep trying.
