https://gcc.gnu.org/bugzilla/show_bug.cgi?id=119614
--- Comment #33 from Martin Jambor <jamborm at gcc dot gnu.org> --- The patch from comment #32 passes LTO-bootstrap and profiled-LTO-bootstrap on x86_64-linux. I have asked Honza to look at it and comment, especially on the decision to put the return VR into clone_info despite that these often do not describe a clone.