https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116219
--- Comment #13 from Richard Biener <rguenth at gcc dot gnu.org> --- I think we do not want any of those fancy names to leak into debug info - those are solely to make dumps easier to follow. Iff we were able to express "this entity was created when vectorizing entity X" (with X not really easy to pin-point) one could use an abstract origin to said entity (but abstract origins are not really designed for this). create_tmp_var_raw already marks those vars as DECL_NAMELESS so they do not show up in final-insns dumps. Having the UID when those variables related to other entities not having a name is very useful - stripping that would be bad. How's exactly code/debug-info differing? Or is it just the dumps that compare-debug compares differing? Can't we avoid dumping DECL_NAMELESS stuff in more places to make that diffing happy?
