https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94734
--- Comment #9 from Richard Biener <rguenth at gcc dot gnu.org> --- (In reply to Richard Biener from comment #8) > In particular tree_could_trap_p woudl consider the load possibly trapping > due to the variable indexing but the patch seems to override that which > I agree is bogus. I think we need to revert it and re-implement for GCC 11. > All testcases contain a dominating load in the condition and what we _can_ > do is treat loads of stack variables the same as stores as far as > non-trappingness is concerned. Actually the non-trapping machinery only considers plain SSA name dereferences so there's much more to extend. Definitely nothing for GCC 10.
