https://gcc.gnu.org/bugzilla/show_bug.cgi?id=128551
--- Comment #5 from Drea Pinski <pinskia at gcc dot gnu.org> --- Note I think it was working by accident in GCC 4.9 and before. Chosing the aliasing info for the load from int64_t . But then again it is hard to see if the issue is coming via the path which PR 125683 fixed or the path which is still broken. Once I debug this further I can see when is the regression really being caused. Note also until GCC 15 (including 15), pre/VN-hoisting would use the correct aliasing which is why `-fno-tree-pre -fno-code-hoisting` is needed there too.
