https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101480
--- Comment #7 from Richard Biener <rguenth at gcc dot gnu.org> --- (In reply to Richard Biener from comment #6) > As Jakub said the behavior is the same for malloc() since years. > ... > So it might work to disable the new/delete/malloc/free optimization when > we see a definition of those functions. But it's surely not enough to disable transforms like x = 0; ... = new ...; if (x != 0) .. to elide the load from x after the 'new'. Like with malloc we assume that state of the runtime (where malloc / new is defined) is only accessible via API calls and not visible to the compiler at the same time its users are.
