https://gcc.gnu.org/bugzilla/show_bug.cgi?id=26388
--- Comment #19 from Martin Sebor <msebor at gcc dot gnu.org> --- You're right, I had overlooked the "provided by the implementation" part. I would not want implementations to have the latitude to omit explicit calls to operator new() unless the operator were the default. That would largely defeat the purpose of the operators' replaceabilty: to be able to control the dynamic allocation policy of a program. But if there were a way to detect whether or not the operator has been replaced (it seems there should be some trick) then dynamic memory allocated at local scope and known not to escape could be replaced with stack space without breaking anything. I think that would be a useful optimization.
