On Wed, Nov 06, 2024 at 01:08:11PM +0100, Jakub Jelinek wrote: > Though, unsure how that > https://eel.is/c++draft/expr.new#14 > interacts with > https://eel.is/c++draft/expr.new#8 > and we'd have to check if we do the size checking before the > ::operator new/new[] calls or it ::operator new just throws/returns NULL.
I've looked at what code we generate and I guess the -fexceptions case is fine, ::operator new[] isn't called if there is overflow, __cxa_throw_bad_array_new_length is called instead. But for -fno-exceptions it is more problematic, the emitted code then uses ::operator new but passes ~size_t(0) to it as length and expects the operator to return NULL (of course, -fno-exceptions is outside of the standard, but by calling the ::operator new which can be optimized away it would no longer guarantee returning NULL). For new (std::nothrow) int[s] with -fexceptions it is even worse, https://eel.is/c++draft/expr.new#8.6 says it shouldn't call ::operator new[] and instead result in nullptr, but the code calls __cxa_throw_bad_array_new_length which throws an exception. Jakub
