On Wed, Oct 05, 2022 at 12:06:10 +0000, nia wrote: > On Tue, Oct 04, 2022 at 12:21:08AM +0200, Joerg Sonnenberger wrote: > > On Mon, Oct 03, 2022 at 12:31:42PM +0000, nia wrote: > > > I'd argue that providing alloca(3) as anything except a compiler > > > builtin is a bug, and that kind of thing should never be used. > > > > Well, if you are in strict C/C++ standard mode, alloca should not exist > > because it is not part of the standard. If you are in GNU-extended mode, > > the builtin logic for it will trigger. Not leaking extra macros has its > > advantage. > > Would you agree that we shouldn't have a definition for alloca in our > headers since it should be handled by the compiler, then?
The compiler needs to see the declaration for the builtin handling to kick in properly. Without one you will get: warning: implicit declaration of function 'alloca' [-Wimplicit-function-declaration] warning: incompatible implicit declaration of built-in function 'alloca' [-Wbuiltin-declaration-mismatch] with a wrong one you will get: warning: mismatch in return type of built-in function 'alloca'; expected 'void *' [-Wbuiltin-declaration-mismatch] or warning: conflicting types for built-in function 'alloca'; expected 'void *(long unsigned int)' [-Wbuiltin-declaration-mismatch] -uwe
