https://gcc.gnu.org/bugzilla/show_bug.cgi?id=117809
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Ever confirmed|0 |1
Keywords| |missed-optimization
Last reconfirmed| |2024-11-28
Status|UNCONFIRMED |NEW
--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
It's difficult for GCC to make use of this case optimally, but sure. On
allocation failure we would need to document that *ptr either becomes
undefined or NULL. We currently handle posix_memalign "by name" only.
Likewise realloc() is not 'malloc' but special-cased by name.
Note we unfortunately overloaded the malloc attribute when it has an argument
to instead specify a corresponding deallocation function but otherwise not
have the same semantic as 'malloc' without argument.
That means 'malloc' semantics on a positional argument will have confusing
implications.
So I'd love to see a proposal, possibly also addressing the realloc case
where the pointer is either the "old" or a new one not aliasing any other
pointer.
