[Bug c/80378] Extend alloc_size attribute for better Linux kernel checking

[amonakov at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80378

Alexander Monakov <amonakov at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |amonakov at gcc dot gnu.org

--- Comment #5 from Alexander Monakov <amonakov at gcc dot gnu.org> ---
Inline wrappers won't work because the second instance of pass_object_sizes
that is supposed to fully fold it is scheduled after IPA passes, and thus
information tied only to the attributes of inlined wrappers is lost.

I think it might be better to introduce __builtin_assume_object_size that can
convey such information directly in the IR.  User code then can do simply:

    #define alloc(a, b) __builtin_assume_object_size(alloc(a, b), a+b)

and GCC itself can properly emit such builtin for existing functions bearing
the attribute, i.e. instead of 

    var = malloc(size);

have in gimple explicitly:

    tmp = malloc(size);
    var = __builtin_assume_object_size(tmp, size);