[Bug c/104633] New: [12 Regression] -Winfinite-recursion diagnoses fortify wrappers

["rguenth at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104633

            Bug ID: 104633
           Summary: [12 Regression] -Winfinite-recursion diagnoses fortify
                    wrappers
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
          Assignee: unassigned at gcc dot gnu.org
          Reporter: rguenth at gcc dot gnu.org
  Target Milestone: ---

extern inline __attribute__((always_inline))
int memcmp (const void * p, const void *q, unsigned long size)
{
  return __builtin_memcmp (p, q, size);
}

are diagnosed with

t.c: In function 'memcmp':
t.c:2:5: warning: infinite recursion detected [-Winfinite-recursion]
    2 | int memcmp (const void * p, const void *q, unsigned long size)
      |     ^~~~~~
t.c:4:10: note: recursive call
    4 |   return __builtin_memcmp (p, q, size);
      |          ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~

This pattern happens in glibc fortify wrappers (but need -Wsystem-headers in
addition to -Wall).  It's reportedly also triggering for kernel wrappers
in its fortify-string.h which does not get the benefit of doubt via
-Wsystem-headers.

These kind of wrappers are not recursions (I think the kernel doesn't have
'extern' on the inline), inline instances will not be called itself and
__builtin_XXX are never inline expanded(?).