Re: scratch/igc warning

[Gerd Möllmann]

Andrea Corallo <acorallo@gnu.org> writes:

> Hi,
>
> GCC 12.3 is giving me this warning building scratch/igc:
>
> ========
>   CC       xfaces.o
> In file included from /usr/include/string.h:535,
>                  from ../lib/string.h:41,
>                  from lisp.h:29,
>                  from xfaces.c:226:
> In function ‘memset’,
>     inlined from ‘make_realized_face’ at xfaces.c:4581:3,
>     inlined from ‘realize_gui_face’ at xfaces.c:6214:10,
>     inlined from ‘realize_face’ at xfaces.c:6124:12:
> /usr/include/x86_64-linux-gnu/bits/string_fortified.h:59:10: warning: 
> ‘__builtin_memset’ offset [164, 303] from the object at ‘face’ is out of the 
> bounds of referenced subobject ‘id’ with type ‘int’ at offset 160 
> [-Warray-bounds]
>    59 |   return __builtin___memset_chk (__dest, __ch, __len,
>       |          ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>    60 |                                  __glibc_objsize0 (__dest));
>       |                                  ~~~~~~~~~~~~~~~~~~~~~~~~~~
> In file included from termhooks.h:27,
>                  from frame.h:22,
>                  from xfaces.c:229:
> dispextern.h: In function ‘realize_face’:
> dispextern.h:1730:7: note: subobject ‘id’ declared here
>  1730 |   int id;
>       |       ^~
> In function ‘memset’,
>     inlined from ‘make_realized_face’ at xfaces.c:4581:3,
>     inlined from ‘realize_tty_face’ at xfaces.c:6641:10,
>     inlined from ‘realize_face’ at xfaces.c:6126:12:
> /usr/include/x86_64-linux-gnu/bits/string_fortified.h:59:10: warning: 
> ‘__builtin_memset’ offset [164, 303] from the object at ‘face’ is out of the 
> bounds of referenced subobject ‘id’ with type ‘int’ at offset 160 
> [-Warray-bounds]
>    59 |   return __builtin___memset_chk (__dest, __ch, __len,
>       |          ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>    60 |                                  __glibc_objsize0 (__dest));
>       |                                  ~~~~~~~~~~~~~~~~~~~~~~~~~~
> dispextern.h: In function ‘realize_face’:
> dispextern.h:1730:7: note: subobject ‘id’ declared here
>  1730 |   int id;
>       |       ^~
> In function ‘memset’,
>     inlined from ‘make_realized_face’ at xfaces.c:4581:3,
>     inlined from ‘realize_face’ at xfaces.c:6130:14:
> /usr/include/x86_64-linux-gnu/bits/string_fortified.h:59:10: warning: 
> ‘__builtin_memset’ offset [164, 303] from the object at ‘face’ is out of the 
> bounds of referenced subobject ‘id’ with type ‘int’ at offset 160 
> [-Warray-bounds]
>    59 |   return __builtin___memset_chk (__dest, __ch, __len,
>       |          ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>    60 |                                  __glibc_objsize0 (__dest));
>       |                                  ~~~~~~~~~~~~~~~~~~~~~~~~~~
> dispextern.h: In function ‘realize_face’:
> dispextern.h:1730:7: note: subobject ‘id’ declared here
>  1730 |   int id;
>       |       ^~
> ========
>
> ATM don't know if it's a real issue or a false positive so I thought was
> worth mentioning.

Hm, that's this one, right?

  static struct face *
  make_realized_face (Lisp_Object *attr)
  {
    enum { off = offsetof (struct face, id) };
  #ifdef HAVE_MPS
    struct face *face = igc_make_face ();
  #else
    struct face *face = xmalloc (sizeof *face);
  #endif
    memcpy (face->lface, attr, sizeof face->lface);
    memset (&face->id, 0, sizeof *face - off);
    face->ascii_face = face;

    return face;
  }

which we can rewrite, for MPS

  static struct face *
  make_realized_face (Lisp_Object *attr)
  {
  #ifdef HAVE_MPS
    struct face *face = igc_make_face ();
    memcpy (face->lface, attr, sizeof face->lface);
  #else
    enum { off = offsetof (struct face, id) };
    struct face *face = xmalloc (sizeof *face);
    memcpy (face->lface, attr, sizeof face->lface);
    memset (&face->id, 0, sizeof *face - off);
  #endif
    face->ascii_face = face;

    return face;
  }

because igc returns something zero-initialized anyway. (I'll do that.).

But I wonder - does GCC 12 also complain for the non-MPS case?