https://gcc.gnu.org/bugzilla/show_bug.cgi?id=120947
Jakub Jelinek <jakub at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |jakub at gcc dot gnu.org
--- Comment #1 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
(In reply to Siddhesh Poyarekar from comment #0)
> struct magic_ {
> char unused[9]; // at least 9
> };
>
> struct magic_map {
> struct magic_ *magic;
> };
>
> static int
> coalesce_entries(struct magic_ **ma)
> {
> size_t slen;
>
> slen = sizeof (**ma);
> *ma = __builtin_malloc (slen);
But note that in this case even a simple
foo ();
call here
>
> for (unsigned i = 0; i < 1; i++)
> {
> char b[1024] = {};
or here
> struct magic_ *ptr = *ma;
> (void) __builtin___memcpy_chk (ptr, b, sizeof (*ptr),
> __builtin_dynamic_object_size (ptr, 0));
or in a loop with more than one iteration also here will mean we need to punt,
at least when the coalesce_entries function isn't inlined into caller. Because
ma pointer
could have been stored e.g. to global variable and foo could have done *ma_copy
= __builtin_malloc (64); or *ma_copy = global_array; etc.
