https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69971
Eric Gallager <egallager at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |egallager at gcc dot gnu.org
--- Comment #1 from Eric Gallager <egallager at gcc dot gnu.org> ---
(In reply to Martin Sebor from comment #0)
> Calling __builtin_return_address with a large argument (say due to a coding
> bug obscured by macros or template arguments) can result in resource
> exhaustion by the compiler as it emits at least one instruction for each
> level. It seems that for levels greater some threshold GCC could emit a
> loop instead to avoid this problem (as unlikely as it may be).
>
> The following is an example run on powerpc64le where GCC emits a stream of
> ld 9,0(9) instructions, one for each level:
>
> $ cat t.c && /build/gcc-trunk/gcc/xgcc -B /build/gcc-trunk/gcc -S -Wall
> -Wextra -Wpedantic -o/dev/stdout t.c | grep "ld 9" | wc -l
> void* foo (void)
> {
> void *p = __builtin_return_address (1024);
> return p;
> }
> t.c: In function ‘foo’:
> t.c:3:11: warning: calling ‘__builtin_return_address’ with a nonzero
> argument is unsafe [-Wframe-address]
> void *p = __builtin_return_address (1024);
> ^
> 1027
Isn't this why the -Wframe-address warning you're getting exists in the first
place?
