https://gcc.gnu.org/bugzilla/show_bug.cgi?id=117359
--- Comment #14 from H. Peter Anvin <hpa at zytor dot com> --- This is something that should be documented, if it is the construct to be relied on to have this effect. In the Linux kernel it has also been used to force the frame pointer to be set up, but that feels fragile to me (see bug 117311) and it doesn't work with "asm goto" in older gccs. I propose in that BR to instead use a dependency on __builtin_frame_address(0) *for that use case* (redzone protection not included.) I don't care if the construct is ugly or confusing, as long as it is documented and supported.
