[Bug target/84521] [8 Regression] aarch64: Frame-pointer corruption with __builtin_setjmp/__builtin_longjmp and -fomit-frame-pointer

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84521

--- Comment #4 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Is the requirement just for functions that contain setjmp?  If so, the backend
could just force frame pointers in cfun->calls_setjmp functions.

If not, even if the default is tweaked again to be -fno-omit-frame-pointer on
aarch64, the code is still wrong with explicit -fno-omit-frame-pointer, even
before that change.

The test uses __builtin_setjmp, can't reproduce it when using normal setjmp,
so is it just __builtin_setjmp that requires frame pointers?  I think we don't
really have a flag about uses of __builtin_setjmp right now, but it could be
added next to calls_setjmp (calls_builtin_setjmp).  Marking the function with
__builtin_setjmp with __attribute__((optimize ("no-omit-frame-pointer"))) fixes
it too.