https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88576

--- Comment #28 from Richard Biener <rguenth at gcc dot gnu.org> ---
(In reply to Rich Felker from comment #26)
> Thanks for fixing! I was looking at hooking up TARGET_DEFAULT_NO_MATH_ERRNO
> for musl targets with this fix now present, but it looks like it's not even
> used even though there's one instance of it being set in config.gcc. Am I
> mistaken? Should I open a tracker ticket for that?

I didn't know of TARGET_DEFAULT_NO_MATH_ERRNO, it was possibly removed.

Hooking this up might be possibly done by amending the
TARGET_OPTION_OPTIMIZATION_TABLE (common/config/i386/i386-common.cc
for x86).  There's currently no OS common mechanism implemented,
honoring TARGET_DEFAULT_NO_MATH_ERRNO might be viable there or in the
consumer, but it should be (re-)documented then.

Reply via email to