[PATCH] Revert "Linux: Keep termios ioctl constants strictly internal"
Andrew Pinski
pinskia@gmail.com
Fri Jul 25 15:46:26 GMT 2025
On Thu, Jul 24, 2025 at 5:38 PM Sam James <sam@gentoo.org> wrote:
>
> "H. Peter Anvin" <hpa@zytor.com> writes:
>
> > On July 24, 2025 3:02:46 PM PDT, Sam James <sam@gentoo.org> wrote:
> >>Andreas Schwab <schwab@linux-m68k.org> writes:
> >>
> >>> On Jul 24 2025, Sam James wrote:
> >>>
> >>>> It's fixed in GCC in r16-360-g1789c57dc97ea2 (don't think it's been
> >>>> backported, but we can handle that, and we're used to such problems with
> >>>> libsanitizer) and in LLVM In 59978b21ad9c65276ee8e14f26759691b8a65763.
> >>>
> >>> Nope, it's still broken.
> >>>
> >>> llvm/llvm-project#138822
> >>
> >>Thanks. I agree that's broken, but I can't reproduce the failure with
> >>glibc trunk and gcc trunk. It looks like it works for me because
> >>linux-headers provides TCGETA and friends via asm-generic/ioctls.h.
> >>
> >>Why does it fail for you, given this? Is your linux-headers patched (or missing)?
> >
> > The issue is with the presence of struct termio.
>
> I'm afraid I still don't grasp why it works for me and fails for
> Andreas, even if I can see that llvm/llvm-project#138822 clearly fixes a problem.
I was still getting a failure on aarch64 too.
Note scripts/build-many-glibcs.py turns off libsanitizer so it does
not test that.
My bet is if you undo the turning off libsanitizer in
scripts/build-many-glibcs.py, you can find all the targets which have
failures.
Thanks,
Andrew
More information about the Libc-alpha
mailing list