This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH 2/6] Use INLINE_SYSCALL_ERROR_RETURN
- From: Florian Weimer <fweimer at redhat dot com>
- To: "H.J. Lu" <hjl dot tools at gmail dot com>
- Cc: Andreas Schwab <schwab at suse dot de>, GNU C Library <libc-alpha at sourceware dot org>
- Date: Tue, 13 Oct 2015 16:02:58 +0200
- Subject: Re: [PATCH 2/6] Use INLINE_SYSCALL_ERROR_RETURN
- Authentication-results: sourceware.org; auth=none
- References: <20151012231944 dot GA27441 at intel dot com> <561D0388 dot 8030206 at redhat dot com> <CAMe9rOqC+Bar9XVrTutMJuo15qtiwts8AL-UMpaPQPc+WxXqNw at mail dot gmail dot com> <mvmfv1ex4iq dot fsf at hawking dot suse dot de> <CAMe9rOq_-0JngUXnHhnKcRJGkMwAjJRTiVguHVZXmgNWEn7orA at mail dot gmail dot com> <561D0DF3 dot 4090507 at redhat dot com> <CAMe9rOoJoJXTjanJKArTPNKK5jNQTvZ2XfnRYfzy+p=6X8O=YQ at mail dot gmail dot com>
On 10/13/2015 04:00 PM, H.J. Lu wrote:
> On Tue, Oct 13, 2015 at 6:58 AM, Florian Weimer <fweimer@redhat.com> wrote:
>> On 10/13/2015 03:56 PM, H.J. Lu wrote:
>>
>>> It looks (ptrdiff_t) isn't needed.
>>
>> It's not needed on x86_64, but we don't know if there will be
>> architectures which will eventually need it (or already do). The
>> existing code had -1L, so I assumed it was better to play it safe and
>> mirror that.
>
> Yes, the cast is also needed on x86-64 if __syscall_error is used:
>
> [hjl@gnu-tools-1 tmp]$ cat x.c
> #include <stddef.h>
>
> extern int __syscall_error (void);
> void *p;
>
> void
> foo (void)
> {
> p = (void *) (ptrdiff_t) __syscall_error ();
> }
> [hjl@gnu-tools-1 tmp]$
Oh, right.
I think it's best not to use the new macro in those two cases. If the
cast is not a no-op, it's no longer a tail call, after all.
Florian