Florian Weimer, le dim. 09 mars 2025 23:32:28 +0100, a ecrit:
> > index 2257cd80d4..49a86f6ab3 100644
> > --- a/rt/aio_suspend.c
> > +++ b/rt/aio_suspend.c
> > @@ -161,7 +161,7 @@ ___aio_suspend_time64 (const struct aiocb *const
> > list[], int nent,
> > struct __timespec64 ts;
> > if (timeout != NULL)
> > {
> > - __clock_gettime64 (CLOCK_MONOTONIC, &ts);
> > + __clock_gettime64 (CLOCK_REALTIME, &ts);
> > ts.tv_sec += timeout->tv_sec;
> > ts.tv_nsec += timeout->tv_nsec;
> > if (ts.tv_nsec >= 1000000000)
>
> Shouldn't this be fixed in the other direction, using
> pthread_cond_clockwait with CLOCK_MONOTONIC?
Ah, I hadn't found it in the manpage, fixed so.
> Either way, I think this needs a bug in Bugzilla.
Ah, I missed filing it before sending the newer patch on the list, I
have fixed my local commit to close
https://sourceware.org/bugzilla/show_bug.cgi?id=32795
Samuel
