Samuel Thibault, le Sun 03 Feb 2013 01:52:20 +0100, a écrit : > Richard Braun, le Sat 02 Feb 2013 16:44:20 +0100, a écrit : > > diff --git a/sysdeps/mach/hurd/pt-docancel.c > > b/sysdeps/mach/hurd/pt-docancel.c > > index 105c6fd..b3a5507 100644 > > --- a/sysdeps/mach/hurd/pt-docancel.c > > +++ b/sysdeps/mach/hurd/pt-docancel.c > > @@ -36,6 +36,8 @@ __pthread_do_cancel (struct __pthread *p) > > assert (p->cancel_pending == 1); > > assert (p->cancel_state == PTHREAD_CANCEL_ENABLE); > > > > + __pthread_mutex_unlock (&p->cancel_lock); > > You should probably document that __pthread_do_cancel now assumes that > the cancel_lock is taken on entry, and released. Apart from that, the
is assumed* to be taken on entry, and released in __pthread_do_cancel Samuel
