Linux port of libdispatch-685 merge

[dgrove-oss]

Minimal set of changes to get sources merged
from libdispatch-685 to compile and run on Linux.

[MadCoder]

that is very good, you should also get rid of all the code guarded with #if USE_FUTEX_SEM that is still around

[dgrove-oss]

revised based on feedback; squashed & force-pushed.

[MadCoder]

that's not the right patch please remove that hunk, and instead do this:

 DISPATCH_NOINLINE
 void
 _dispatch_queue_wakeup(dispatch_queue_t dq, pthread_priority_t pp,
        dispatch_wakeup_flags_t flags)
 {
    dispatch_queue_wakeup_target_t target = DISPATCH_QUEUE_WAKEUP_NONE;

    if (_dispatch_queue_class_probe(dq)) {
        target = DISPATCH_QUEUE_WAKEUP_TARGET;
    }
    if (target) {
        return _dispatch_queue_class_wakeup(dq, pp, flags, target);
+#if HAVE_PTHREAD_WORKQUEUE_QOS
    } else if (pp) {
        return _dispatch_queue_class_override_drainer(dq, pp, flags);
+#endif
    } else if (flags & DISPATCH_WAKEUP_CONSUME) {
        return _dispatch_release_tailcall(dq);
    }
 }

[MadCoder]

the reason is that _dispatch_queue_class_override_drainer() only makes sense if you have QoS support which linux doesn't so there's no point having the code in the first place, as you can see _dispatch_queue_class_override_drainer() is not reachable because it is itself under #if HAVE_PTHREAD_WORKQUEUE_QOS

[dgrove-oss]

Thanks. updated. Also had to guard calls to _dispatch_queue_class_override_drainer in source.c

[MadCoder]

Makes perfect sense.