[Bug 750397] CRITICAL: Race condition in GstBus

[GStreamer (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=750397

--- Comment #76 from Sebastian Dröge (slomo) <slomo at coaxion.net> ---
(In reply to Matt Gruenke from comment #75)
> (In reply to Sebastian Dröge (slomo) from comment #74)
> > I've updated your patch now to set EBADF as errno if something goes wrong on
> > Windows. That's probably the closest that makes sense (write() and read()
> > can both return that, unlike the previous EACCES).
> 
> Okay, so do you need anything more from me?

No, thanks. I didn't want you to waste time on such trivial things ;)

> (In reply to Sebastian Dröge (slomo) from comment #72)
> 
> > And then hopefully get rid of the mutex once it's all
> > working, at least for Linux. Somehow.
> 
> Well, the non-atomic conversion and conditional update of control_pending
> will impede that effort.  I think it's possible to revert it back to being
> atomic, but I haven't been paying attention to whether we'd get burned by
> anything else.
> 
> At this point, I think it's safer to leave the mutex in there.  It should be
> cheap, especially by comparison with calls to select() and when we actually
> read() and write().

Yes, probably. I leave that as an exercise for someone else when it becomes a
problem / important enough. It will have to keep the API in any case, or become
a new API.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.