[Bug 757488] gst-rtsp-server: multicast functionality is broken if using the same port ranges for multicast and unicast

[GStreamer (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=757488

--- Comment #79 from Sebastian Dröge (slomo) <slomo at coaxion.net> ---
(In reply to Patricia Muscalu from comment #77)

> > Just that udpsrc does not do anything with that flag if we provide it our
> > own socket :) That's why I went with the hammer solution now and just
> > disabled it directly on the socket from the very start.
> 
> Yes, I understand, but in that case we should remove these two lines above
> from the code. Maybe setting multicast loopback should be also done in the
> same function just to make sure that all configuration that is really needed
> is done in one place. What do you think?

For after 1.8.0 I would like all this code to be changed. Either we use
udpsrc/udpsink to set up all things for us, including providing the sockets...
or we do everything with the socket setup ourselves and provide them only to
the elements.

(In reply to Patricia Muscalu from comment #78)
> (In reply to Sebastian Dröge (slomo) from comment #76)
> > (In reply to Patricia Muscalu from comment #73)
> > > gst_rtsp_stream_get_multicast_address () will only get an address from the
> > > address pool if addr_v4 or addr_v6 are null.
> > 
> > Ah I missed that, thanks! But if I understand this correctly this can happen:
> > 1) get_multicast_address(), stores it in the stream then
> > 2) alloc_ports_one_family()
> > 3) Frees that address from 1) and gets a new one
> > 
> > I think instead 3) should use that address.
> 
> It's not good that we actually can get addresses in two ways. We should
> probably only allocate the multicast address in alloc_ports() and return
> this address in _get_multicast_address().

And make it an error to call get_multicast_address() before? If that works,
sure.

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