[gst-devel] threads in 0.7

[Martin Janzen]

Benjamin Otte wrote:

 > I said it that way because GstQueue needs a rewrite because it's ugly
 > code (mostly because of the ability of being leaky that a) isn't used and
 > therefore b) doesn't work) and should have some more signals. People
 > already discussed that quite some time ago (once when 0.6 was new and
 > Colin did buffering for iradio and once when someone else who I don't
 > remember was interested in doing stuff with the queue), but people don'T
 > like touching ugly parts of the code.


Er, that might have been me.  I use a (leaky) queue to play video 
contained in MPEG-2 transport streams received over a network, something 
like this:

     udpsrc ! queue leaky=LEAK_DOWNSTREAM ! \
       { [my MPEG-2 TS demuxer] ! mpeg2dec ! queue leaky=NO_LEAK } ! \
       { xvideosink }

The first queue is my transport stream buffer, and leaks downstream in 
order to drop the oldest packets if the decoder thread can't keep up.

(I'm not sure why Benjamin says it doesn't work.  The 0.6.0 version -- 
with the #108268 and #109313 patches -- seems to be behaving itself so 
far.  Is there some subtle misbehaviour that I'm missing?)

Anyway, while I agree that GstQueue could use a bit of work, and admire 
the brave soul who takes on that project, I beg you to keep the 
leakiness feature!

-- 
Martin Janzen
janzen at pixelmetrix dot com