[mythtv] jack layer rewrite?

Ed W lists at wildgooses.com
Tue May 20 12:59:01 UTC 2008


cal wrote:
> With the utmost respect and humility toward the dev(s) involved, I thought
> I'd float my wistful wonderings regarding mythtv and jack output. The Way
> Back Machine shows a thread where Ed W hinted at some interest in updating
> the mythtv jack layer.
>   <http://www.gossamer-threads.com/lists/mythtv/dev/308886#308886>
>
> No request, no expectations, just wondering.
>   

It's still in my todo list...

It's fairly simple, except that one of the snags is that you really need 
a lockless audio buffer because the audio buffer callback is potentially 
very time sensitive.  At present the way the ringbuffer is coded (well 
at least how it was a few months ago...) is that it has a shared lock 
for reading and writing and this means that a writer can block a 
reader.  Ideally we need to split reading and writing locks and also 
quite a few of the locks can go altogether if we assume that there is 
only one writer thread (but this may or may not be true?)

Anyway it needs some slightly invasive updates to AudioOutputBase to get 
the locking a little better.  I really wanted to wait until all of 
Mark's patches were in before tackling it.  Also there is quite a bit of 
stuff crept in from the transcoder which I really think could be 
cleanedup quite a lot - seems like a good point to tackle it

However, a few days ago we gained a new little boy (Luke), so even more 
out of time right now

What's wrong with the current code though?  Working just fine here?

Ed W


More information about the mythtv-dev mailing list