On Fri, 25.12.09 20:23, David Henningsson (launchpad.web at epost.diwic.se) wrote: > > Corking means that the application asked for the audio stream to be > > "corked" temporarily, so that data flow stops. "uncorking" then makes > > things flow again. > > > > It's mostly synonymous to application triggered pause/unpause. > > Instantaneous pause, i e without "drain buffers" first, right? No data is dropped. On tsched enabled systems the pausing should happen almost immediately, just a few safety samples later than the current hw playback index. On non-tsched systems the pausing will happen as soon as "possible", i.e. with a delay of once the hw buffer size. As soon as you uncork playback starts right-away again. > > In both cases this is triggered by the app explicitly. > > Thanks, that answers some questions but also raises new ones... > > There are rows saying "Requesting rewind due to rewrite" but there are > no rows following up that rewind, as seen in other logs[2]. How > come? Not all the same a rewind is even possible. > Second, can you (or anyone else) think of a scenario where these lines > typically reoccurs frequently, or are we looking at an error, either in > GStreamer[1] or in PulseAudio? It is not necessarily an error if these lines appear. For example if the client does not provide data to PA when the buffer actually runs empty but based on its own client side timing. Lennart -- Lennart Poettering Red Hat, Inc. lennart [at] poettering [dot] net http://0pointer.net/lennart/ GnuPG 0x1A015CC4