Attaching a couple of patches for SVN. The first one modifies libpolyp's clock interpolation, so that it correctly interpolates instead of updating the ipol_timestamp on each get_time request. It also add some code paths to a) Try and ensure time doesn't advance past the moment when a stream is corked b) Not to advance the time past the current client-side write/read pointer - to avoid time advancing further than the client has written/read. I'm not sure if that is correct behaviour - should the server side clock be advancing anyway? At http://noraisin.net/~jan/gst/polyp/clock-jitter-raw.polyp.gnumeric you can find some charts showing the difference in clock output with this patch attached. Also, please take a look at the 3rd chart in the file, which shows a plot when using a 50ms period size in the client. It clearly shows that there are still artifacts at 10ms intervals, which I guess is based on the period that libpolyp is requesting timing updates from the server. I'm not sure how to fix those yet. The 2nd patch is some small cleanups for gst-polyp. The most important is removing gst_tag_list_free from the event handler - the tag list is owned by the event and will be freed when the event is. Let me know if you've any questions. Cheers, Jan. -- Jan Schmidt thaytan at noraisin.net Open Source Software: Free as in Free Speech, not Free Beer -------------- next part -------------- A non-text attachment was scrubbed... Name: polypaudio-clockinterp.patch Type: text/x-patch Size: 1917 bytes Desc: not available URL: <http://lists.freedesktop.org/archives/pulseaudio-discuss/attachments/20060605/b750b081/attachment.bin> -------------- next part -------------- A non-text attachment was scrubbed... Name: polypsink-1.patch Type: text/x-patch Size: 1559 bytes Desc: not available URL: <http://lists.freedesktop.org/archives/pulseaudio-discuss/attachments/20060605/b750b081/attachment-0001.bin>
