Trouble playing HTTP-URIs gapless in "playbin"

[Henner Zeller]

Did any of the developers had a chance to take a look ? I might want to
investigate the reasons next weekend, but it might be helpful to have some
rough plan of action.

This is essentially a regression from 0.10: gapless playing of non-file
URIs used to work with the suggested about-to-finish callback, but doesn't
with 1.0.
(but in 0.10 it was leaking threads...)



On 15 April 2013 23:13, Henner Zeller <h.zeller at acm.org> wrote:

> Hi,
> I am maintainer of gmrender-resurrect [1], a UPnP renderer (forked from
> the dormant 'gmrender' project). It is using gstreamer as backend.
>
> Using the gstreamer "playbin" feature to set the next URI in an
> "about-to-finish" callback, this UPnP renderer successfully implements
> gapless playback, alas, not without trouble.
>
> It turns out, that playing gapless with inputs from HTTP-URLs is leaking
> threads, up to the point that things stop working. This is with
> gstreamer-0.10. With gstreamer-1.0 setting HTTP URLs in the
> "about-to-finish" callback does not seem to work at all.
>
> This _does_ work with local file-URIs, so I believe this has to do with
> different handling of network resources vs. local resources, i.e.
> buffering, probably a separate thread handling the network input (and it
> would explain why this hasn't been noticed before, because the usual
> use-case is to read from files).
>
> To reproduce, I created simplified test-code including Makefile and
> test-sound in this public git repository that illustrates the behavior:
>
>   https://github.com/hzeller/gstreamer-gapless-test
>
> All it does is to re-set the original URI back once the "about-to-finish"
> callback is called ... and exhibiting the described behavior:
>   - gstreamer-0.1: runs out of threads on gapless playing URIs
>   - gstreamer-1.0: does not handle well having the URI set in
> about-to-finish callback.
>    - both work with file URIs
>
> Instructions to reproduce are under
>   https://github.com/hzeller/gstreamer-gapless-test/blob/master/README.md
>
> Maybe, this is a usage error, but I suspect it to be a bug. If you can
> confirm that this _should_ work, then I am happy to file a bug, and
> probably dig into the code to help working on a fix.
>
> Thanks,
>   Henner.
>
> [1] https://github.com/hzeller/gmrender-resurrect
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/gstreamer-devel/attachments/20130417/850d1422/attachment.html>