GNOME Bugzilla – Bug 614306
output-selector, input-selector: move to core
Last modified: 2010-12-31 02:04:21 UTC
I notice that due to bug #586356 that a private copy of the selector plugin has already been moved to gst-plugins-base. Some applications like songbird depend on the selector plugin. It seems that this plugin should be fully integrated. What work, if any, is still needed? I don't see any bugs filed about this already.
I think this should be moved to core. The two elements are media agnostic and we have a copy in gst-plugin-base for playbin2 right now. The elements have docs and a few unit tests. I have pending patches for code cleanups (use BOILERPLATE macros).
Wim, you said some time ago, that input-selector is more or less ready but that there are some things in output-selector that need to be changed. Do you remember details?
-bad/tests/icles/output-selector-test just errors out with not-negotiated for me, that should be fixed.
Works fine for me. I'll try this on a few different machines as well. Tim, if you could attach a debug log.
Also tried at home with same result - works as expected. The systems are: intel core duo under ubuntu 10.04 amd phenon II X6 under opensuse 11.3 I'll try in on my netbook in the coming days as well.
Created attachment 174996 [details] debug log of failing output selector test (gzipped)
Can you run two xvimagesinks? The example hardcodes this :/ gst-launch v4l2src ! video/x-raw-yuv,width=320,height=240 ! tee name=t ! queue ! xvimagesink t. ! queue ! xvimagesink gst-launch v4l2src ! video/x-raw-yuv,width=320,height=240 ! timeoverlay ! tee name=t ! queue ! xvimagesink t. ! queue ! xvimagesink
Yes, those pipelines work fine here (if I couldn't run two xvimagesinks it would fail with an appropriate error and not not-negotiated).
Created attachment 175152 [details] [review] improve the test Tim, this is some test cleanup. According to you log videotestsrc and timeoverlay have issues negotiating the the same format, which is a bit weired as they are pure software and for me they negotiate fine. The attached patch adds a ffmpegcolorspace there and also uses autovideosinks instead of hardcoded sinks. Do these changes make it work?
> Tim, this is some test cleanup That makes it work, thanks!
Wim said on IRC he was fine with these moving as well, so moved to core, removed from -bad.