We currently test with 1.2.1 which is one year old. I think we should bump to a more recent release especially because of the gst upstream improvements in HLS. I have a patch locally but need to check tests with a Debug build.
Created attachment 241854 [details] Patch
Comment on attachment 241854 [details] Patch Fewer patches to maintain!. I guess this doesn't affect the layout tests results.
I have found only one affected test, see bug 138867
This can't be landed yet, the HLS tests are broken with this version... :(
Created attachment 242166 [details] patch Carlos, can you please review this again? I had to make a small change in the player as well.
Comment on attachment 242166 [details] patch View in context: https://bugs.webkit.org/attachment.cgi?id=242166&action=review > Source/WebCore/platform/graphics/gstreamer/MediaPlayerPrivateGStreamer.cpp:299 > + // Make sure that decodebin2 emits buffering messages on the > + // bus for non-local media files. We need those to manage the > + // ready and network states of the player. > + unsigned flagBuffering = getGstPlayFlag("buffering"); > + unsigned flags; > + g_object_get(m_playBin.get(), "flags", &flags, NULL); > + g_object_set(m_playBin.get(), "flags", flags | flagBuffering, NULL); How is this related to gst 1.4.4, why we didn't this before? Use nullptr, instead of NULL
(In reply to comment #6) > Comment on attachment 242166 [details] > patch > > View in context: > https://bugs.webkit.org/attachment.cgi?id=242166&action=review > > > Source/WebCore/platform/graphics/gstreamer/MediaPlayerPrivateGStreamer.cpp:299 > > + // Make sure that decodebin2 emits buffering messages on the > > + // bus for non-local media files. We need those to manage the > > + // ready and network states of the player. > > + unsigned flagBuffering = getGstPlayFlag("buffering"); > > + unsigned flags; > > + g_object_get(m_playBin.get(), "flags", &flags, NULL); > > + g_object_set(m_playBin.get(), "flags", flags | flagBuffering, NULL); > > How is this related to gst 1.4.4, why we didn't this before? I'm not sure yet. I still need to find the gst commit that introduced this change but in any case this change is harmless for any gst 1.x version. Since we've been relying on BUFFERING messages for quite a while now it's good anyway that we make sure those are emitted.
Created attachment 242265 [details] patch The http/tests/media/hls tests pass now \o/
Committed r176564: <http://trac.webkit.org/changeset/176564>
Re-opened since this is blocked by bug 139093
Relanded in r176627