On Mon, Aug 08, 2016 at 06:00:23PM +0200, Francois Gouget wrote: > > For reference the crash happened in display_handle_stream_data() when > calling trying to queue the frame because video_decoder is NULL (same > sort of issue as on the server with video_encoder). > > st->video_decoder->queue_frame(st->video_decoder, in, latency) > > But if not there the call would happen in > display_session_mm_time_reset_cb() when calling rescheduling frames: > > st->video_decoder->reschedule(st->video_decoder); > > So it's simpler to just not create the stream object if we cannot create > the corresponding video decoder. > > > Note that currently create_gstreamer_decoder() would only fail if > gstvideo_init() fails but I'm looking at moving the create_pipeline() > call there so it only returns non-NULL if that succeeded.
I'd add that to the commit log rather than having the rationale/bug description buried in the mailing list archives. Christophe
signature.asc
Description: PGP signature
_______________________________________________ Spice-devel mailing list Spice-devel@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/spice-devel
