Sending the flush-start event forward before taking the stream lock actually
works, in contrast to deadlocking in downstream preroll_wait (hunk 1).
After that we get the chain function being stuck in a busy loop. This is fixed
by updating the minimum frame size inside the synchronization loop because the
subclass asks for more data in this way (hunk 2).
Finally, this leads to a very probable crash because the subclass can find a
valid frame with a size greater than the currently available data in the
adapter. This makes the subsequent gst_adapter_take_buffer call return NULL,
which is not expected (hunk 3).
No need to unref buffers if pad_push returns something non-FLOW_OK.
In fact, this will cause assertions or crashes. However, we do need
to unref the input buffer whenever we don't pass it downstream. Also,
a non-OK flow return is not an error, so don't post error messages on
the bus - the pipeline may just be shutting down. Miscellaneous other
clean-ups and crack removal. Plenty of work left for those who feel
like it.
When sending the stream change message(s) to the demuxer, also include the
logical stream id along with the physical ID, so that the demuxer can remap
logical->physical MPEG stream ID.
Handle the new DVD commands, so that we can handle commands from a player
to change angle, jump to menus etc. Use the new GstNavigation event parsing
functions, instead of hand-rolled stuff.
Send GstNavigation notification messages when the mouse enters a button
or leaves it, so UI can turn the mouse cursor to a hand icon.
Make the highlights re-appear correctly when jumping back into menus
by making sure to set the flushing_seek flag for user-action initiated
seeks.
Fiddle some debug related to tracking down the issue.
Add a workaround for an apparent libdvdnav bug where it loses nav packets
during multiangle titles, and add some keypress bindings to switch the angle.
Add a 'fast-start' property to the rsndvdsrc element, that attempts to
jump directly to the DVD menu when starting. Doesn't work correctly on all
titles yet.
Add workarounds for issues with multiple angles in libdvdnav: Use a heuristic
to avoid detecting discontinuities during multiple-angle titles, it seems
caused by libdvdnav losing some NAV packets in multiangle titles. Fix
seeking in multi-angle titles by aligning our sector calculation logic
with libdvdnav's. Also, use libdvdnav's dvdnav_get_current_time() method
to determine the logical position of the current cell when it changes, as the
cell_start value in the cell_change event provides a number that doesn't
compensate for angle cell blocks.
If one side has a preference for a particular sample rate or set of sample rates, we
should honour this in the caps we advertise and transform to and from, so that elements
actually know about the other side's sample rate preference and can negotiate to it
if supported. Also add unit test for this.
State change to READY and then back to PAUSED should still provide
the proper structures as are otherwise freshly available following
a request_new_pad.
Pointed out by Thiago Santos.
