audio: Wait for state change instead of async done in scanner.
This increases the robustness by quite a bit, we no longer: - Fail on elements such as mmssrc that can't go to playing ASYNC. - Try going to playing if we've already tried before.
This commit is contained in:
parent
c0836d9d97
commit
74cf673171
@ -35,7 +35,7 @@ Feature release.
|
||||
|
||||
- Audio: Fix handling of MMS (and any other streams) that can't switch to
|
||||
playing ``ASYNC``. Previously we would time out trying to get a duration from
|
||||
these. (Fixes: :issue:`1553`, PR :issue:`1575`)
|
||||
these. (Fixes: :issue:`1553`, PR :issue:`1575`, :issue:`1576`)
|
||||
|
||||
|
||||
v2.0.1 (2016-08-16)
|
||||
|
||||
@ -194,13 +194,14 @@ def _process(pipeline, timeout_ms):
|
||||
missing_message = None
|
||||
duration = None
|
||||
|
||||
# TODO: Turn this into a callback table to cleanup code.
|
||||
types = (
|
||||
Gst.MessageType.ELEMENT |
|
||||
Gst.MessageType.APPLICATION |
|
||||
Gst.MessageType.ERROR |
|
||||
Gst.MessageType.EOS |
|
||||
Gst.MessageType.ASYNC_DONE |
|
||||
Gst.MessageType.DURATION_CHANGED |
|
||||
Gst.MessageType.ELEMENT |
|
||||
Gst.MessageType.EOS |
|
||||
Gst.MessageType.ERROR |
|
||||
Gst.MessageType.STATE_CHANGED |
|
||||
Gst.MessageType.TAG
|
||||
)
|
||||
|
||||
@ -235,25 +236,25 @@ def _process(pipeline, timeout_ms):
|
||||
raise exceptions.ScannerError(error)
|
||||
elif msg.type == Gst.MessageType.EOS:
|
||||
return tags, mime, have_audio, duration
|
||||
elif msg.type == Gst.MessageType.ASYNC_DONE:
|
||||
success, duration = _query_duration(pipeline)
|
||||
if tags and success:
|
||||
return tags, mime, have_audio, duration
|
||||
elif msg.type == Gst.MessageType.STATE_CHANGED and msg.src == pipeline:
|
||||
old_state, new_state, pending = msg.parse_state_changed()
|
||||
|
||||
# Workaround for upstream bug which causes tags/duration to arrive
|
||||
# after pre-roll. We get around this by starting to play the track
|
||||
# and then waiting for a duration change.
|
||||
# https://bugzilla.gnome.org/show_bug.cgi?id=763553
|
||||
logger.debug('Using workaround for duration missing before play.')
|
||||
result = pipeline.set_state(Gst.State.PLAYING)
|
||||
if result == Gst.StateChangeReturn.FAILURE:
|
||||
return tags, mime, have_audio, duration
|
||||
if pending == Gst.State.VOID_PENDING:
|
||||
success, duration = _query_duration(pipeline)
|
||||
if tags and success:
|
||||
return tags, mime, have_audio, duration
|
||||
|
||||
# Try again to get duration since we changed without async, then
|
||||
# just give up as there is nothing more that can happen.
|
||||
if result == Gst.StateChangeReturn.SUCCESS:
|
||||
_, duration = _query_duration(pipeline)
|
||||
return tags, mime, have_audio, duration
|
||||
if new_state == Gst.State.PAUSED:
|
||||
# Workaround for upstream bug which causes tags/duration to
|
||||
# arrive after pre-roll. We get around this by starting to
|
||||
# play the track and then waiting for a duration change.
|
||||
# https://bugzilla.gnome.org/show_bug.cgi?id=763553
|
||||
logger.debug('Using workaround for duration missing.')
|
||||
result = pipeline.set_state(Gst.State.PLAYING)
|
||||
if result == Gst.StateChangeReturn.FAILURE:
|
||||
return tags, mime, have_audio, duration
|
||||
else:
|
||||
return tags, mime, have_audio, duration
|
||||
|
||||
elif msg.type == Gst.MessageType.DURATION_CHANGED:
|
||||
# duration will be read after ASYNC_DONE received; for now
|
||||
|
||||
Loading…
Reference in New Issue
Block a user