last few seconds of video are not played

[Mark Spieth]

when a recording or video is played, the last few seconds are not played.

There are 2 reasons for this:

1. appended audio with no more video is sent and discarded but the audiooutput.

so if not paused, feed the audio into the output buffer carefully but not with infinite wait.

2. eof is detected and any frames in the queue are discarded.

delay eof activation until all frames are played

needs some testing with unusual cases such as audio only and video only.

also see #4359

this is another of my annoyances fixed.

patch attached.

[anonymous]

This is great news if this patch fixes the problem - I really hope this patch gets applied. Also see ​http://svn.mythtv.org/trac/ticket/6148 which is a ticket opened after the last fix didn't work.

[sphery]

Allows playing recording closer to the end when end-of-recording prompt is enabled. Must be used with play-to-end.patch .

[sphery]

I attached a patch, mythtv-6974-IsNearEnd_margin_for_end_of_recording_prompt.patch , which should be used in conjunction with Mark's play-to-end.patch .

mythtv-6974-IsNearEnd_margin_for_end_of_recording_prompt.patch changes TV::HandleEndOfRecordingExitPromptTimerEvent?() to check to see if we're within 1/3 second of the end of the recording when determining whether to display the end-of-recording prompt. Without the patch, it's checking to see if we're within 2 seconds of the end of the recording, meaning we display the end-of-recording prompt about 2 seconds from the end of the recording. The patch factors in timestretch when calculating the 1/3-second margin and, due to the timer's checking for the end of the recording every 1/4 second, seemed to work reliably on my systems in all cases. If 1/3 second isn't reliable on all systems/video files, it could be increased to 1/2 second.

Note that mythtv-6974-IsNearEnd_margin_for_end_of_recording_prompt.patch can *not* be used without play-to-end.patch , as we'll hit the issues Mark mentioned before the prompt is displayed and exit back to the Watch Recordings screen without ever displaying the prompt for those users who have enabled the end-of-recording prompt. This is probably why we're currently using a 2-second margin.

Note that the 2-second margin is the issue reported in #6418 , mentioned above.

[Mark Spieth]

I was thinking of handling the prompt with a secondary loop outside the main loop which is activated on eof exit and prompt menu is required. That way it will truely be the end and no nearness test is required.

one of the menu options isnt needed as you cant go back and play more (cancel) because you are at the end already.

Havent had time to implement yet.

[sphery]

As mentioned in ​http://www.gossamer-threads.com/lists/mythtv/dev/452617#452617 , play-to-end.patch is no longer working as intended, and mythtv-6974-IsNearEnd_margin_for_end_of_recording_prompt.patch relies on play-to-end.patch (but should be reimplemented as at comment:4 , anyway), so since we don't have any usable patches for this ticket, I'm closing it for now. Anyone should feel free to reopen the ticket with an updated/fixed patch(es).

[Mark Spieth]

update to play to end patch.

reopening as a result.

[stuartm]

Maybe this won't get fixed for 0.25, but it is a bug with a patch supplied so ...

[Mark Spieth]

updated reasonably current patch for play to end

[Jim Stichnoth]

Pushing back to 0.26, but could go into 0.25-fixes.

[Mark Spieth]

current patch to ease integration

[Jim Stichnoth <jstichnoth@…>]

In 91ec76e8afba1833a7cb64f17429910f29b39404/mythtv:

Play closer to the end of the video.

Refs #6974. When the end of the recording is reached, don't exit the
player immediately. Instead, pause playback and let exiting be
handled by either the end-of-playback timer event or the
end-of-recording-delete-prompt timer event. This removes IsNearEnd?()
from end-of-playback consideration, as well as the randomness of where
in the 250ms window the timer events fire.

IsNearEnd?() is still used in other places (e.g. to control playback
speed and bookmarking behavior near the end of a recording). The
IsNearEnd?() calculation is changed to depend on the number of frames
played by the player rather than the number of frames read by the
decoder thread, making IsNearEnd?() more predictable and reliable.

Note that this does not address the original issue in #6974, which
involves allowing the accumulated decoded frames to be displayed after
the decoder reaches EOF.

[Jim Stichnoth <jstichnoth@…>]

In 31126568be61528ed05eef04ca0bf7f0a3d4b004/mythtv:

Fixes #6974. Play all the way to the end of the video.

Playback is allowed to continue even after the decoder thread reaches
the EOF state, until decoded and buffered frames are drained from the
decoder. This may represent anywhere from 0.25s to 1s of content.

Thanks to Mark Spieth for providing and maintaining the patch.

[Jim Stichnoth <jstichnoth@…>]

In cef316f803117339a3e954531aedd00b462a81e9/mythtv:

Avoid premature playback exit for audio-only recordings.

Stop playback only when the decoder, video output, and audio output
all drain. Refs #11357. Refs #6974.