Another day another regression: this time it’s the Downloads feature. Some specific files will transcode (HW) to 100%, fail, and then retranscode (SW):
[fc#0 @ 0x7f1cdc6a1040] Error while filtering: Out of memory
Failed to inject frame into filter network: Out of memory
Error while filtering: Out of memory
I tried changing a bunch of variables but I couldn’t reproduce this with other files, so I guess I’ll share this 10 second sample (it’s just the studio intro):
I tried different videos with different resolutions and with different flavors of DTS but I couldn’t get it to (not) work. I guess I didn’t try to ffprobe each frame but I’ll leave that up to you.
P.S. While testing I only noticed that the Transcoder doesn’t like it when there’s no audio in the file. That might not be a regression though (too lazy to check) so you can put it on your WONTFIX pile if you want.
Are you sure about that? The transcode is so fast, you might have missed that it happend twice. Or maybe it’s related to Android / the new app? It’s definitely a server regression though.
I did go through the log and there’s nothing interesting in it:
VERBOSE - Didn't receive a request from 127.0.0.1:41966: End of file
VERBOSE - JobManager: child process with handle 3983747 exited
DEBUG - Jobs: '/usr/lib/plexmediaserver/Plex Transcoder' exit code for process 3983747 is 244 (failure)
DEBUG - [Req#55a18] Streaming Resource: Terminated session 0x7fc1d0299ac8:fb4fd5e1-b313-418a-810e-d414421e186f with reason Client stopped playback.
DEBUG - [Req#55a18] Streaming Resource: Removing session 0x7fc1d0299ac8:fb4fd5e1-b313-418a-810e-d414421e186f
“with reason Client stopped playback” is certainly false. The Transcoder just doesn’t like the file for some reason (btw it’s also certainly not “out of memory”).
Anyway for my 3rd (and hopefully last regression): PGS burn-in subtitles are green (instead of white). It’s better than the latest release (where it would fall back to SW or just crash the server) but still not as good as it was before.
@bur115 when i said client specific, each client specifies different parameters for the download (supported codecs, ect), It would have been much easier to reproduce if I had those parameters for your device.
That being said, I have been able to reproduce and address all your issues except the green subtitles. Does that only happen for certain files? If so can you send me a sample? if its every file I wonder if its a driver/device issue.
(Note that it’s not a valid MKV file because I had to truncate it to 8 MB but it still plays and the subtitles are still green).
Edit: By the way I should mention that as per HEVC Encoding Forum Preview - #686 by Ari it will not work if there’s any scaling involved, so if your device can play DTS, make sure to set the quality to Original and to disable Direct Play/Stream.
@bur115 I can’t seem to replicate the green subtitles even with the sample you sent me, which makes me suspect this is a driver/device dependent issue. Are the subtitles green with tonemapping, without tonemapping, or both?
@pl_5309 If you have time and can, could you repeat this and capture the logs after observing the issue?
@Anaerin checking in on you too. The fix is reportedly in but we have not been able to reproduce. This seems file specific. If you could try to reproduce with the latest forum preview build and report back (with logs if there is an issue) that would be extremely helpful.
Closed Captions not saved with recording when using Intel.
12:00 - 12 :18 Started a live stream one at a time on CBS, NBC and ABC then activated CC. ABC failed the first time, succeeded on retry.
12:30 - 2 PM 3 Recordings on CBS, NBC ABC saving as HEVC, no subs saved.
2:18 - 2:25 3 Live Streams one at a time PBS, FOX, CW activating CC, had a failure on the third station so something about 3 tunes in short succession is a problem.
Older logs show Nvidia recording to HEVC and saving CC successfully with version 1.42.1.10060 (Intel fails)
If the take away is that FFMPEG 6.1 is working and the failure of saving EIA_608 with the recording while using the Intel path is already an existing problem then I agree.
and capture the logs after observing the issue?