devout trout
I'm somewhat often getting this error on the screen under the ETV logo when a show starts:
Impossible to convert between the formats supported by
the filter 'Parsed_hwupload_9' and the filter 'auto_scale_2'
Error reinitializing filters!
Failed to inject frame into filter network: Function not implemented
Error while processing the decoded data for stream #0:0
Any idea how to get around this?
idle forge
can you give info on your gpu and how you have accel configured in your ffmpeg profile
devout trout
GPU is Intel onboard AlderLake-S GT1 Vaapi container, ETV version 0.7.9-4af2d7aa-docker-vaapi
ffmpeg config
idle forge
which kernel do you have on the host
devout trout
Running under Docker.
Switched the vaapi driver to "default" and rebooted. Show stared to play. Beyond Docker I'm on Unraid which runs Linux kernel v6.1.33 under Uneraid 6.12
devout trout
Still happening with default.
idle forge
can you paste ffmpeg hls arguments from when it happens
would also be interested in the vainfo -a output, which should also be available in troubleshooting page
devout trout
Reboot does allow it to play, though.
idle forge
the same content that later fails?
devout trout
Fails, I restart the container, channel plays fine.
idle forge
same content?
like, make a test channel with one item, and that item will stop working but when you restart the container that item will work?
devout trout
Yep, All I did was reboot the channel then have the same client try to play the channel again and it plays through fine.
Not sure.
idle forge
can you send today's log file in a dm
idle forge
vainfo -a output would also help (can also find in troubleshooting page of latest release)
devout trout
where are the full logs? Docker is only showing me stuff since reboot.
idle forge
config folder / logs subfolder
got them thanks, will look later
ah right, one possible fix is to set your ffmpeg profile to 1 thread, not 0 (auto)
hw accel should always use one thread
devout trout
Ok, I'll drop that in and monitor, thanks!
stable warren
would it be possible to force it to 1 when hw accel is enabled?
That was something that caught me early on, until I realized that multiple GPU transcoding threads didn't make sense.
devout trout
Just got the error again. This time upload 10 and scale 4 with 1 set.
idle forge
i previously had it forced to 1 but there were complaints, i'll think about re-implementing that
the other thing you could do is temporarily enable ffmpeg troubleshooting reports (switch in etv's settings page), and when you get the error again, zip up the ffmpeg-reports config subfolder and dm it to me (then turn the switch back off). that might contain more info than the normal log
devout trout
I know this is super old and not sure which version change helped it out, but now when this happens it is only for the duration of the particular video, I believe, then ETV carries on. Nice fix.