Unable to find encoder about sunshine

[cyl3x]

I fixed my issue by adding -DLIBSWSCALE_INCLUDE_DIR=/usr/include/ffmpeg4.4 -DLIBSWSCALE_LIBRARIES=/usr/lib/ffmpeg4.4/libswscale.so as CMAKE Flags in the PKGBUILD from AUR

[emmyoh]

Perhaps this is related to the FFmpeg 5.0 release? The Arch Linux package ffmpeg4.4 was used to build Sunshine on my system, so the recent 5.0 release is not the cause.
How may I help with debugging this?

[Theslees]

I'm also having this issue, same OS and terminal output, but using a different GPU. (AMD)

[cyl3x]

same issue here, Arch and NVIDIA Card

[ReenigneArcher]

@hadogenes any thoughts?

[JPenuchot]

Same issue, same config. I switched to this fork because of an AUR update and now I can't get Sunshine to run. However I tried building the package at its last version targeting the original repo and I still run into this problem, so it might not be a regression. I'm also seeing CMake not detecting any CUDA compiler despite CUDA_HOME and CUDA_PATH both being set to /opt/cuda, so I'm suspecting a build system issue to be the root cause.

PS: CUDA compiler not being detected happens on both the original repo and this fork. I guess it's worth fixing this repo as it's more reactive to PRs and already being tracked by the AUR package.

[JPenuchot]

Update: The CMake script actually finds the CUDA install present on my system, although not when being called through makepkg when building the AUR package. Not sure the lack of a CUDA compiler is what's causing the error, but maybe what's causing CMake to not find nvcc could also be causing it to not find an essential library.

[JPenuchot]

After a bit of debugging it turns out the issue happens here:
https://github.com/SunshineStream/Sunshine/blob/4a483078ad983b159d349e8056fccf61d918dc74/sunshine/video.cpp#L740

Both this fork and the original repo are impacted.

[JPenuchot]

Another note: I don't get this issue on my laptop with Intel integrated graphics. It's possible that it was caused by a libdrm or Nvidia update too.

[emmyoh]

OK … so it's not a regression introduced from the fork (so is not due to a code change), is not unique to Nvidia, but may possibly still be connected to something with Arch Linux? I'm quite lost as to what the cause could be. I had the pre-fork version of Sunshine installed an hour before opening this issue, and my system was 100% up-to-date at the time, so the only updates that could've possibly affected things is anything statically-linked to the compiled binary. The fact that it works on Intel is interesting.

[ReenigneArcher]

The AUR was recently updated and there were changes in the build besides pointing to this repo. There were 3 commits a few days ago https://aur.archlinux.org/cgit/aur.git/log/?h=sunshine

[JPenuchot]

OK … so it's not a regression introduced from the fork (so is not due to a code change), is not unique to Nvidia, but may possibly still be connected to something with Arch Linux? I'm quite lost as to what the cause could be. I had the pre-fork version of Sunshine installed an hour before opening this issue, and my system was 100% up-to-date at the time, so the only updates that could've possibly affected things is anything statically-linked to the compiled binary.

The issue points to avcodec, and it is statically linked indeed. I'll have a look at their bug tracker to find out if there's been any sort of regression.

The fact that it works on Intel is interesting.

Not 100% sure but as far as I know my 6300U doesn't have any hardware encoder

[JPenuchot]

And there it is: https://trac.ffmpeg.org/ticket/9666

[JPenuchot]

Sorry I just found out I forgot to build Sunshine with ffmpeg 4.4, so I just polluted this issue with something that's already been solved here: #63

I've added the CMake flags to link against ffmpeg 4.4 and I'm back to the Couldn't expose some/all drm planes for card: /dev/dri/card0 error

[pogojotz]

I can confirm that @cyl3x's solution works. Cheers for that 🍻

[JPenuchot]

Back to business indeed. Thanks a lot :)

[pogojotz]

This has been resolved in the AUR. Can be closed.

[mdeguzis]

In my case, I needed to make sure I set the DISPLAY in the systemd unit file. I noticed things would run fine if I didn't launch the app image via systemd

[gamer@chimeraos tmp]$ cat ~/.config/systemd/user/sunshine.service 
[Unit]
Description=Sunshine self-hosted game stream host for Moonlight.
StartLimitIntervalSec=500
StartLimitBurst=5

[Service]
ExecStart=%h/software/sunshine.AppImage
Restart=on-failure
RestartSec=5s
Environment=DISPLAY=chimeraos:10.0
#Flatpak Only
#ExecStop=flatpak kill dev.lizardbyte.sunshine

[Install]
WantedBy=graphical-session.target

[VRKade11]

do you mind if you elaborate on the solution a little bit I have the code but I don't know where to put it

[ReenigneArcher]

@VRKade11 you need to set the DISPLAY environment variable

[VRKade11]

Well I figured it out it was the output name and adapter name needed to be blank or I messed something up with the name