PipeWire 0.3 Released With Redesigned Scheduling Code To Offer JACK2-Like Performance

[pIng2000]

I dont see any audio/video time delay offset in this. Its pretty much useless if this feature isn'nt implemented. Anybody seen parts in the code pointing at a feature like this?

[MrCooper]

Yeah... The only other way is with CAP_SYS_ADMIN but that's pretty much half-root...

I have read somebody tried making a solution which involves a CAP_SYS_ADMIN'd stub that grabs the file descriptor of the screen data and passes it to the OBS plugin, but I am not sure...

The KMS API doesn't allow applications like OBS to reliably capture the output of a display server which uses page flipping without the help of that display server. By the time OBS gets around to processing the FB retrieved via KMS, the display server / fullscreen application presenting to it may already be drawing another frame to that FB.

In general, the KMS API is for the display server only, not for direct use by applications (other than console applications presenting directly using KMS).

[pal666]

*I* want to e.g. ssh into my PC and play a Song/record over ssh and (more importantly) that both seats of my dual seat system have audio, no matter who logs in.

ok, so it's *you* want to spy on someone else. but this stuff works both ways, doesn't it?

[pal666]

I dont see any audio/video time delay offset in this.

it's hard to have video offset without having video in the first place

[QuImUfu]

ok, so it's *you* want to spy on someone else. but this stuff works both ways, doesn't it?

I want to use my PC in the way i see fit. If that would entail spying (it doesn't) then that's a usecase developers might (or might not) want to support. I currently have no "positive" usecase for audio recording **over ssh**, but i am sure there are some.

You can give the same argument for every feature.
Listing processes of other users in top? privacy validation!
Ssh as a whole? Can be used to control a botnet!
Loud audio playback? You surely want to annoy your room-mates!

My main usecase (that requires this) is however still my multi-seat system.
I want user two, physically sitting next to user one to be able to record sound while user one listens to something on headphones. Or user one have game sound while user to watches a Youtube video.
It is a feature i find useful and i want to know if there is support for that in PipeWire, because otherwise i might have to switch (if more applications drop ALSA support like Firefox) to an more... ugly solution like configuring PulseAudio on every user to stream the audio to localhost and have one user that takes all those streams and plays them. (because that is how well PulseAudio "supports" multiple users, last time i checked).

[FireBurn]

There is, in theory, by using DRM/KMS (see kmsgrab). It is not part of PipeWire though...

...plus you need a workaround for AMD cards (which have a terrible encoder by the way).

This took me a while to get working but proved useful to show a Plasma Wayland issue I was having, the big issue for me was stopping and starting X meant ffmpeg stopped recording

Would also be great if ffmpeg could list the possible providers, it took me digging in /sys/debug for drm to find the right values

[profoundWHALE]

If Pipewire is able to do what Apple did with CoreAudio then I think it'll be a homerun.

[Danielsan]

pipewire : pulseaudio = X : systemd.

And popcorn...