See also #109362 which may also add PipeWire support
Not to say your PR is worse or anything, just wanted to let you know about this PR too :)
But I'm not sure which one may be better or worse, or just different.

I rebased the commits so that this PR no longer depends on #109500.

f1c07cad65925fa942086cbae37c27d858af8d90 now uses print_verbose instead of ERR_FAIL_NULL_V macro if pw_context_connect failed in AudioDriverPipeWire::init, this is for suppressing errors on audio driver fallback when PipeWire library meets version requirement but the socket is not accessible.

Tested briefly on Fedora 43 (using PipeWire as audio server with default config) and it seems to work fine. I haven't reviewed the code yet.

How to validate that it's indeed using PipeWire directly and not via the pipewire-pulse plugin?

What kind of benefits are expected from this new driver, aside from sidestepping pipewire-pulse on systems defaulting to PipeWire?

Edit: I just read the proposal in godotengine/godot-proposals#10723. It mentions lacking audio input in ALSA, and too much latency in PulseAudio. Does this PR solve those two shortcomings?

Regarding the vendored headers in thirdparty/linuxbsd_headers/pipewire/, there's quite a lot of them. If possible, we like to trim down which thirdparty files we vendor to the minimum subset needed by Godot. I asked Copilot to assess which files could be safely removed and it came up with this: akien-mga@f8efbd3

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
                       Files        Lines         Code     Comments       Blanks
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
 Before                  205        27222        16279         6702         4241
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
 After                    79        13007         7115         3916         1976
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

The downside is that we'd have to document which files should be copied / removed when updating the library.

I check if a program is using pipewire directly by doing pw-dump > ~/pw-dump, searching for the program's binary name in the file and checking its "client.api", if there is no "client.api" for the program - it's using pipewire directly.

What kind of benefits are expected from this new driver, aside from sidestepping pipewire-pulse on systems defaulting to PipeWire?

If I did the profiling correctly, we can expect the new driver having much lower driver time (audio_driver in EngineDebugger) when compared with existing PulseAudio driver, and the new driver also implements audio input that ALSA driver lacks, which should solve both issues mentioned in godotengine/godot-proposals#10723.

If possible, we like to trim down which thirdparty files we vendor to the minimum subset needed by Godot.

That makes sense, I'll try to trim down unused headers in next push.

Are we ready for the next round of review?

Are we ready for the next round of review?

The PR is ready for the next round of review. Should I re-request the review, or wait for the reviewers?

I think re-requesting a review would be good, I asked on RocketChat for a review and that's how the previous round happened. Calinou also said he'd review but I haven't seen him so I think you could ping him too when asking for review

Hi @Calinou, can you help reviewing this PR? Thank you in advance!

Tested locally on Fedora 43, it mostly works as expected.

Some feedback:

* PipeWire appears to be used natively, according to the [above instructions](https://github.com/godotengine/godot/pull/111056#issuecomment-4074777513):

$ pw-dump > ~/pw-dump
$ rg -i "platformer" pw-dump
10101:        "application.name": "Platformer 2D",
10168:        "node.name": "Platformer 2D",
10563:        "object.path": "Platformer 2D:output_0",
10565:        "port.alias": "Platformer 2D:output_FL",
10648:        "object.path": "Platformer 2D:output_1",
10650:        "port.alias": "Platformer 2D:output_FR",

* Audio input doesn't seem to work in the [Mic Record](https://github.com/godotengine/godot-demo-projects/tree/master/audio/mic_record) demo (nothing is recorded, and the app doesn't appear in the list of recording applications). In comparison, it works in PulseAudio (but not in ALSA).

* PipeWire sounds noticeably quieter than PulseAudio/ALSA. I double-checked all audio drivers are at 100% for the per-app volume in the system settings.

Music from the Platformer 2D demo run through PulseAudio, ALSA, then PipeWire:

pulseaudio_alsa_pipewire.mp3

Waveform of the above audio normalized in Audacity for easier comparison (from left to right: PulseAudio, ALSA, PipeWire):

* I haven't done latency testing since I don't have the hardware for it currently, but it would be good to investigate this in the future.

Thanks for the review, the PR has been updated to 506940f with the following change:

• Audio position for input stream has been added (the primary reason why input stream didn't work as expected).
• When initializing input stream, set_input_device("Default") is called to update input channel count.
• Init / Start logic for output stream has been updated to remove the redundant active SafeFlag from the driver.

For the volume issue across different drivers, I can't seem to reproduce the same problem. The default per-app volume can be different when switching to different driver, but the waveform magnitude looks like the same when comparing with PulseAudio(first) and PipeWire(second) at 100% volume. More testing might be needed for this part.