owntone: New package variant: owntone-pulseaudio

[davidandreoletti]

Maintainer: @ejurgensen

Compile tested:

• arch: armv8
• model: NanoPI R4S
• OpenWrt version: 23.05

Run tested:

• arch: armv8
• model: NanoPI R4S
• OpenWrt version: OpenWRT version: OpenWrt 23.05.2 r23630-842932a63d / LuCI openwrt-23.05 branch git-23.306.39416-c86c256
• test done:
• [x] owntone-pulseaudio 28.5-2 installed on NanoPi R4S
  • I compiled + tested the changes above to this repo's master. Why ? master uses a new version of libsodium incompatible with the lib sodium version installed on 23.05 openwrt releases.
• [x] setup shairport-sync to write audio to named pipe
• [x] setup owntone read audio named pipe and stream to pulseaudio's default bluetooth sink (thanks for the write up on the owntone doc)
  • 🔊 People can hear music sent from an iPhone to OpenWRT (shairport-sync[0] -> named pipe -> owntone -> pulseaudio -> bluetooth stack) to Bluetooth Speaker !

[0] https://github.com/openwrt/packages/pull/22923

Description:

• Goal: Create package variant with pulseaudio support
• Packages size difference: + 4076 bytes
  • owntone_28.8-2_aarch64_generic.ipk: 734443 bytes
  • owntone-pulseaudio_28.8-2_aarch64_generic.ipk: 738519 bytes
  • Size reported with ls -ls --block-size 1 /tmp/shairport-sync-openssl_4.3.2-0_aarch64_generic.ipk

Irrelevant content now:

• @ejurgensen Building the package "owntone-pulseaudio" fails with this error.
  • I am hoping you can provide some guidance as to how to resolve the issues mark "FIXME".
    • @neheb's rpath-link solved the issue.

[neheb]

If I had to guess, rpath-link hack as used in mpd is needed.

[davidandreoletti]

If I had to guess, rpath-link hack as used in mpd is needed.

@neheb Good guess! Thank you.

[davidandreoletti]

owntone-pulseaudio is defined with DEPENDS+= +AUDIO_SUPPORT:pulseaudio.

• pulseaudio is a virtual package, declared in:
  • pulseaudio-daemon
  • pulseaudio-daemon-avahi

On my router: 0. pulseaudio-daemon-avahi is installed 1 opkg install /tmp/owntone-pulseaudio_28.8-2_aarch64_generic.ipk fails:

Installing owntone-pulseaudio (28.8-2) to root...
Installing pulseaudio-daemon (14.2-10) to root...
Collected errors:
 * check_conflicts_for: The following packages conflict with pulseaudio-daemon:
 * check_conflicts_for:         pulseaudio-daemon-avahi *
 * opkg_install_cmd: Cannot install package owntone-pulseaudio.

@neheb / @ejurgensen How to specify owntone-pulseaudio is fine with either dependency pulseaudio-daemon or pulseaudio-daemon-avahi installed (as long as one of them installed) ?

[davidandreoletti]

~Moving the discussion to https://forum.openwrt.org/t/owntone-support-pulseaudio/182872?u=dandreolett to get help from the wider community~ Problem solved. Discussion closed.

[davidandreoletti]

@ejurgensen

• Should this PR be merged in https://github.com/owntone/owntone-openwrt/tree/master instead ?
• Let me know if further work needed on the PR itself.

[ejurgensen]

Should this PR be merged in https://github.com/owntone/owntone-openwrt/tree/master instead ?

I think here is the right place, the owntone repo is for testing/CI. You are welcome to make a PR to that repo too, but otherwise I will pull it myself from here.

Let me know if further work needed on the PR itself.

Looks good to me, but I should say that I'm not an OpenWrt packaging expert.

People can hear music sent from an iPhone to OpenWRT (shairport-sync[0] -> named pipe -> owntone -> pulseaudio -> bluetooth stack) to Bluetooth Speaker !

Super cool!

[BKPepe]

1. Commit description is missing.

Ad) the size difference. Does it make sense to create variant in the first place instead of enabling it by default? Doing that for +4076 bytes. Hmmm. I would rather see that enabled it by default. You know. Every year requirements are higher and higher. Software is using more storage, etc.

[davidandreoletti]

I will pull it myself from here.

Ok

[davidandreoletti]

Commit description is missing.

done.

[davidandreoletti]

Does it make sense to create variant in the first place instead of enabling it by default? Doing that for +4076 bytes. Hmmm. I would rather see that enabled it by default.

I agree: Enabled by default. With DEPENDS+= +AUDIO_SUPPORT:pulseaudio, targets not supporting audio support will gracefully compile without pulseaudio support, ie current owntone package.

The PR has 2 commits:

• (a) https://github.com/openwrt/packages/pull/22942/commits/b46a8fe8a43356c5abead46869d81911d2e496b0: new owntone-pulseaudio package variant
• (b) https://github.com/openwrt/packages/pull/22942/commits/194d4c86b9e66aa58d5dc3265b902983496a52a7: original owntone package with conditional pulseaudio support added

@BKPepe / @ejurgensen Once a decision is taken, I will remove the irrelevant commit.

I prefer (b).

[davidandreoletti]

Does it make sense to create variant in the first place instead of enabling it by default? Doing that for +4076 bytes. Hmmm. I would rather see that enabled it by default.

I agree: Enabled by default. With DEPENDS+= +AUDIO_SUPPORT:pulseaudio, targets not supporting audio support will gracefully compile without pulseaudio support, ie current owntone package.

The PR has 2 commits:

• (a) b46a8fe: new owntone-pulseaudio package variant
• (b) 194d4c8: original owntone package with conditional pulseaudio support added

@BKPepe / @ejurgensen Once a decision is taken, I will remove the irrelevant commit.

I prefer (b).

@BKPepe / @ejurgensen (ping)

[ejurgensen]

I agree, just having it enabled makes sense if the difference is that small.

With DEPENDS+= +AUDIO_SUPPORT:pulseaudio, targets not supporting audio support will gracefully compile without pulseaudio support

Not sure I understand this - for package building, the "target" is the buildbot, right? Shouldn't they all build the same type of package?

[davidandreoletti]

This is how I understood how it works.

• A target is a hardware device.
• Each hardware device has different audio capabilities, some supported by OpenWRT.
• AUDIO_SUPPORT enables theowntone package to build against pulseaudio when the kernel / .config file support/enable it.

For my particular case, I don't think AUDIO_SUPPORT is required (audio over bluetooth stack). Yet I think it might be relevant for other device targets.

@BKPepe Would mind amending gaps in my understanding, if any please ?

[davidandreoletti]

@BKPepe Would mind amending gaps in my understanding, if any please ?

(Ping @BKPepe)

[BKPepe]

I will look at it. It looks like we dont have enough reviewers and I am trying to do it as much as I can in my free time, which I dont have much these days.

[davidandreoletti]

I will look at it. It looks like we dont have enough reviewers and I am trying to do it as much as I can in my free time, which I dont have much these days.

@BKPepe

I would like to merge this (once ready) before the next OpenWRT stable release.

So far the next release (timeframe seems not defined in any new meetings) and no release target set either.

Does that give you more or less time to review + answer this question ?

[davidandreoletti]

@BKPepe This PR needs your input: https://github.com/openwrt/packages/pull/22942#issuecomment-1911997425