Support Opus codec

[rubdos]

(untested)

Compiles, seems to index the files, will test a bit more extensively on the road!

It plays! I'll have to test whether gapless works too.

[rubdos]

It plays! I'll have to test whether gapless works too.

Looks like this works quite well, have been listening to Tool and Apocalyptica for a few hours now.

[rubdos]

Looks like this works quite well, have been listening to Tool and Apocalyptica for a few hours now.

It skipped from 'Parabol' to 'Ticks & Leeches', that was very weird. Not sure whether that's only Opus, or whether that'd happen with other formats too. Any reports of that happening before in gapless mode?

I am so looking forward to finally have gapless :D

[rubdos]

I'm starting to think that the "about-to-finish" signal gets fired once too often.

g_signal_connect(pipeline, "about-to-finish",
        G_CALLBACK(prepare_next_stream), this);

The bug shows up to me when I'm playing Opus files from start to end; when I just scroll to the end of the song, the gapless transition is perfect, but when I listen to the whole song, it skips the next one.

[Olf0]

when I listen to the whole [Opus encoded] song, it skips the next one.

Any reports of that happening before in gapless mode?

"Not that I am aware of", but this statement is mostly equivalent to "I have no idea". Thus, please try other formats, e.g. classic .mp3.

[rubdos]

Thus, please try other formats, e.g. classic .mp3.

Uuuh, then I need to get my hands dirty and get some of those albums in MP3 for testing purposes. Will do.

[Olf0]

Uuuh, then I need to get my hands dirty …

Sorry, but occasionally this is a natural consequence of being a purist. 😉

[rubdos]

Good that I kept the source FLAC files :-)

[rubdos]

Ugh I can't distinguish which ones are Opus and which ones are FLAC 🤣

Test in progress, feedback in 6 minutes. Those Tool songs are looooong.

[rubdos]

Test in progress, feedback in 6 minutes. Those Tool songs are looooong.

FLAC worked :-/

[Olf0]

FLAC worked :-/

I was suggesting to also test .mp3, if you have access to some album encoded as .mp3, because that is what most people are still using.
(Superfluous side note: Fraunhofer's patents expired long ago, the drawback which remains is that it is > 30 years old technology. Hence IMO it is not that bad, only a bit space consuming for decent quality relative to modern codecs.)

If it really looks like only OPUS is affected by the double "about-to-finish" signal, this is unfortunate for you personally, but would not affect many. Consequently I would gladly review and merge this, because it makes OPUS encoded audio playable by FlowPlayer for the first time.

Then you can take your time to file a bug report to whatever component you deem to be responsible for this mishap.

P.S.: And whatever causes the double "about-to-finish" signal, it is completely unrelated to this changeset, any way.

[rubdos]

I was suggesting to also test .mp3, if you have access to some album encoded as .mp3, because that is what most people are still using.

I can definitely still do that, if you think it's still relevant!

(Superfluous side note: Fraunhofer's patents expired long ago, the drawback which remains is that it is > 30 years old technology. Hence IMO it is not that bad, only a bit space consuming for decent quality relative to modern codecs.)

Another problem would be that I have to re-transcode about 1TB of FLAC files into MP3, which would more than double the occupied size on my laptop and SD-card, giving in to a slight quality loss (arguably imperceptible to me with modern Lame). This kind of reasoning is kinda chicken-and-egg: if players don't support Opus, there will never be an incentive to move on. But you're definitely right in principle!

Don't take that as a personal attack though. Just stating my situation, and my opinion on the matter.

If it really looks like only OPUS is affected by the double "about-to-finish" signal, this is unfortunate for you personally, but would not affect many. Consequently I would gladly review and merge this, because it makes OPUS encoded audio playable by FlowPlayer for the first time.

Yeh, I think that's fine by me!

Then you can take your time to file a bug report to whatever component you deem to be responsible for this mishap.

Do you have any idea what it could be? The "about-to-finish" signal seems to be rather the default for implementing gapless through gstreamer, so I imagine this issue exists elsewhere too.

P.S.: And whatever causes the double "about-to-finish" signal, it is completely unrelated to this changeset, any way.

I'd think so too...

[Olf0]

I was suggesting to also test .mp3, if you have access to some album encoded as .mp3, because that is what most people are still using.

I can definitely still do that, if you think it's still relevant!

I would appreciate that, because I think it is potentially relevant for most FlowPlayer users, but unfortunately many do not mind or are used to playing albums with gaps between songs (one cannot miss something which one is not aware of). Note that I do not use FlowPlayer (I do not listen to music when on the move), I only inherited it from CepiPerez when I offered to maintain FileCase; later I decided to apply the infrastructure changes I made for FileCase also to FlowPlayer, because I thought people may be interested in it: It turned out to cause greater resonance than FileCase to my surprise.

Another problem would be that I have to re-transcode about 1TB of FLAC files into MP3, …

No, IMO you only have to transcode two consecutive songs or a single album, and only temporarily, as you can delete the transcoded MP3-files after the test. 😜
_{I am quite sure that is not what you meant to address, but it was easy to misread it in this way.}

Then you can take your time to file a bug report to whatever component you deem to be responsible for this mishap.

Do you have any idea what it could be?

No, "audio on Linux" always has been and appears to continue to be a hell: I stopped to care about the technical details after ~ 10 years, i.e. before pulseaudio was implemented. Since then, either my Linux distributor manages to make audio work or some things simply don't: I care about too many things, in some areas I adopted the "some update may fix it (or not)"-strategy of proprietary software users.

The "about-to-finish" signal seems to be rather the default for implementing gapless through gstreamer, so I imagine this issue exists elsewhere too.

Likely: Most problems occur for many and often somebody has a solution. The issue is to discover it, if it was publicly documented at all.

P.S.: Is this PR ready to be reviewed and merged from your point of view, or do you intend to keep working on it or review it by yourself, first?

[rubdos]

No, IMO you only have to transcode two consecutive songs or a single album, and only temporarily, as you can delete the transcoded MP3-files after the test. 😜 I am quite sure that is not what you meant to address, but it was easy to misread it in this way.

lol. I'll have a listen to an MP3 V0 transcode of 10000 days, and make a bug report after that :-)

I now also wonder whether the Opus playback skips with gapless disabled. Never tested that, actually. Maybe I'm blaming the wrong part of the code.

No, "audio on Linux" always has been and appears to continue to be a hell: I stopped to care about the technical details after ~ 10 years, i.e. before pulseaudio was implemented. Since then, either my Linux distributor manages to make audio work or some things simply don't: I care about too many things, in some areas I adopted the "some update may fix it (or not)"-strategy of proprietary software users.

Completely understand that :)

P.S.: Is this PR ready to be reviewed and merged from your point of view, or do you intend to keep working on it or review it by yourself, first?

No I think it can be merged, at the mercy of your review now!

[rubdos]

lol. I'll have a listen to an MP3 V0 transcode of 10000 days, and make a bug report after that :-)

MP3 played flawlessly :)

[Olf0]

I now also wonder whether the Opus playback skips with gapless disabled. Never tested that, actually. Maybe I'm blaming the wrong part of the code.

Well, does it?

MP3 played flawlessly :)

Thank you! I was afraid something was broken by the recent (2023 / 2024) changes.

No I think it can be merged, at the mercy of your review now!

It is always "fun" for me to review code in languages of which I just know the basics, hence usually failing to get the "big picture". But often (e.g. here) I understand the changes made and they are looking good. And I am quite sure that no extant functionality may become broken.

[Olf0]

LGTM

[Olf0]

I now also wonder whether the Opus playback skips with gapless disabled. Never tested that, actually. Maybe I'm blaming the wrong part of the code.

Well, does it?

Still curious … if you find the time.

[rubdos]

I now also wonder whether the Opus playback skips with gapless disabled. Never tested that, actually. Maybe I'm blaming the wrong part of the code.

Well, does it?

Still curious … if you find the time.

I'm not closing this tab before I figure it out. Latest finding was that gapless-disabled didn't play the next song at all in my car... but I hadn't used Bluetooth in my car with flowplayer yet at all!

[rubdos]

I hadn't used Bluetooth in my car with flowplayer yet at all!

Turns out I just had some spurious unrelated process playing with MPRIS, so playing gapless-disabled in the car was fine on the last trip!

[Olf0]

Thank you. From the pieces of information you evaluated, I created issue #74 "FlowPlayer skips Opus-encoded audio tracks when gap-less playback is enabled", so users can find this issue documented when observing it.

Hence I only captured the user-facing aspects: If I captured any aspect incorrectly or missed an aspect, please denote that in a comment there.