request.dynamic replays same song

[ScuttleSE]

I have a fairly simple setup. I want Liquidsoap to run a pythonscript to get what file to play, then when that song is over, execute the script again and play the song the script outputs.

The song plays fine the first time, but when it is over, it just repeats.

Here is the .liq file I'm running

def get_request() =
  result = get_process_lines("python3 /home/liq/getsong.py")
  log("next song: #{result}")
  request.create(list.hd(default="", result))
end

gbsfm = request.dynamic(id="gbsfm", get_request)

output.icecast(
    %mp3(bitrate=96),
    fallible = true,
    host = "localhost",
    port = 8000,
    password = "sourcepassword",
    mount = "96", gbsfm)

In the logfile, I see that the script is being executed and it gets a new song, but the old one just replays.

2020/10/09 14:44:07 [128:3] Connection setup was successful.
2020/10/09 14:44:07 [decoder:3] Method "FFMPEG" accepted "/srv/music/music4/6/6fa53f0680b7fca4d9df2e48fb60d3ab85ed54e0.mp3".
2020/10/09 14:44:07 [96:3] Connecting mount 96 for source@localhost...
2020/10/09 14:44:07 [96:3] Connection setup was successful.
2020/10/09 14:50:02 [lang:3] next song: ["/srv/music/music4/8/86a1a7fc75b3137f102a2b9cfc5480e694913332.mp3"]
2020/10/09 14:50:02 [decoder:3] Method "FFMPEG" accepted "/srv/music/music4/8/86a1a7fc75b3137f102a2b9cfc5480e694913332.mp3".
2020/10/09 14:50:12 [gbsfm:3] Prepared "/srv/music/music4/6/6fa53f0680b7fca4d9df2e48fb60d3ab85ed54e0.mp3" (RID 1).

As you can see, it first plays the song starting with 6..., then it finds the one starting with 8..., but it plays 6... instead.

[ScuttleSE]

After some more digging I have discovered that liquidsoap seems to do multiple requests to the getsong.py script when starting up.
It looks like it first makes a request to get the song it is supposed to play, it then makes another request to get the next song to play?

Is there any way to have it only do the one request? And then one request when the current song is over.

[add-n2x]

Am I right, that this was resolved by updating to Liquidsoap 1.4.3?

[ScuttleSE]

Sorry, yes, running 1.4.3 seems to have helped.

[ScuttleSE]

Reopening this since the behavior has reappeared in 2.0.1

Setting prefetch=0 does nothing, it actually makes things worse by just skipping that source completely.

[toots]

Thanks. Do you have any logs and/or a script to reproduce the issue?

[CatVahe]

Hi, I have the same problem, when "prefetch" = 0 silence plays, when "prefetch" = 1 2 songs are loaded

script

#!/usr/bin/liquidsoap

%include "../data.cf"

#for print log
settings.log.stdout.set(true)
settings.log.level.set(3)
settings.log.file.path.set("#{main_dir}/#{station}/#{logs}/<script>.log")
def get_playlist () =
	result = process.read("php  #{main_dir}/#{station}/#{gettrack}")
	log("next song: #{result}")
	request.create(result)
end

radio = request.dynamic(prefetch = 0, get_playlist)

#for sound
radio = crossfade(fade_out=3., fade_in=3., duration=5., radio)
radio = normalize(radio)
radio = mksafe(radio)

output.icecast(%mp3,
  host = "my.host",
  port = 8443,
  user = "#{user}",
  password = "#{pass}",
  mount = "#{station}_320",
  radio
)

logs

2022/10/20 11:20:42 >>> LOG START
2022/10/20 11:20:40 [main:3] Liquidsoap 2.1.2
2022/10/20 11:20:40 [main:3] Using: bytes=[distributed with OCaml 4.02 or above] pcre=7.5.0 sedlex=3.0 menhirLib=20220210 curl=0.9.2 uri=4.2.0 dtools=0.4.4 duppy=0.9.2 cry=0.6.7 mm=0.8.1 mad=0.5.2 dynlink=[distributed with Ocaml] lame=0.3.6 ffmpeg=1.1.6 samplerate=0.1.6 taglib=0.3.10 ssl=0.5.12 camomile=1.0.2
2022/10/20 11:20:40 [dynamic.loader:3] Could not find dynamic module for fdkaac encoder.
2022/10/20 11:20:40 [clock:3] Using builtin (low-precision) implementation for latency control
2022/10/20 11:20:42 [frame:3] Using 44100Hz audio, 25Hz video, 44100Hz main.
2022/10/20 11:20:42 [frame:3] Video frame size set to: 1280x720
2022/10/20 11:20:42 [frame:3] Frame size must be a multiple of 1764 ticks = 1764 audio samples = 1 video samples.
2022/10/20 11:20:42 [frame:3] Targeting 'frame.duration': 0.04s = 1764 audio samples = 1764 ticks.
2022/10/20 11:20:42 [frame:3] Frames last 0.04s = 1764 audio samples = 1 video samples = 1764 ticks.
2022/10/20 11:20:42 [sandbox:3] Sandboxing disabled
2022/10/20 11:20:42 [video.converter:3] Using preferred video converter: ffmpeg.
2022/10/20 11:20:42 [audio.converter:3] Using samplerate converter: libsamplerate.
2022/10/20 11:20:42 [clock.main:3] Streaming loop starts in auto-sync mode
2022/10/20 11:20:42 [clock.main:3] Delegating synchronisation to CPU clock
2022/10/20 11:20:42 [video.text:3] Using native implementation
2022/10/20 11:20:42 [mksafe:3] Switch to safe_blank.
2022/10/20 11:20:42 [416b9d94-4d49-11ed-af4b-93b71599bd52_320:3] Connecting mount 416b9d94-4d49-11ed-af4b-93b71599bd52_320 for user@localhost...
2022/10/20 11:20:42 [416b9d94-4d49-11ed-af4b-93b71599bd52_320:3] Connection setup was successful.

script

#!/usr/bin/liquidsoap

%include "../data.cf"

#for print log
settings.log.stdout.set(true)
settings.log.level.set(3)
settings.log.file.path.set("#{main_dir}/#{station}/#{logs}/<script>.log")
def get_playlist () =
	result = process.read("php  #{main_dir}/#{station}/#{gettrack}")
	log("next song: #{result}")
	request.create(result)
end

radio = request.dynamic(prefetch = 1, get_playlist)

#for sound
radio = crossfade(fade_out=3., fade_in=3., duration=5., radio)
radio = normalize(radio)
radio = mksafe(radio)

output.icecast(%mp3,
  host = "my.host",
  port = 8443,
  user = "#{user}",
  password = "#{pass}",
  mount = "#{station}_320",
  radio
)

logs

2022/10/20 11:30:35 >>> LOG START
2022/10/20 11:30:33 [main:3] Liquidsoap 2.1.2
2022/10/20 11:30:33 [main:3] Using: bytes=[distributed with OCaml 4.02 or above] pcre=7.5.0 sedlex=3.0 menhirLib=20220210 curl=0.9.2 uri=4.2.0 dtools=0.4.4 duppy=0.9.2 cry=0.6.7 mm=0.8.1 mad=0.5.2 dynlink=[distributed with Ocaml] lame=0.3.6 ffmpeg=1.1.6 samplerate=0.1.6 taglib=0.3.10 ssl=0.5.12 camomile=1.0.2
2022/10/20 11:30:33 [dynamic.loader:3] Could not find dynamic module for fdkaac encoder.
2022/10/20 11:30:33 [clock:3] Using builtin (low-precision) implementation for latency control
2022/10/20 11:30:35 [frame:3] Using 44100Hz audio, 25Hz video, 44100Hz main.
2022/10/20 11:30:35 [frame:3] Video frame size set to: 1280x720
2022/10/20 11:30:35 [frame:3] Frame size must be a multiple of 1764 ticks = 1764 audio samples = 1 video samples.
2022/10/20 11:30:35 [frame:3] Targeting 'frame.duration': 0.04s = 1764 audio samples = 1764 ticks.
2022/10/20 11:30:35 [frame:3] Frames last 0.04s = 1764 audio samples = 1 video samples = 1764 ticks.
2022/10/20 11:30:35 [sandbox:3] Sandboxing disabled
2022/10/20 11:30:35 [video.converter:3] Using preferred video converter: ffmpeg.
2022/10/20 11:30:35 [audio.converter:3] Using samplerate converter: libsamplerate.
2022/10/20 11:30:35 [video.text:3] Using native implementation
2022/10/20 11:30:35 [clock.main:3] Streaming loop starts in auto-sync mode
2022/10/20 11:30:35 [clock.main:3] Delegating synchronisation to CPU clock
2022/10/20 11:30:35 [lang:3] next song: https://my.other.host/e0d26fd6-493c-11ed-ae5d-93b71599bd52/a035d2fc-495e-11ed-ae67-93b71599bd52.mp3
2022/10/20 11:30:35 [mksafe:3] Switch to safe_blank.
2022/10/20 11:30:35 [416b9d94-4d49-11ed-af4b-93b71599bd52_320:3] Connecting mount 416b9d94-4d49-11ed-af4b-93b71599bd52_320 for user@host...
2022/10/20 11:30:35 [416b9d94-4d49-11ed-af4b-93b71599bd52_320:3] Connection setup was successful.
2022/10/20 11:30:36 [request.dynamic_0:3] Prepared "/tmp/liq-processaf2afb.mp3" (RID 0).
2022/10/20 11:30:36 [mksafe:3] Switch to amplify_1 with transition.
2022/10/20 11:30:36 [lang:3] next song: https://my.other.host//e0d26fd6-493c-11ed-ae5d-93b71599bd52/a154f6f4-495e-11ed-ae67-93b71599bd52.mp3

[CatVahe]

@toots

[toots]

Hi,

This seems like the default behavior, yes. One song is fetched and starts playing immediately and the next song gets fetched to be queued.

If you need something more finer-grained you can use the fetch method. For instance:

def fetch(s) =
  thread.run(fun () -> begin
    if not s.fetch() then
      # Do something when fetch failed
    end
  end)
end

s = request.dynamic(prefetch=0, ...)

# Initial request
fetch(s)

# Fetch next song 10. before the current one ends
s = source.on_end(delay=10., fun (_, _) -> fetch(s), s)

(code untested)

However, you should be careful that the request resolution succeeds in less than 10. otherwise the source will become unavailable, which is why this is not the default.

[smimram]

Note: you might also want to test native.request.dynamic which is a re-implementation of request.dynamic in liquidsoap, which might suit you (and you can modify it if not!)

[rrolla]

Yeah native.request.dynamic works different, do you have any documentation, how native stuff differs?