mpd 0.24.13 > mpd.conf > audio_output > type: shout > proto: icecast2 > opustags: yes => causing icecast 2.5.0 to crash?

[iconoclasthero]

Interleaving the metadata of the audio stream into the opus stream sent to icecast seems like it would be preferable, but it's causing icecast 2.5.0 (bare-metal & docker) to seg fault at the song boundary.

Has anyone found a way to make this work? Is there a config setting on mpd or icecast that needs to be set?
mpd.conf:

audio_output {
  name          "Icecast OPUS:  96 kbps"
  enabled       "no"
  type          "shout"
  encoding      "opus"            # optional
  host          "localhost"
  port          "8000"
  mount         "/mpd96.opus"
  user          "source"
  bitrate       "96000"           # do not define if quality is defined
  format        "48000:16:2"
  protocol      "icecast2"        # default is icecast2
  opustags      "no"              # causes icecast to crash if enabled ("yes")
}

I'm not going to post the whole icecast.xml because of the size, nor any portion thereof because I have no idea what would be relevant however, the entire mount section is commented out.

I'm also seeing this setting metadata issue with the icecast ouput enabled, interestingly though, it only occurs at natural song boundaries, it doesn't occur seem to occur (n=2 observed) when mpc next is called as shown:

2026-08-16T15:35:15 player: played "Tom Petty/Tom Petty and the Heartbreakers -- Live at the Fillmore, 1997 (2022)/Tom Petty and the Heartbreakers -- 03-04 - Drug Store Truck Drivin' Man.flac"
2026-08-16T15:35:15 shout_output: error setting shout metadata
2026-08-16T15:40:40 alsa_output: Decoder is too slow; playing silence to avoid xrun
2026-08-16T15:40:40 player: played "Tom Petty/Tom Petty and the Heartbreakers -- Live at the Fillmore, 1997 (2022)/Tom Petty and the Heartbreakers -- 03-05 - Eight Miles High.flac"
2026-08-16T15:40:40 shout_output: error setting shout metadata
2026-08-16T15:45:14 player: played "Tom Petty/Tom Petty and the Heartbreakers -- Live at the Fillmore, 1997 (2022)/Tom Petty and the Heartbreakers -- 03-06 - Crazy Mama.flac"
2026-08-16T15:45:14 shout_output: error setting shout metadata
2026-08-16T15:48:40 alsa_output: Decoder is too slow; playing silence to avoid xrun
2026-08-16T15:48:40 player: played "Delaney & Bonnie/Delaney & Bonnie and Friends -- Motel Shot (1971)/Delaney & Bonnie and Friends -- 01-04 - Long Road Ahead.flac"
2026-08-16T15:48:40 shout_output: error setting shout metadata
2026-08-16T15:49:05 player: skipped "Delaney & Bonnie/Delaney & Bonnie and Friends -- Motel Shot (1971)/Delaney & Bonnie and Friends -- 01-05 - Faded Love.flac"
2026-08-16T15:50:16 player: skipped "Delaney & Bonnie/Delaney & Bonnie and Friends -- Motel Shot (1971)/Delaney & Bonnie and Friends -- 01-06 - Talkin' About Jesus.flac"
2026-08-16T15:53:00 player: played "Delaney & Bonnie/Delaney & Bonnie and Friends -- Motel Shot (1971)/Delaney & Bonnie and Friends -- 01-07 - Come On in My Kitchen.flac"
2026-08-16T15:53:00 shout_output: error setting shout metadata

[gbcox]

Seems to me that the segfault is an Icecast bug. But what feature of Icecast do you actually need here that MPD's httpd output doesn't provide? MPD can serve the Opus stream directly, including OpusTags and multiple clients, so unless you need some Icecast-specific functionality, it seems like an unnecessary extra layer.

[iconoclasthero]

I'm trying to report the icecast issue, but their gitlab seems to be hosted on a 386 machine with 512 kb of RAM and a 14.4k dialup. I posted here because I figured that this would also be a nexus point for people using mpd and icecast. I'll take the httpd stream suggestion under consideration as a possible alternative to icecast.

[MaxKellermann]

Don't post Icecast bugs here, that's pointless. Maybe MPD is misbehaving somehow, but it's too difficult to analyze that as long as Icecast crashes. First get the Icecast bug fixed.

[iconoclasthero]

I'm trying to find out if other mpd users in this community have gotten this working or if they're experiencing the same issue. That's the point of the discussion.

[MaxKellermann]

That discussion is pointless, because this isn't a social problem, but a technical problem, and needs to be solved by fixing the bug.
But now I remember your name from IRC, and I remember very well that you're that one guy who'd never listen to me. Why am I even replying?

[iconoclasthero]

https://docs.github.com/en/discussions

GitHub Discussions documentation

GitHub Discussions is a collaborative communication forum for the community around an open source or internal project. Community members can ask and answer questions, share updates, have open-ended conversations, and follow along on decisions affecting the community's way of working.

I feel like I'm looking to gather some information here about a particular issue related to mpd and a particular audio output. I don't see why you need to bring all this hostility into what should be a "collaborative communication forum." Perhaps someone has already solved this problem, perhaps not. I'm not posting this as an mpd issue that needs to be solved so there's really no concern from that standpoint. Thanks.