CNA-BUG-142: Without a mixer (SDL2 or NULL audio) songs never end: queued copies carry no Duration, so the elapsed-time fallback never fires

CNA snapshot 009d40f5  ·  Known Issues › Current bugs  ·  source links pinned to 009d40f5

LoadSong copies only a Song's file and name, so every queued Song has a zero Duration and, in builds without SOUND_ENABLED, DetectSongEndedByElapsedTime never reports an end: MediaPlayer stays Playing and the queue never advances.

Identifier

CNA-BUG-142

Category

Bug

Subsystem

Audio & media

Status

Open

Verified against

CNA 009d40f5 (009d40f5dd085c4e674d3479675fac84b12b3e0a)

Severity

Low (a triage suggestion, not a project priority)

Evidence basis

Source-verified: read at TARGET, not executed

Tests touching this area

Yes: see Current tests

Affected contract

MediaPlayer::Update() (via FrameworkDispatcher::Update), MediaState, ActiveSongChanged and queue advance under CNA_AUDIO_PLATFORM=SDL2 or NULL

Expected behaviour

DetectSongEndedByElapsedTime is documented in MediaPlayer.hpp as the song-end detector for builds without a native track-stopped signal, reporting an end whenever the duration is genuinely known. A Song loaded from .xnb carries its duration (Song(path, name, durationMS)), as do library songs whose duration was probed, so their end should be detected and the queue should advance as in XNA and FNA.

Actual behaviour at TARGET

modules/CMakeLists.txt defines SOUND_ENABLED only for SDL3 and ALSA audio. In the other builds Update in MediaPlayer.cpp asks DetectSongEndedByElapsedTime(activeSong, TimerElapsed()), where activeSong is the queue's copy. LoadSong builds that copy with new Song(song->getHandle(), song->getNameProperty()), dropping the duration, and PlaySong sets none without a mixer. The detector ignores a zero duration, so it never fires: the player stays Playing forever, MediaStateChanged and ActiveSongChanged are never raised for a natural end, and repeat, shuffle and queue advance never happen. The fallback is reachable only by calling the static function directly, which is what its tests do.

Source locations

• modules/media/src/Xna/MediaPlayer.cpp — MediaPlayer::LoadSong, PlaySong (non-SOUND_ENABLED branch), Update, DetectSongEndedByElapsedTime
• modules/media/include/Microsoft/Xna/Framework/Media/MediaPlayer.hpp — DetectSongEndedByElapsedTime documentation
• modules/CMakeLists.txt — SOUND_ENABLED only for CNA_AUDIO_PLATFORM SDL3 and ALSA
• modules/media/tests/Microsoft/Xna/Framework/Media/MediaPlayerTests.cpp — MediaPlayerNoSoundFallbackTest cases call the function directly

Evidence

Checked by reading at 009d40f5; not executed. The copy never receives a duration on this path in any configuration, so the conclusion does not depend on timing.

Independently observed as a separate finding (merged): LoadSong queues a copy built from the Song's handle and name only, so it has no Duration, and the elapsed-time end detection used in builds without SOUND_ENABLED never fires.

Focused reproduction

Illustrative; not compiled or run (CNA_AUDIO_PLATFORM=NULL).

Song s("Content/jingle.ogg", "jingle", 500);   // known 0.5 s duration
MediaPlayer::Play(&s);
// ... one second of frames calling FrameworkDispatcher::Update()
MediaPlayer::getStateProperty();   // still MediaState::Playing

Current tests

The four MediaPlayerNoSoundFallbackTest cases in MediaPlayerTests.cpp call DetectSongEndedByElapsedTime with a hand-built Song that has a duration; none plays a Song through Play and Update in a no-mixer build.

Regression test

In a NULL-audio build: Play a Song constructed with a 100 ms duration, advance Update past 100 ms and require Stopped (or the next song). The fix is for LoadSong to copy the duration.

Blast radius

SDL2 and NULL audio selections only, where music is silent anyway: games whose logic waits for a song to end, playlists, repeat and shuffle. SDL3 and ALSA builds end songs through the mixer's stopped callback.

Workaround

End or advance songs from game code with a timer in these profiles.