# Real Pulse daemon fixture The M4.4 fixture is opt-in because it connects to the current user's real audio session. Its normal phase creates a uniquely named, silent stereo stream with `paplay`, changes only that stream, and always terminates the child process. It never parses shell output in the production adapter. ## Prerequisites - A reachable PulseAudio-compatible server. This can be PulseAudio itself or PipeWire's `pipewire-pulse` service. - The `paplay` executable and libpulse development/runtime files. On Arch these are supplied by `libpulse`. - Permission to access the user audio socket. Sandboxed test runners must expose that socket deliberately. Run the non-disruptive phase with: ```sh cargo test --locked -p volume_lib --test pulse_daemon --all-features \ real_stream_commands_external_edits_and_churn_converge -- \ --ignored --exact --nocapture ``` It verifies discovery of the isolated stream, worker volume and mute commands, readback, changes made by an independent Pulse client, removal, and a replacement stream with a fresh handle. Existing desktop streams are observed but never selected for writes. The fixture fails rather than silently passing when the daemon, `paplay`, expected notifications, or cleanup behavior is unavailable. Set `PULSE_SERVER` before the command to target an explicitly isolated daemon. This is the intended way to run the same fixture against a standalone PulseAudio CI instance without touching the desktop session. ## Disruptive PipeWire restart phase This phase intentionally restarts the current user's PipeWire, pipewire-pulse, and WirePlumber services. It interrupts all desktop audio and must not run in the ordinary test suite. Close or pause important audio applications first, then acknowledge the interruption explicitly: ```sh FADER_MASTER_RESTART_PIPEWIRE=1 \ cargo test --locked -p volume_lib --test pulse_daemon --all-features \ pipewire_restart_disconnects_and_a_new_epoch_resnapshots -- \ --ignored --exact --nocapture ``` The fixture requires a visible disconnect diagnostic, retries with bounded backoff, reconnects under a new backend epoch, and waits for a complete fresh snapshot. Without the acknowledgement variable it fails before restarting anything.