Merge pull request 'The 272 ms audio buffer was legal: the PipeWire callback filled the buffer ceiling, not the graph's request' (#106) from fix/pw-playback-requested into main

Reviewed-on: unom/punktfunk#106

crates/pf-client-core/Cargo.toml

@@ -118,7 +118,11 @@ rand = "0.9"
 # need the hidapi driver). Linux links the system SDL3; Windows builds it from source
 # (no system SDL3 there — same choice as clients/windows).
 [target.'cfg(target_os = "linux")'.dependencies]
-pipewire = "0.9"
+# `v0_3_49` for `Buffer::requested` — the graph's per-cycle frame ask, without which the
+# playback callback can only size writes from the buffer CEILING (quantum-limit, ~170 ms).
+# Pure cfg gate; needs libpipewire ≥ 0.3.49 (2022-03) at runtime, which every ship target
+# (SteamOS, flatpak runtimes, Arch, Ubuntu ≥ 22.10) clears.
+pipewire = { version = "0.9", features = ["v0_3_49"] }
 sdl3 = { version = "0.18", features = ["hidapi"] }
 # Native VAAPI decode (M6 of the native-decode program): the hand-declared libva buffer
 # layouts, the profile/format/surface decisions, the AuPlan → picparams/IQ/slice

crates/pf-client-core/src/audio.rs

@@ -274,14 +274,41 @@ fn pw_thread(
                     chunk.clear();
                     let _ = ud.recycle.try_send(chunk);
                 }
+                // The graph asks for `requested` frames this cycle (one quantum, after
+                // rate-matching); the mapped buffer is sized for the WORST case — PipeWire's
+                // `quantum-limit`, 8192 frames ≈ 170 ms — not for this cycle. Filling to
+                // capacity queued ~170 ms per buffer downstream of the ring and, worse, taught
+                // the jitter policy that the device drains 170 ms per callback, which lifted
+                // the underrun floor (`want` + one frame) above any depth the A/V sync loop is
+                // allowed to ask for: audio sat a stable ~270 ms late and, by the continuity
+                // rule, sync was FORBIDDEN from draining it. Capacity is only the ceiling;
+                // `requested == 0` (no adapter suggestion) falls back to it.
+                let requested = usize::try_from(buffer.requested()).unwrap_or(0);
                 let stride = 4 * ud.channels; // F32LE interleaved
                 let datas = buffer.datas_mut();
                 if datas.is_empty() {
                     return;
                 }
                 let data = &mut datas[0];
-                let want_frames = data.data().map(|s| s.len() / stride).unwrap_or(0);
+                let max_frames = data.data().map(|s| s.len() / stride).unwrap_or(0);
+                let want_frames = if requested > 0 {
+                    requested.min(max_frames)
+                } else {
+                    max_frames
+                };
                 let want = want_frames * ud.channels;
+                // Once per stream, in the shape of the host's per-capture-open quantum log:
+                // whether the graph's request or the buffer ceiling is sizing our writes is
+                // exactly what an on-glass latency report needs to say.
+                if ud.callbacks == 0 {
+                    tracing::info!(
+                        requested_frames = requested,
+                        capacity_frames = max_frames,
+                        write_frames = want_frames,
+                        write_ms = want_frames / 48,
+                        "audio playback quantum"
+                    );
+                }
 
                 // A/V sync: take whatever depth the decode thread's sync loop last asked for, and
                 // publish where the ring actually is so it can measure the result. The policy