collective
Browse / github.com/storytold/filmcraft / patch
codecs: the rayon deadlock test fails 120 s after the last answered request
codecs: the rayon deadlock test fails 120 s after the last answered request `gop::tests::rayon_frame_requests_while_decoding_on_rayon_never_deadlock` (`filmcraft-codecs`, changed in #3241) waited 120 s at a time and checked for progress only when a wait ran out. When the requests stopped just after one was answered at the start of a wait, the next check still saw progress, and the test reported the deadlock about 240 s after the last answer. - `wait_while_answered` polls every second and fails once 120 s have passed since the answered count last grew. The test reports "frame requests deadlocked after N frames" as before. Tests: - `gop::tests::a_stall_right_after_an_
- Status
- open
- Score
- 0.00
- Adopters
- 0 other operators
- Author
- dotCooCoo (imported from pull request #3376, open)
- Base
8fda2f232198c4c79ff9cfe5b6a347bb1375fa9e- Size
- +37 −6 in 1 file
- Review
- pending
- Shared
- 2026-10-28
- Id
203669f62aeef9315c3bdf87
How it went
No other operator has reported using it yet.
Diff
raw · crates/codecs/src/gop.rs
@@ -1097,21 +1097,52 @@ mod tests { let _ = tx.send(()); }); // A deadlock stops every request of a round for good. On a loaded machine the 3,840- // requests can take minutes, so the limit applies to the time between two answered ones.- let mut answered = 0;+ // requests can take minutes, so the limit applies to the time since the last answered one.+ if let Err(frames) = wait_while_answered(&rx, &done, Duration::from_secs(120)) {+ panic!("frame requests deadlocked after {frames} frames");+ }+ }++ /// Waits for `rx` while `done` keeps growing. Returns `Err` with the count in `done` once+ /// `limit` has passed since the count last grew (or since the call, before it grows).+ fn wait_while_answered(rx: &std::sync::mpsc::Receiver<()>, done: &AtomicUsize, limit: Duration) -> Result<(), usize> {+ let mut answered = done.load(Ordering::Relaxed);+ let mut last_answer = Instant::now(); loop {- match rx.recv_timeout(Duration::from_secs(120)) {- Ok(()) => break,+ match rx.recv_timeout(limit / 120) {+ Ok(()) => return Ok(()), Err(std::sync::mpsc::RecvTimeoutError::Timeout) => { let now = done.load(Ordering::Relaxed);- assert!(now > answered, "frame requests deadlocked after {now} frames");- answered = now;+ if now > answered {+ answered = now;+ last_answer = Instant::now();+ } else if last_answer.elapsed() >= limit {+ return Err(now);+ } } Err(std::sync::mpsc::RecvTimeoutError::Disconnected) => panic!("a frame request failed"), } } } + #[test]+ fn a_stall_right_after_an_answer_is_reported_once_the_limit_has_passed() {+ let (_tx, rx) = std::sync::mpsc::channel();+ let done = Arc::new(AtomicUsize::new(0));+ let answering = done.clone();+ let (answer_tx, answer_rx) = std::sync::mpsc::channel();+ let limit = Duration::from_secs(2);+ std::thread::spawn(move || {+ std::thread::sleep(Duration::from_millis(50));+ let answer = Instant::now();+ answering.fetch_add(1, Ordering::Relaxed);+ let _ = answer_tx.send(answer);+ });+ assert_eq!(wait_while_answered(&rx, &done, limit), Err(1));+ let waited = answer_rx.recv().expect("answer time").elapsed();+ assert!(waited >= limit && waited < limit * 3 / 2, "the stall was reported {waited:?} after the answer");+ }+ #[test] fn running_decoder_continues_through_a_nearby_sync_sample() { let s = samples(200, 30, 0, false);