Skip to content
5 changes: 5 additions & 0 deletions quest/m0/test-flakes-2/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -199,10 +199,15 @@ Added 2026-11-07 (maintainer, /quest-complete): moq-uring's qlog accounting
test, which spins a fixed count waiting on a background thread, from
[moq-dev/moq#8208](https://github.com/moq-dev/moq/pull/8208)'s follow-ups.

Added 2026-11-08 (maintainer, /quest-complete): `js/net`'s widened max-age
re-drop test, which failed on [moq-dev/moq#8230](https://github.com/moq-dev/moq/pull/8230)
with the `Cancel` reset [moq-dev/moq#8195](https://github.com/moq-dev/moq/pull/8195) intends for a reply still pending at the stream's end; the test races its finish against that reply.

Public API: none. Wire: none.

## Required

- [Fallback playback](/quest/m0/test-flakes-2/ietf22-fallback.md) - a browser viewer follows a fallback within the media lane's bound without dropping its session
- [moq-relay's forged-dial test waits for bob's endpoint](/quest/m0/test-flakes-2/forged-dial-announce.md) - `a_forged_dial_never_connects` requests bob's endpoint only after its announce reaches mallory, and its `#[ignore]` is removed
- [moq-uring's qlog accounting test waits on the accounting](/quest/m0/test-flakes-2/qlog-staging-spin.md) - `an_open_trace_accounts_for_staging_and_close` waits for the accounting, not 10,000 spins
- [js/net's widened max-age re-drop test ends cleanly, not with a reset](/quest/m0/test-flakes-2/widened-max-age-reset.md) - the test finishes the track only after the re-drop reply arrives, so the end is a FIN, and drops its wall-clock assert
15 changes: 10 additions & 5 deletions quest/m0/test-flakes-2/qlog-staging-spin.md
Original file line number Diff line number Diff line change
Expand Up @@ -12,10 +12,15 @@ Planned 2026-11-07 (maintainer, /quest-complete), a follow-up of
[moq-dev/moq#8208](https://github.com/moq-dev/moq/pull/8208).

The test (`rs/moq-uring/src/quic/qlog.rs`) spins up to 10,000
`std::thread::yield_now` calls until `sink.inner.memory` reaches the
expected value, which a background thread sets. On a loaded host the thread
can run late and the spins run out first. Wait on the accounting itself, for
example a condition the sink signals when it charges memory, or a blocking
read of the staged state. Don't raise the spin count.
`std::thread::yield_now` calls until `sink.inner.memory` drops to
`CHUNK + size_of::<Msg>()`. `Inner::reserve` charges synchronously; the drop
comes from the writer thread releasing the `Msg::Open` charge
(`memory.fetch_sub`) once it handles it. On a loaded host the writer can run
late and the spins run out first. Wait for that release itself, for example
a signal the writer raises after it handles a message. Don't raise the spin
count.

It has not failed in CI yet, so it stays enabled; quarantine it with
`#[ignore]` and a `sh/gh/test-inventory.allow` entry if it does.

Public API: none. Wire: none.
35 changes: 35 additions & 0 deletions quest/m0/test-flakes-2/widened-max-age-reset.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,35 @@
# [XS] js/net's widened max-age re-drop test ends cleanly, not with a reset

## Goal

`js/net`'s "an update keeping the budget drops the stale groups again on
lite-07" cannot flake. It finishes the track only after the re-drop reply has
reached the subscriber, so the end arrives as a FIN, and it detects a missing
re-drop without a wall-clock deadline.

## Plan

Planned 2026-11-08 (maintainer, /quest-complete), a follow-up of
[moq-dev/moq#8230](https://github.com/moq-dev/moq/pull/8230), whose iterate
saw it fail there and on unrelated branches, then pass on rerun.

The observed failure (8230's `Check`, job 128649627407) took 9.30 ms:
`recvGroup()` rejected with `reset stream: 1` (`StreamCode.Cancel`), thrown
from `js/net/src/lite/publisher.ts`'s `response()`. When the stream ends
while a reply write is blocked on the write budget, `response()` resets the
writable half with `Cancel`. Here the reader's priority-only update sends a
reply while the producer finishes, so the clean end can lose to the reset.
The test's 500 ms elapsed assert never ran.

The reset is intended:
[moq-dev/moq#8195](https://github.com/moq-dev/moq/pull/8195) rejected letting a
clean FIN override a pending reply write, since that write could block on flow
control for a peer that has left. So the race is in the test, which finishes
the producer right after `reader.update()` without waiting for the re-drop
reply. Fix the test, not the publisher: wait until the subscriber has received
the re-drop of groups 1 and 2 under the new update, then finish the producer.
That wait also replaces the 500 ms elapsed assert. Bound it on the `Manual`
clock, or by event, so a missing re-drop (which would otherwise wait out the
1.5 s grace) still fails while a CI pause cannot.

Public API: none. Wire: none.
3 changes: 2 additions & 1 deletion rs/moq-vaapi/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -69,7 +69,8 @@ It lived in [moq-dev/vaapi](https://github.com/moq-dev/vaapi) through 0.2.1.
- **Loads system libva at runtime on Linux.** The generated `Va` function table
loads `libva` / `libva-drm` with `libloading`. Building needs no system libva,
and a binary can start without it; opening a display reports a missing driver
or library. The crate also compiles on macOS for development.
or library. Off Linux the crate compiles empty (`#![cfg(target_os = "linux")]`),
so its tests run only on Linux.

## Upstream

Expand Down
Loading