Repository navigation
Open
5 / 85 of 8 issues completedLabels
bugSomething isn't workingSomething isn't workinginternalTooling, CI, tests and code health; no user-facing changeTooling, CI, tests and code health; no user-facing change
Description
Goal
The host test suites pass without reruns. Each flake is fixed at its root in the test rig, not hidden with a retry, a sleep or #[ignore].
Reports
Grouped so the pieces stay together. The causes differ; each report stays open until its own cause is fixed or shown not to recur.
- Late-audio tests flake: the fixture script's summary print, and a racing queue sampler #982 Late-audio tests: the fixture script's summary print, and a racing queue sampler. Both reported failures are already fixed on main (CI flake: tests/mkv_audio_lead.py crashes on a CRC-32 that parses as a huge TrackNumber #1123 by tests/mkv_audio_lead.py: read only TrackEntry children of Tracks, so ffmpeg's CRC-32 cannot crash the fixture build #1124, CI flake: late_audio_tests lane sampler can finish without taking a single sample #1125 by late_audio_tests: the lane sampler always takes a sample after the demuxer ends, so a fast demux cannot read 0 #1126); the draft Late-audio fixture script: bound integer fields by width on read and the Cluster Timestamp by value on write #1181, which bounds the fixture script's remaining reads and was set to close this issue, was closed as stale on 2026-11-13 and can be reopened. No PR is open for it now.
- ff body-heal test flakes: an unsized body's read returns before the server speaks again #997
ffbody-heal test: an unsized body's read returns before the server speaks again. The test-only draft Grade the bytes, not a clock, in the unsized-body silence test #1180 was closed as stale on 2026-11-13 and can be reopened. No fix has landed. - Flaky hostsim test: an end of stream can still be on the player when the lifecycle test checks the route #1111 Hostsim lifecycle test: an end of stream can still be on the player when the test checks the route. The draft Pin the interleaving behind #1111, and make an end-of-stream test failure say what it saw #1185 (pins the interleaving and makes the failure say what it saw; a diagnosis aid, not a proven fix) was closed as stale on 2026-11-13 and can be reopened. No PR is open for it now.
Done
- Intermittent: library_route_tests crumb-root test failed once under load #1001
library_route_testscrumb-root test failed once under load:World::settlefixed in Host-suite flakes fixed at their roots (test code only) #1189; reopen if it recurs. - Flaky: plx_screens tests that build a screen read the Under posters setting without a lock #1178
plx_screenstests that read the Under posters setting without a lock, and Flaky:app::diagnostics::tests::an_unknown_firmware_is_named_as_unknownreadsResolving#1179 an app diagnostics test that readResolving: the flake each was filed for is fixed in Host-suite flakes fixed at their roots (test code only) #1189. - The
plx_mediahostsim testqueue_predecessor_tests::the_window_reread_waits_for_the_outgoing_stop_within_its_boundfailed in CI on 2026-11-12 and 2026-11-14 (waited 249 ms against a 250 ms bound): its clock started inside the reader thread, after the stop's sleep had begun. Fixed in Media: the window-reread test times its wait from before the stop's sleep (flaky on loaded runners) #1382 (2026-11-14; backported torelease/v0.10in [v0.10] Media: the window-reread test times its wait from before the stop's sleep (flaky on loaded runners) #1467, 2026-11-19); no issue was filed.
Metadata
Metadata
Assignees
Labels
bugSomething isn't workingSomething isn't workinginternalTooling, CI, tests and code health; no user-facing changeTooling, CI, tests and code health; no user-facing change
Activity
app::diagnostics::tests::an_unknown_firmware_is_named_as_unknownreadsResolving#1179