Open source Loom alternative. Beautiful, shareable screen recordings.
Rust appcapcossloommacnextjsnextjs15open-sourceossreactrecordrustscreen-capturescreen-recorderscreenshotsolidjstauritauri-apptypescriptvite
test(verify): expect the offline export frame reorder under load on Windows
Richie McIlroyd08a5f8300c0
diff --git a/crates/verify/expectations.toml b/crates/verify/expectations.tomlindex 2bb3ab2fd..796aceabe 100644--- a/crates/verify/expectations.toml+++ b/crates/verify/expectations.toml@@ -134,6 +134,14 @@ link = ""# (recording track); the product anchors the Instant audio track at the# recording epoch on Linux only.+[[xfail]]+case = "export/*"+class = "timestamps"+platforms = ["windows"]+flaky = true+reason = "NEW: under load an offline export reorders frames. A screen decode that misses its 2,000 ms deadline (crates/rendering/src/decoder/mod.rs, request_frame returns None) makes the render loop emit a fallback frame (crates/rendering/src/lib.rs, the decode-failure fallback) ahead of the frame still waiting for its GPU readback, so the output order becomes 32, 34, 33 and the encoder's monotonic pts repair (EncoderBase::update_pts via queue_frame_reusable, crates/enc-ffmpeg/src/video/h264.rs) shifts every later frame by one slot: frames 33 onward sit off the 60 fps grid while the count stays right. Reproduced deterministically by delaying the two screen responses at 34/60 past the deadline (sol-starved-export.md). Seen on a starved windows-2022 runner (Engine verify 38818870844, windows shard 4, export/basic/camera-hidden/mp460/ffmpeg); the code is shared, but no macOS or Linux run has missed a decode deadline. Fix branch fix/export-frame-order-under-load (frames are retried in order, never substituted)"+link = ""+[[xfail]]case = "export/instant*/output/mp4/native"class = "timestamps"@@ -147,8 +155,8 @@ case = "export/instant-ungated-system-continuous-mic-late/output/mp4/native"class = "sync"platforms = ["windows"]flaky = true-reason = "NEW: the recorder deletes real audio samples when audio delivery latency falls. AudioGapTracker::detect_overlap (crates/recording/src/output_pipeline/core.rs, called from process_audio_frame) compares a frame's capture time against a wall-clock delta, so a delivery catch-up reads as overlap and the frame is trimmed; every later sample plays early. The saved Windows outputs lost exactly 279, 250 and 842 samples (5.8, 5.2 and 17.5 ms; mic and system shifted together, one step), reproduced deterministically by feeding those delivery schedules to the real function (sol-windows-audio-offset.md). Seen 3 times on windows-2022 (38764978019, 38780016448, 38804993000), about 1 in 35 recordings, and 0 of 102 in probes 38782202975 and 38785649744; never on macOS (30 recordings, -0.3 to +2.9 ms) or Linux (16, 0.0 ms), whose delivery in this harness does not catch up the same way. A fix PR is being built"-link = ""+reason = "NEW: the recorder deletes real audio samples when audio delivery latency falls. AudioGapTracker::detect_overlap (crates/recording/src/output_pipeline/core.rs, called from process_audio_frame) compares a frame's capture time against a wall-clock delta, so a delivery catch-up reads as overlap and the frame is trimmed; every later sample plays early. The saved Windows outputs lost exactly 279, 250 and 842 samples (5.8, 5.2 and 17.5 ms; mic and system shifted together, one step), reproduced deterministically by feeding those delivery schedules to the real function (sol-windows-audio-offset.md). Seen 3 times on windows-2022 (38764978019, 38780016448, 38804993000), about 1 in 35 recordings, and 0 of 102 in probes 38782202975 and 38785649744; never on macOS (30 recordings, -0.3 to +2.9 ms) or Linux (16, 0.0 ms), whose delivery in this harness does not catch up the same way. Fixed by #2635"+link = "https://github.com/CapSoftware/Cap/pull/2635"[[xfail]]case = "export/instant-ungated-mic-late/output/mp4/native"