Skip to content

armgit / Cap

Public Imported from GitHub
0

Open source Loom alternative. Beautiful, shareable screen recordings.

test(verify): expect the decode-deadline losses seen on starved Windows

Richie McIlroy4a209fcffde5
diff --git a/crates/verify/expectations.toml b/crates/verify/expectations.toml
index 6683dfb64..a8016b7da 100644
--- a/crates/verify/expectations.toml
+++ b/crates/verify/expectations.toml
@@ -44,7 +44,7 @@ case = "export/jitter-camera-120s/full/mp460/avf"
class = "nondeterministic"
platforms = ["macos"]
flaky = true
-reason = "Two exports of the 120 s jitter-camera project can differ (4001 of 7200 frames, first at frame 3163, in 1 of 3 runs on blacksmith-6vcpu-macos-15, 38838622145) inside the camera runaway of the future entry above: once the pts-hole map has lost the hold, which frame a request gets depends on how far the AVAssetReader has decoded when the request arrives, so the wrong answers themselves vary (5081 wrong frames in that run, 5146 in the others). The IOSurface part of this nondeterminism is gone with #2625 (0 of 3 sol-avf proof runs, 0 of 5 in probe 38784185050); #2595's decoder changes remove the runaway and with it this class (5 of 5 identical exports in that probe's decoder configuration, 2 of 2 on Apple silicon)"
+reason = "Two exports of the 120 s jitter-camera project can differ (4001 of 7200 frames, first at frame 3163, in 1 of 3 runs on blacksmith-6vcpu-macos-15, 38838622145) inside the camera runaway of the future entry above. The decoder drains requests in batches with try_recv, so a batch's makeup depends on timing, and the reader's pool position is capped to the batch's newest request (update_decoder_position(decoder_idx, capped)); once the pts-hole map has lost the hold, that decides which reader answers, so the wrong answers themselves vary. Measured on blacksmith-6vcpu-macos-15: alone the case is identical in 16 of 16 runs (5146 wrong frames, probe 38843099696); run beside the rest of the export suite it differs in 3 of 8 (5134 and 5201 wrong frames, probe 38844461913). The IOSurface part of this nondeterminism is gone with #2625 (0 of 3 sol-avf proof runs, 0 of 5 in probe 38784185050); #2595's decoder changes remove the runaway and with it this class (5 of 5 identical exports in that probe's decoder configuration, 2 of 2 on Apple silicon)"
link = "https://github.com/CapSoftware/Cap/pull/2595"
[[xfail]]
@@ -144,6 +144,30 @@ 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 = "playback/backward-hold/backward-jumps/seq30-from0/ffmpeg"
+class = "error"
+platforms = ["windows"]
+flaky = true
+reason = "NEW: under load offline playback gets no frame at all: a decode that misses its 2,000 ms deadline (crates/rendering/src/decoder/mod.rs, request_frame returns None) leaves the request unanswered (\"decoders returned no frame for clip 0 at 5.000s\"). Seen once on a starved windows-2022 runner (Engine verify 38841241061, windows shard 1, sentinel longest wake gap 3.1 s, 38 gaps over 100 ms); same deadline as the export frame reorder entry, fix branch fix/export-frame-order-under-load"
+link = ""
+
+[[xfail]]
+case = "export/early-camera/full/mp430/ffmpeg"
+class = "frame-count"
+platforms = ["windows"]
+flaky = true
+reason = "NEW: under load an offline export loses its first frame: when the first decodes miss their 2,000 ms deadline the render loop advances before any usable frame exists (crates/rendering/src/lib.rs, initial decode failure path; sol-starved-export.md), so the export has 179 of 180 frames and every frame shows the source one or two frames ahead. Seen once on a starved windows-2022 runner (Engine verify 38841241061, windows shard 1, sentinel longest wake gap 4.2 s, 133 gaps over 100 ms); fix branch fix/export-frame-order-under-load"
+link = ""
+
+[[xfail]]
+case = "export/early-camera/full/mp430/ffmpeg"
+class = "future"
+platforms = ["windows"]
+flaky = true
+reason = "NEW: same initial decode-deadline loss as the frame-count entry: with the first frame gone every later frame shows the source one or two frames ahead (146 of 178 screen frames). Seen once on a starved windows-2022 runner (Engine verify 38841241061); fix branch fix/export-frame-order-under-load"
+link = ""
+
[[xfail]]
case = "export/instant*/output/mp4/native"
class = "timestamps"