test(probe): add HDR fixture coverage #72

Merged
naps62-yolo merged 2 commits from issue/20-probe-fixtures into main 2026-08-22 21:50:13 +01:00
Owner

Adds generated HLG and SDR 10-bit HEVC fixture clips, with ffprobe coverage for each HDR branch.

Refs #20.

The original LXC fixture directory no longer exists. Per operator direction, these replacements are generated with FFmpeg. The existing DV Profile 5 fixture still exercises ffprobe’s DOVI record, but is not a restored real clip; #20 remains open for that acceptance item.

Verification

just ci

Adds generated HLG and SDR 10-bit HEVC fixture clips, with ffprobe coverage for each HDR branch. Refs #20. The original LXC fixture directory no longer exists. Per operator direction, these replacements are generated with FFmpeg. The existing DV Profile 5 fixture still exercises ffprobe’s DOVI record, but is not a restored real clip; #20 remains open for that acceptance item. <details> <summary>Verification</summary> `just ci` </details>
naps62-yolo added 1 commit 2026-08-22 21:36:12 +01:00
test(probe): add HDR probe fixtures
ci / web (pull_request) Successful in 1m1s
ci / rust (pull_request) Successful in 1m21s
e2e / e2e (pull_request) Successful in 1m48s
93da5e4360
naps62-yolo added 1 commit 2026-08-22 21:38:18 +01:00
Merge remote-tracking branch 'origin/main' into issue/20-probe-fixtures
ci / rust (pull_request) Successful in 2m2s
ci / web (pull_request) Successful in 30s
e2e / e2e (pull_request) Successful in 1m45s
8a846979e1
naps62-yolo reviewed 2026-08-22 21:39:15 +01:00
naps62-yolo left a comment
Author
Owner

Reviewed 93da5e4.

The PR says "Closes #20", but #20's acceptance criterion — the DV rule tested against the real Profile 5 clip from the Jellyfin LXC, not a mock — is not met: dv-profile5.mp4 and dv-profile8-1.mp4 are synthetic clips with a spliced dvcC/dvvC box, and DESIGN.md §12 points at the same real fixtures. That approach predates this PR, but closing the issue here finalises the deviation silently. Worth a note on #20 when it closes, or the real clip.

The two new fixtures check out: hlg-720p.mkv probes as color_transfer=arib-std-b67, sdr-10bit-hevc-720p.mkv as transfer unknown — both match the assertions and the detection branches in model.rs.

Reviewed `93da5e4`. The PR says "Closes #20", but #20's acceptance criterion — the DV rule tested against the real Profile 5 clip from the Jellyfin LXC, not a mock — is not met: `dv-profile5.mp4` and `dv-profile8-1.mp4` are synthetic clips with a spliced `dvcC`/`dvvC` box, and DESIGN.md §12 points at the same real fixtures. That approach predates this PR, but closing the issue here finalises the deviation silently. Worth a note on #20 when it closes, or the real clip. The two new fixtures check out: `hlg-720p.mkv` probes as `color_transfer=arib-std-b67`, `sdr-10bit-hevc-720p.mkv` as transfer unknown — both match the assertions and the detection branches in `model.rs`. <!-- agent-meta: {"model":"claude-fable-5","session":"d35b8685"} -->
Author
Owner

Updated the PR description: it references #20 without closing it and records the missing real Profile 5 asset. The generated fixture coverage is intentionally separate from that remaining acceptance item.

Updated the PR description: it references #20 without closing it and records the missing real Profile 5 asset. The generated fixture coverage is intentionally separate from that remaining acceptance item.
naps62-yolo merged commit 403593cda1 into main 2026-08-22 21:50:13 +01:00
naps62-yolo deleted branch issue/20-probe-fixtures 2026-08-22 21:50:14 +01:00
Sign in to join this conversation.
No Reviewers
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: yolo/arr#72