• Home
  • Features
  • Pricing
  • Docs
  • Announcements
  • Sign In

Eyevinn / hi265 / 41621941041
81%
main: 81%

Build:
Build:
LAST BUILD BRANCH: release/v0.6.2
DEFAULT BRANCH: main
Ran 03 Nov 2026 10:46AM UTC
Jobs 1
Files 40
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst

03 Nov 2026 10:45AM UTC coverage: 81.237% (+0.3%) from 80.925%
41621941041

Pull #21

github

tobbee
fix: refuse quantization parameters outside the range the standard allows

Nothing bounded the slice QP, 26 + init_qp_minus26 + slice_qp_delta: mp4ff stores the PPS fields unchecked, and pkg/decoder used the sum as it came. Spec 7.4.7.1 bounds SliceQpY to -QpBdOffsetY..51. Below zero, a QP that is not a multiple of 6 indexed LevelScale at -1 in transform.Dequantize and panicked; every other out-of-range value decoded without error into a picture no conforming decoder produces (an IDR coded at QP 26 and decoded against a PPS saying -6 or 52 came out flat 128 for a source of 200). FFmpeg refuses these streams. The encoder wrote such a PPS on request: EncodeParams{QP: -1} gave init_qp_minus26 = -27, and GenerateIDR then panicked in the forward quantizer, as did EncodeIDRSliceFromSPSPPS with an external PPS like it.

internal/transform/qprange.go now holds the bounds for both sides: SliceQpY and init_qp_minus26 (7.4.3.3.1, 7.4.7.1), the picture- and slice-level chroma QP offsets and their sums, each within +-12 (7.4.3.3.1, 7.4.7.1), and CuQpDeltaVal (7.4.9.14). The decoder checks the slice QP and chroma offsets in one helper both slice header paths now share, naming every term of the sum since the fault can be in the PPS or the slice, and checks cu_qp_delta where it is read. The generators refuse an EncodeParams.QP outside 0..51 (0 stays the default, 26), and the writers that take an external PPS refuse an init_qp_minus26 out of range at the SPS's bit depth; the residual writers also refuse PPS chroma offsets out of range.

Every new test fails on the previous main, by panicking, by decoding, or by accepting the QP. The P-skip path was refused there too, but only by accident: CABAC initialised at the wrong QP misread the skip CUs, and the error named motion compensation. The roadmap records this as 0.26.

Claude-Session: https://claude.ai/code/session_01EagxH7FZ72UothFqTUDaZj
Pull Request #21: fix: refuse quantization parameters outside the range the standard allows

96 of 101 new or added lines in 5 files covered. (95.05%)

7291 of 8975 relevant lines covered (81.24%)

0.93 hits per line

Uncovered Changes

Lines Coverage ∆ File
3
79.71
0.93% pkg/decoder/decoder.go
2
92.8
-0.15% internal/slice/slicedata.go
Jobs
ID Job ID Ran Files Coverage
1 41621941041.1 03 Nov 2026 10:46AM UTC 40
81.24
GitHub Action Run
Source Files on build 41621941041
  • Tree
  • List 40
  • Changed 5
  • Source Changed 0
  • Coverage Changed 5
Coverage ∆ File Lines Relevant Covered Missed Hits/Line
  • Back to Repo
  • Pull Request #21
  • PR Base - main (#41056290966)
STATUS · Troubleshooting · Open an Issue · Sales · Support · CAREERS · ENTERPRISE · START FREE TRIAL · SCHEDULE DEMO
ANNOUNCEMENTS · TWITTER · TOS & SLA · Supported CI Services · What's a CI service? · Automated Testing

© 2026 Coveralls, Inc