docs(ledger): narrow the video gap to the CI4 + TLUT texel path

[doublegate]

Two suspects eliminated, the third narrowed to one code path

The previous entry left three suspects for "the RDP rasterises real geometry to
black". Inspecting the RDP state Ocarina actually leaves settles two of them and
turns the third into a specific, testable path.

Z path — REFUTED

z_compare_en = false, z_update_en = false. Nothing is depth-rejected.
SET_Z_IMAGE is issued (z_image = 0x12c700), but the compare is off — so the
"every triangle is the 0x0F Z-buffered variant" reasoning, while true, does not
lead anywhere.

Combiner — REFUTED as a collapse, and it is the key evidence

cyc1 decodes to rgb_a=15, rgb_b=15, rgb_c=31, rgb_d=1:

(0 − 0) × 0 + TEXEL0   →  a pure texture pass-through

The pixel colour is the texel. So a black frame means the texel fetch
resolves to 0 — the combiner is faithfully passing through black.

Texture path — narrowed to CI4 + TLUT

• Tiles are fmt=2, size=0 = CI4 — 4-bit colour-indexed, resolved via a TLUT.
• Both TMEM halves are populated after 4,374 LOAD_BLOCKs: 1,508/2,048
  non-zero bytes in the texture half, 760/2,048 in the TLUT half.
• TLUT entry 0 reads 0x0000 (black), with entries alternating
  0000 ffff 0000 ffff at the 8-byte stride tlut_lookup uses.

So real indices and a real palette are both present, and the resolution yields
black.

Why this is a good stopping point for a docs PR

The gap is now one code path — CI4 index → TLUT lookup — rather than a
subsystem. It is directly testable against the existing Angrylion .rvec
harness: a CI4-with-TLUT vector is the obvious next probe, in the same shape as
the CI/TLUT work already recorded under R-7.

I am deliberately not guessing at a fix in this PR. The last three times I named a
cause before measuring its boundary I was wrong (exception vector, presentation
path, Z), and the pattern each time was reasoning ahead of the oracle. The vector
comes first.

Gates

cargo fmt --all --check, cargo test --workspace, pre-commit run markdownlint --all-files — each run separately, exit status checked.

Oracle: docs-only, no emulation behaviour change, so n64-systemtest is
unchanged at 90 suite-wide, Phase 1 Failed: 0, per docs/STATUS.md.

🤖 Generated with Claude Code

[coderabbitai]

Important

Review skipped

No new commits to review since the last review.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Pro Plus

Run ID: 22ee6330-d175-4620-a85a-74a07d9e1ad9

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

📝 Walkthrough

Walkthrough

The PR updates the R-18 accuracy record and changelog with the corrected hle_boot stack-pointer diagnosis, observed retail execution results, revised RSP evidence, and a narrowed CI4 plus TLUT texel-fetch hypothesis.

Changes

R-18 diagnosis

Layer / File(s)	Summary
Corrected boot diagnosis and CI4 probe docs/accuracy-ledger.md, CHANGELOG.md	The documentation records the missing initial sp seeding in hle_boot, corrected retail boot and RSP execution evidence, retirement of “lit pixels” as a rendering metric, and the CI4 plus TLUT path resolving texels to zero with an Angrylion-testable vector.

Estimated code review effort: 1 (Trivial) | ~3 minutes

Possibly related PRs

• doublegate/RustyN64#120: Adds core hle_boot CPU state and register seeding matching the corrected R-18 diagnosis.
• doublegate/RustyN64#177: Updates the same R-18 ledger entry and related retail execution diagnostics.
• doublegate/RustyN64#178: Refines the same “lit pixels” and CI4 plus TLUT diagnostic narrative.

🚥 Pre-merge checks | ✅ 10

✅ Passed checks (10 passed)

Check name	Status	Explanation
Title check	✅ Passed	Matches Conventional Commits, stays within 72 characters, and the scope/subject align with the docs-only change.
Description check	✅ Passed	It clearly describes the same CI4 + TLUT narrowing and docs-only intent as the change set.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Oracle Number Is Stated	✅ Passed	Docs-only diff (CHANGELOG.md only); the oracle-number rule is exempt here, and docs/STATUS.md still records 90 suite-wide.
Docs-As-Spec Sync	✅ Passed	PASS: HEAD changes only CHANGELOG.md; no rustyn64-* chip-crate files or docs/.md changed, so there is no observable chip behaviour to sync.
Changelog Entry For User-Visible Changes	✅ Passed	Docs-only PR; no user-visible product change to log, and CHANGELOG.md already has the entry under [Unreleased].
Measured, Never Tuned	✅ Passed	Only ASCII normalisation in the combiner expression changed; no new hardware constants or timings were introduced.
Unsafe Stays Out Of The Chip Crates	✅ Passed	PASS: the patch only touches CHANGELOG.md; no Rust sources change, no unsafe is added, and no #![forbid(unsafe_code)] is removed.

---

_{Comment @coderabbitai help to get the list of available commands.}

[doublegate]

Antigravity — Unicode vs ASCII in the combiner expression. ADOPTED. The same formula appeared as (0−0)×0 + TEXEL0 (U+2212, U+00D7) in CHANGELOG.md and (0 - 0) x 0 + TEXEL0 in docs/accuracy-ledger.md. It is one formula and now reads identically in both, in ASCII — which also survives grep and terminals that mangle U+2212. Same class as the comma/space split you caught on #179: one measurement rendered two ways reads as two.

Nitpick — commas in 4-digit byte fractions (1,508/2,048). NOT ADOPTED, with a reason. They are there deliberately so the pair scans as counts out of a total rather than as a ratio or a version number, and they match the comma convention this file settled on last PR. Dropping separators only at four digits would reintroduce a per-number judgement call, which is what the standardisation was meant to remove.

[github-actions]

Antigravity review (Gemini via Ultra)

This PR updates CHANGELOG.md and docs/accuracy-ledger.md (row R-18) to document the technical narrowing of the RDP video rendering gap specifically to the CI4 + TLUT texel fetch path.

Blocking issues

None found.

Suggestions

• CHANGELOG.md (lines 31–38) / docs/accuracy-ledger.md (line 413): The documentation update states that the issue is "directly testable with a CI4 .rvec vector against Angrylion," but no reproduction vector or test case is included in this pull request. If the vector was produced during this investigation, it should accompany this documentation change to ensure test coverage.

Nitpicks

• CHANGELOG.md (line 32) / docs/accuracy-ledger.md (line 413): (0 - 0) x 0 + TEXEL0 uses ASCII x instead of the multiplication symbol × used elsewhere in the ledger (e.g., × 63/64).

_{Automated first-pass review by agy on a self-hosted runner -- not a human review.}

[doublegate]

@coderabbitai review

[coderabbitai]

✅ Action performed

Review finished.

Note: CodeRabbit is an incremental review system and does not re-review already reviewed commits. This command is applicable only when automatic reviews are paused.