docs(ledger): localise the video gap to RDP rasterisation, not presentation

[doublegate]

Correcting my own localisation

The previous entry blamed the "RDP → VI presentation path". That was wrong,
and this PR replaces it with a measured boundary.

Presentation is not at fault

Ocarina of Time's framebuffer at VI_ORIGIN is uniformly 0x0001 — RGBA5551
with R=G=B=0 and the coverage bit set. The VI is faithfully presenting a
genuinely black buffer, and scanout is correct.

(VI_ORIGIN = 0x3b5280 sits exactly one scanline into the RDP's colour image
0x3b5000, which is normal — so even the addresses line up.)

Submission is not at fault either

A DPC-seam opcode census over 300 frames shows the real F3DEX stream arriving:

Opcode	Command	Count
0x0F	TRIANGLE — shade + texture + Z	7,412
0x35	SET_TILE	7,705
0x33	LOAD_BLOCK	4,374
0x24	TEXTURE_RECTANGLE	1,630
0x3C	SET_COMBINE	1,589

…of 74,508 commands. Geometry, textures and tile state all reach the RDP.

So the gap is between those two

The RDP rasterises real geometry to black. That is far more useful than "does
not reach video", and unlike the previous framing it is directly testable.

First suspects, each isolatable against the Angrylion vector harness that already
exists:

• the Z path — every one of those 7,412 triangles is the 0x0F Z-buffered
  variant, so a z_compare that rejects universally produces exactly this result;
• texture fetch returning 0;
• the combiner collapsing to 0.

A note on my own process

Two localisations in two turns have now been wrong — presentation, and before
that the exception vector — both because I named a subsystem before measuring
its boundary. In both cases a census or a correlated capture settled it in one
run. Recorded in the ledger alongside the finding, since the pattern is more
useful than the individual corrections.

Gates

cargo fmt --all --check, cargo clippy --workspace --all-targets -- -D warnings,
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]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Pro Plus

Run ID: dff2d12f-3da6-4fc5-950b-22f5de277f13

📥 Commits

Reviewing files that changed from the base of the PR and between 3b95d25 and c7568e7.

📒 Files selected for processing (2)

• CHANGELOG.md
• docs/accuracy-ledger.md

---

📝 Walkthrough

Summary by CodeRabbit

• Documentation
  • Updated the accuracy ledger to mark the Ocarina of Time execution issue as substantially resolved.
  • Clarified that missing CPU initialisation prevented proper video setup.
  • Corrected evidence and measurement guidance for RSP execution and framebuffer output.
  • Added graphics-command census data and narrowed remaining video issues to black RDP rasterisation.
  • Updated the changelog with more concrete, testable evidence and investigation targets.

Walkthrough

R-18 documentation now records the missing sp initialisation in hle_boot, corrects RSP execution measurements, retires “lit pixels” as evidence, and narrows the remaining video issue to black RDP rasterisation.

Changes

R-18 evidence correction

Layer / File(s)	Summary
Corrected rendering diagnosis docs/accuracy-ledger.md, CHANGELOG.md	R-18 is marked substantially resolved with the hle_boot stack-pointer fault identified, authoritative SP_STATUS measurements documented, “lit pixels” retired as a pass metric, and remaining black-output evidence narrowed to RDP rasterisation suspects.

Estimated code review effort: 2 (Simple) | ~10 minutes

Possibly related PRs

• doublegate/RustyN64#120: Adds or migrates the hle_boot and boot wiring implicated by the R-18 diagnosis.
• doublegate/RustyN64#174: Corrects RSP status and PC observation using authoritative state.
• doublegate/RustyN64#178: Updates R-18 interpretation of “lit pixels” as rendering evidence.

🚥 Pre-merge checks | ✅ 8 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Title check	⚠️ Warning	The title is relevant, but it exceeds the 72-character Conventional Commits limit.	Shorten it to 72 characters or fewer, for example: docs(ledger): localise video gap to RDP rasterisation.},{

✅ Passed checks (8 passed)

Check name	Status	Explanation
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	PASS: only CHANGELOG.md and docs/accuracy-ledger.md changed, and they either say “not measured” or quote the current 90-suite oracle; docs/STATUS.md confirms 90.
Docs-As-Spec Sync	✅ Passed	PASS: the diff touches only CHANGELOG.md and docs/accuracy-ledger.md; no rustyn64-* chip crate files changed, so the docs-sync rule is not triggered.
Changelog Entry For User-Visible Changes	✅ Passed	CHANGELOG.md has an [Unreleased] fixed entry for the visible Ocarina/R-18 correction, so the user-facing change is recorded.
Measured, Never Tuned	✅ Passed	PASS: the new changelog numerals are mirrored in R-18, which records the 300-frame census and framebuffer measurement method.
Unsafe Stays Out Of The Chip Crates	✅ Passed	PASS: Diff is docs-only; no Rust sources changed, and chip crates plus rustyn64-core still have #![forbid(unsafe_code)].

---

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

[doublegate]

Adjudicating the Antigravity review.

Inconsistent number formatting — ADOPTED. The same census appeared with commas in CHANGELOG.md and spaces in docs/accuracy-ledger.md, so one measurement read as two. Both now use commas for this content: the CHANGELOG is decisively comma already (38 to 2), and it is the more standard convention.

Scoped deliberately: docs/accuracy-ledger.md is historically mixed (26 space-separated vs 22 comma-separated), so there was no existing convention there to violate. I changed only the census text — a whole-file sweep would bury this correction inside a large unrelated diff, and the ledger is an append-mostly record where that matters.

Incomplete opcode breakdown — ADOPTED. CHANGELOG.md omitted 1,589 SET_COMBINE. The two lists now match exactly. That one mattered more than the formatting: a census with a silently different member list is a different census.

Nitpick: the R-18 row is an overly dense block — AGREED, not actioned here. This is the second review to raise it and it is right; the row has accumulated many appended corrections and is now hard to read. I have not restructured it in this PR because rewriting the row layout would obscure the correction the PR exists to make. It belongs in a dedicated ledger pass, and I would rather it stay visibly overdue than be half-done inside an unrelated change.

[github-actions]

Antigravity review (Gemini via Ultra)

This documentation PR updates CHANGELOG.md and docs/accuracy-ledger.md (ledger item R-18) to clarify that the video rendering gap in Ocarina of Time is localized to RDP geometry rasterisation rather than the VI presentation layer.

Blocking issues

None found.

Suggestions

• CHANGELOG.md#L26-L28: The line wrap places SET_TILE at the start of line 28 with leading whitespace ( SET_TILE), disrupting line continuation inside the list item.
• docs/accuracy-ledger.md#L413: The resolution text in row R-18 appends ~1,200 characters onto an already extremely long single-line table cell. Consider whether historical investigation logs belong directly in the ledger table or in dedicated tracking docs to avoid unmaintainable table rows.

Nitpicks

• CHANGELOG.md#L27: Z-buffered TRIANGLEs has a dangling plural 's' outside the backticks, whereas docs/accuracy-ledger.md#L413 uses ``TRIANGLE (0x0F: shade + texture + Z). Standardize the opcode formatting across both files.

_{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.