[DrunkOnJava]

Summary

Brings docs/streaming/task-ledger.md and the README's "Checkpoint contract" section up to what is on main, from the code and from the evidence files of a green Streaming E2E run; anything without a run or file behind it is marked NOT MEASURED. Documentation only.

Spec reference

None; no decoder changes. Refs #205, #206, #207 (they stay open: their remaining items are what the ledger now lists), and #209, #210, #211, #212 for their ledger rows.

Evidence: the required Streaming E2E job of run 40532346151 on main at e7c53e1, success, artifact streaming-evidence (one *-evidence.json per step). The same commit's ci.yml run is 40532346172.

Test plan

• Every number was read from the artifact, not from earlier prose: walk-evidence.json (17 file cases, 5,835 records, each with legacy_parity true and the oracle "independent framing parse"), lz77-evidence.json (25 files, 266 pages, 6,097,392 bytes), map-evidence.json, sections-evidence.json, file-evidence.json, objects-evidence.json, r2007-evidence.json, r2007-objects-evidence.json, source-evidence.json, the four wasm*-evidence.json (all status: passed).
• Checkpoint kinds, markers, CRC-32 trailers, restore(input, expected_source) signatures and size constants read from src/streaming/checkpoint.rs and r2007_checkpoint.rs: ten kinds (DWGH, DWGW, DWG4, DWGZ, DWGM, DWGS, DWGF, DWGO, DWGR, DWGQ); every restore checks marker, version, length and the CRC-32, and every one but DWGZ takes the expected source identity.
• The manifests in tools/streaming/*-corpus.json (54 entries) name all 52 pinned drawings in tools/corpus/pins.tsv.
• The committed ruleset (.github/rulesets/main-protection.json) has no bypass actor and the same twelve required checks as the live one; git worktree list --porcelain shows no prunable worktree.
• python3 tools/ci/spec_coverage.py --check and python3 tools/maintainer/train/arch.py check: clean.
• No code, no build.

What changed:

• task-ledger.md: a "last reviewed" line (2026-10-27, e7c53e1, the run); the rows for Streaming 1: replace whole-file parsing with explicit byte-demand states #205, Streaming 2: transport-independent pausable byte sources #206, Streaming 3: portable bounded checkpoints with fresh-process resume #207, Streaming 5: bounded browser input and checkpoint API in native/WASM CI #209, Streaming 6: proptest chunk schedules and automatic streaming invariant gate #210, Streaming 7: real golden corpus through multi-gigabyte DWG and honest reader benchmarks #211 and Watch ODA specification revisions and open deduplicated grammar-review issues #212 rewritten from "header only" to what landed (feat(streaming): walk the R13-R2000 container with byte-demand states #240, feat(streaming): bounded, resumable LZ77 decoder for R2004-family pages #246, feat(streaming): asynchronous byte source driven by its host #255, feat(streaming): verify the R2004-family file header with byte-demand states #260, feat(streaming): walk the R2004 page map and section map with byte-demand states #276, feat(streaming): assemble R2004 sections from their data pages with byte-demand states #299, test(streaming): restore a native checkpoint in a wasm32 kernel and the reverse #301, feat(streaming): walk a whole R2004-family file behind one checkpoint #312, feat(streaming): stream the object records of an R2004-family file (#205) #333, feat(streaming): walk a whole R2007 file with every checksum checked before its bytes are handed out (#205) #340, feat(streaming): stream the object records of an R2007 file (§20.1, #205) #363, the perf series docs(streaming): measure the streaming walks against the legacy reader per release band (#205) #343, perf(streaming): fold eight bytes per step in the DWG CRC-8 (#205) #360, perf(streaming): add the §4.2 page checksum in eight lanes (#205) #362, perf(streaming): copy a record's first bytes as a slice (#205) #364 to docs(streaming): record where the streaming walks' time went (#205) #367) with the run's counts and a per-row list of what is open or NOT MEASURED; the stale "Discovered" notes resolved (ruleset drift: resolved by docs(maintaining): main is governed by the main-protection ruleset alone, with its ten required checks and no bypass (#117) #220 and docs(ci): Corpus and Spec coverage are required on main #292; the two real inputs are pinned and walked in CI; the foundation PR is merged); the "prunable worktree" row closed.
• README.md, Checkpoint contract: was the DWGH header checkpoint alone, ending "Full object/decompression checkpoints remain open in Streaming 3: portable bounded checkpoints with fresh-process resume #207". It now states the contract for all ten kinds in one table (marker, what each holds, size bound as the section of each walker states it), the common restore rules, the caller's custody of emitted output, and what no checkpoint holds (a partly decoded object, because none is decoded). The DWGH byte table stays.
• README.md, other edits in the same vein: a paragraph about the R2007 object stream's wasm32 run, which sat at the end of the contract section, moves to the end of "Cross-runtime restore" where it belongs; "Current evidence" (which cited a 2026-10-26 header-only run) now cites run 40532346151 with a per-step table; the closing sentences of "Repeatable acceptance" no longer say the manifests record no object counts or that no independent oracle has passed (they say what each manifest records), and keep the R13, multi-gigabyte and browser gaps as NOT MEASURED.

Left as it was, because it is still true: the per-walker sections of the README, the throughput table and the audit.md file.

Source-provenance declaration

By submitting this PR I certify that my contribution does not incorporate executable code from any source whose license is incompatible with Apache-2.0. Specifically:

• No Autodesk DWG SDK source (RealDWG, ObjectARX, ObjectDBX) or NDA-protected Autodesk documentation was consulted or imported.
• No Open Design Alliance Teigha / Drawings SDK source was consulted or imported. The ODA's public Open Design Specification for .dwg files PDF is explicitly allowed (see CLEANROOM.md).
• No GPL-licensed DWG implementation source (LibreDWG, etc.) was consulted or imported. This matters because dwg-rs is Apache-2.0 and must remain compatible.

My contribution is based on:

• The ODA's freely-redistributable Open Design Specification for .dwg files (v5.4.1);
• Public reverse-engineering notes, research papers, or blog posts cited inline;
• Reading algorithm-description comments (not executable code) in permissively-licensed projects (MIT / Apache / BSD) to resolve spec ambiguities — disclose any such reading in this PR body so it can be recorded in CLEANROOM.md.

See CLEANROOM.md for the full allowed/forbidden source list and the honest scope of what "clean-room" means for this project.

Developer Certificate of Origin

By submitting this PR, I certify that the contribution complies with the
Developer Certificate of Origin 1.1.
Specifically, I affirm that the contribution is my own work (or rightfully
licensed to me) and that it can be distributed under the project's Apache-2.0
license.

I will sign off commits with git commit -s when asked, which adds a
Signed-off-by trailer to the commit message asserting the DCO.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-28T17:11:09.284031Z	f16c8c7	New commits

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.