The map gains harness:reference-apps-compat (EVM-02 to EVM-06, partial coverage named per case), harness:reference-apps-ver
(VER-03 to VER-06 and VER-08) and harness:reference-apps-ver-trust-anchor (VER-01 and VER-02, FAIL by design and disclosed).
Batches ra-20261008T1744-evm (22 of 22 rows on the 2.0 devnet, node 4cdcc488) and ra-20261008T1746-ver (evidence per case
under tools/reference-apps/ver/evidence) are recorded with node tools/ci/test-record.mjs --record; the harness map page is
regenerated; the registry rows move with their evidence (site/lc/verify-receipt.js rebuilt in this landing).
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Run 201-ef0f2ed8-2826f37e: node ef0f2ed8 (291ee6ae plus the 2.0.1 version bump) on the key-succession pairing, kaspad check and the five node-side suites green on build-4; the miner cut tip 2826f37e (app crate identical to 2ea7b43f), pow and app green on build-2; evidence at build-1:/srv/artefacts/tas/201-ef0f2ed8-2826f37e/. GOV-01's row takes F0 (docs/plans/igneum-2.0-f0-manifest.md) as its evidence through the batch f0-manifest-20261008-b (the earlier f0 batch landed before the replay existed and never recorded). The map splits the finality cells at the finality lane's ask: harness:finality-sim keeps FIN-08 (the model side), harness:finality-sim-fin02 carries FIN-02's simulator half, node:finality-realnode carries the node lane's real-node lines for FIN-02 and FIN-07.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
docs/plans/igneum-2.0-f0-manifest.md is the fixture F0 every case pins to: the identity pins as the steward read them on 8 October 2026 (the miner cut tip 2ea7b43f, the node roll sha ef0f2ed8, igneum-devnet-4, the object digest be5f4068, the class v5 freeze and its fingerprint, the roles and kits), the four fields main ruled on at 19:3x (P07's workload W as the pilot plan's shape; P00 and P08's fault model and thresholds as the finality spec's own at a1b3977a6; P10's study NOT RUN until unaffiliated participants are recruited, never a staff run; P02's AMD cells the 9070 XT and RX 7600 on PC 1, PARTIAL until a rented AMD cell exists), an unresolved-field register with owners and defaults, and the signing block for 23:30. GOV-01's row (through the batch f0-manifest-20261008, replayed at the merge) adds the manifest to its evidence path and stays RUNNING until the signatures are in.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The acceptance registry is a hot file: three landings in a row lost rule 26's race to another lane's rows during their own seven-minute gate. Now a registry landing carries its run batches as tools/ci/batches/<run id>.json (the files test-record.mjs --record took); merge-to-master.sh, for a branch that added batch files, lets rule 26 pass the registry path (the branch's copy is never what lands), and in the merge worktree takes master's copy of the registry, replays every added batch through the recorder (idempotent), commits the merge and runs the evidence rules on the merged result before the push. Self-test: a fixture where master gained a row after the branch point and the branch carries a batch; the merge lands with both rows. The batches: 200-417c4a57-b2 (the 2.0.1 gate, 30 cases RUNNING), review-harness-20261008T180034Z (POW-01's cross-check, all four anchor groups agree), p01-partb-20261008-01 (POW-06, three parsers, zero panics); the map gains crosscheck:review-harness and part B's command and coverage.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Pre-2.0 ledger ids pointed at 2.0 entries or dropped (M32 to D3, E19 to A7,
M4 to C1, P1/P3 to D5; E20, F7, M12 dropped); the dataset row and sentences
read the served dataset policy; engineering-log citations name
docs/bench-log.md; the node1-dn3 source notes gone; the earlier devnet named
in governance; the four words on /address, /proving and /scenes; the tx fee
sentence matches economics; provenance's SP1 sentence reads the current fact;
receipt capitals and the unlabelled timing dropped; the lock tile reads active
and total weight; TEAM-REPORTED labels on app, income, compatibility, build,
light and the dev fee row. New /docs page listing the 2.0 documents by their
master paths, wired into the build, nav, nav check, sitemap, capture and og.
docs/site/audit-2.0.md moves each fixed item into "What was rewritten".
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
tools/ci/registry-evidence-check.sh runs in merge-to-master.sh after rule 26: (1) a landing that sets a case's run_status to PASS must name an evidence_path that exists, in the tree at the landing or on a build box over ssh (a box that does not answer is a line, not a refusal); (2) a landing that changes a file under docs/analysis/ or a path a registry row names without moving that row's updated field is refused (GOV-04: the row and its evidence move together); (3) a PASS whose evidence record pins another manifest than the registry's pinned_manifest_sha is refused (GOV-08: stale evidence reads NOT RUN); (4) a run_status written while the registry carries no approval block is refused (GOV-02: thresholds before results). Self-test on a fixture repo covers each rule both ways.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The page shows each of the 128 cases in 16 suites as it is run, in the standard's words (NOT RUN, RUNNING, PASS, FAIL,
BLOCKED, DEFERRED): a header with the approval line and a thin segmented count bar (never a percentage), G0 and the five
gates G1 to G5 plus the three named gates as cards whose state follows the standard's rule, one section per suite with an
expandable row per case (setup, steps, accept, required evidence, the run's evidence record when present), the profiles
(approved as proposed, P13 deferred, the P04 line read as a FAIL to work against), the fixtures, and the closing notes.
site/lib/acceptance-render.mjs is the one renderer: build.mjs imports it for the build-time page and prints the same functions
into the page's inline script, which polls /api/acceptance (site/api/acceptance.mjs: the registry from git.igneum.network
master, cached 60 s, ok:false with status 200 on failure) and re-renders in place when a case state or source_sha256 moves.
Hooks: the Learn panel, the nav check, the sitemap, capture.sh, the og card row, a lead link on /scorecard, the facts rows 1
to 7 name the case that will prove each. The never-served check strips the plan's own marked never-served list. Node tests
(known-failed first) run in the gate; the registry and the standard text land with the page.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
tools/ci/test-map.json gains bench:pc1-packs (GPU-02, GPU-03, GPU-05, POW-08 on PC 1's 5090), bench:pc1-amd (GPU-01, GPU-05 on the RX 7600), bench:fleet-pods (GPU-01, GPU-03, POW-08 on the rented cards), census:class-v6 (POW-02), POW-07 under suite:pow (the FP32 branch unreachable on master) and UX-08's published-tuning half; GPU-04 and GPU-07 read NOT RUN with the Devnet 3 off order as the reason; the harness map page is regenerated. tools/ci/p01-vectors.py is P01 part A as the hash lane shaped it: the real worker driven with pool.rs's job lines at an all-pass target, every found hash diffed against the CPU reference list, the evidence JSON with the counts, the first ten disagreements, the pack and program ids, the device line and the manifest sha; exit 1 on any disagreement or unanswered nonce; self-tested with a fake worker (clean PASS, one wrong hash FAIL naming the nonce, one unanswered nonce FAIL naming it).
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
tools/ci/test-map.json maps every automated case of the Test and Acceptance Standard's registry to the cell that runs it (the seven crate suites of the release matrix, the pre-push gate, the freeze check, the fast-time harness, the finality simulator, the P01 vectors and parser-malformed harnesses as they land), each cell with its command, box class and fixtures (F0 onward), each case with what the cell proves and what remains; 31 of 109 automated cases map tonight, 78 read NOT RUN with the reason. tools/ci/test-map-doc.mjs writes docs/plans/igneum-2.0-test-harness-map.md from the JSON (--check in the gate refuses a stale page). tools/ci/test-record.mjs --check refuses an unmapped automated case without a reason or an unknown id; --record <batch> writes run_status, run_id, evidence_path, updated and the evidence record (cell, manifest sha, coverage) to the mapped cases only, leaves every accept text byte-identical (the self-test diffs them), gives an unmapped automated case NOT RUN with its reason and a manual case nothing. Rule: a mapped cell's green writes RUNNING; PASS only when coverage is full and the evidence file exists; never PASS by inference.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The markdown loop in site/build.mjs takes paths under docs/ (build/build.md, build/grants.md, build/compatibility.md, site/audit-2.0.md); /audit joins the Learn panel (both lists), the nav-check groups, the sitemap, the capture routes and the og cards.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
tools/ci/rule26-no-revert.sh: for every path under site/ or docs/ that master changed since the merge base and the branch also touched, every non-blank line master added must be present in the branch's copy (a merge of master carries them; a replace from an older copy loses them); a path the branch deleted counts as every line lost; a landing whose commit messages name the path lifts the refusal for it. merge-to-master.sh runs it after rule 24, before the merge loop, and refuses with the paths and the fix named. Self-test on a fixture repo: a replace and a deletion are refused and named; a branch that merged master, one on a disjoint path, and one that names the path as intentional pass.
Class (main through the coordinator, 8 October 2026, 17:5x UK): the og lane's a68a36c0 and the release-manifest ffd6751e put the reference-apps lane's receipt.html and then all of site/lc back to older copies on master, restored by its own merges.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
tools/ci/rule24-crate-gate.sh maps every .rs, Cargo.toml, Cargo.lock, build.rs or .cargo/config in the diff to its crate (the nearest [package] Cargo.toml; the workspace root when none nearer; the core pair igneum-pow and app/igneum-app for a root config; a vendor crate named and left to its lane) and runs cargo check then the crate suite on the box at gate priority; docs/ and site/ alone (docs-only-check.sh) run nothing; any other diff takes the full gate alone. merge-to-master.sh runs it against the merge base before its merge loop; the hook runs it on a push of master or release-* against the remote tip. Self-test with a fixture repo and a recording runner: the crate map for each file class, documents-only runs nothing, check then test per crate, a red suite fails the gate.
Class (main through the coordinator, 8 October 2026, 17:2x UK): master's igneum-pow stopped compiling at 17:07 under landings that never built it.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Documents-only copy of counter-asic-4 df7c85c1c (its delta over master 919293896: docs/analysis/class-v6, docs/design, docs/spec, the export list; nothing outside docs/ differs at that tip; the branch itself stays unmerged, its history carrying the igneum-pow research commits)
