Repository navigation
Conversation
… (0.69.4 #4318 cuda-unit red) cuda-unit on #4318 (f230832) failed gpu_every_serve_route_prefills_through_the_batched_prefill with left 2, right 1. Its log shows the chat request DID prefill batched, twice: 14 tokens from position 0, then 7 from position 14. e6004d7 (#4214) splits a prompt at its checkpoint (before a chat's last <|im_start|>, else before its last token) and updated the session test to 2 and 4; the #4250 serve test kept "one per request". Expected per route, from Session::prepare_prompt + Qwen35Session::checkpoint_at: - chat, fresh: history span + generation-header span = 2 - chat again, streamed: restores request 1's checkpoint, one span = 3 - /v1/completions, raw: one span up to the last token, then a one-token step the counter never sees = 4 Per-token serve adds 0 everywhere, so #3596 stays RED. The messages keep their "prefilled its prompt one token at a time" prefix, which the cuda-nightly RED control greps. Refs #4250 #4214 #4318 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit 9bc06c3)
…KIP there is RED (#4251) ci.yml's cuda-only step selects gpu_every_serve_route_prefills_through_the_batched_prefill. There a missing Qwen3.5 file counts as a legitimate skip, and the runners carry none, so the test passed without running. The gx10 job now gets a step. It resolves APR_QWEN35_MODEL, whose default path gx10 carries (vars override, as for PERF-053). If the file is absent, the step fails as UNMEASURABLE. It then runs the test with --exact --nocapture and fails on: - a nonzero exit, - any `SKIP:` line, - any result other than exactly `1 passed`. It carries !cancelled(), so an earlier red falsifier cannot hide it. The test module reads APR_QWEN35_MODEL, and falls back to the old hardcoded path. Case table on the step's own run: block, driven by a fake cargo: pass=0, skip=1, fail=101, zero-tests=1, model-absent=1. All 5 matched. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit 0a27696) (cherry picked from commit 86d3775) backport note: non-workflow part applied; WITHHELD for 36: .github/workflows/cuda-nightly.yml hunks (this session may not edit .github/workflows) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…r 0 tests run The step and #4251's done_when both used the bare function name with --exact. libtest matched 0 tests (17978 filtered out). The step's `1 passed` check would have turned that RED, but it would never have run the test. The test's path is api::openai_handlers::qwen35_serve_tests::gpu_every_serve_route_prefills_through_the_batched_prefill. On lambda (RTX 4090), through gpu-q, with the binary built from this branch: - default: ok, 1 passed, and every route logs `[qwen35] batched prefill: N tokens` - APR_QWEN35_SESSION_PREFILL=per-token: FAILED (exit 101), panics at qwen35_chat_backend_tests.rs:448 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit bfcf8dc) (cherry picked from commit f6da455) backport note: WITHHELD for 36: workflow-only (.github/workflows/cuda-nightly.yml); this session may not edit .github/workflows Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…per-machine model path guard-cargo: #4110 inserted FALSIFY-VERSION-005 before -004 in apr-version-traceability-v1, a new "test ID gap" violation that took contract_data_integrity to 445 against its 444 ceiling. Put 005 after 004. validate_contracts: 10/10 pass; pv validate: 0 errors. guard-tree: #4254's effective-config test hardcoded /home/noah/models/Qwen3.5-0.8B-Q4_K_M.gguf, taking out-of-workspace fixture paths 85 -> 86. The path now comes from APR_QWEN35_08B_GGUF (skip names the variable when it is unset). check_test_fixture_paths: 85/85 PASS. With the variable set, the test ran and passed (1 passed, no SKIP). Refs #4318 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit deb10aa) backport note: PARTIAL: effective_config_route_pp2.rs per-machine model path applied; contract ID reorder SUPERSEDED (car already lists FALSIFY-VERSION-005 after -004, pointing at aprender-build-sha emit_tests) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…set checker (#4292) build-apr-darwin builds apr (default features, no cuda) natively, proves arm64 + no libcuda + --version, records minos, and uploads apr-<tag>-aarch64-apple-darwin-cpu.tar.gz(+sha256) over REST. check_release_assets.sh now expects 18 assets; verify-apr-assets waits for the lane. Runner label is PENDING an operator ruling (#3073 no-hosted rule). Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> (cherry picked from commit f1ff895) (cherry picked from commit 134df95) backport note: SUPERSEDED on car by #4327: binary-release.yml build-apr-darwin (mini-builder) + check_release_assets.sh 18 assets incl darwin Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…hosted macos-14 (#4292) Operator 2026-09-24: "register the mini as self-hosted mac builder". Adds a tool preflight for the self-hosted box. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> (cherry picked from commit f2743de) (cherry picked from commit 0dece9f) backport note: SUPERSEDED on car by #4327: build-apr-darwin runs-on [self-hosted, macOS, ARM64, mini-builder] Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…atchet — mini-builder declared (infra#1042) - check_runner_labels: mini-builder is a one-box discriminating label (the mini's dedicated repo-level release runner, infra#1042) - actionlint.yaml: declare mini-builder as a known self-hosted label - check_release_assets(.sh|_test.sh): here-strings instead of pipe-into-grep -q (ratchet stays at baseline 74) Refs #4292 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit bc27a3c) (cherry picked from commit 995a52c) backport note: SUPERSEDED on car: mini-builder already in check_runner_labels.sh and .github/actionlint.yaml; car's check_release_assets(.sh|_test.sh) are the #4327/#4352 rewrite Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…r on the Qwen3.5 hybrid (#4254) The batch's #4254 fix (cbc571a) is pinned by a test that calls effective_config() directly. The issue is that the two ENDPOINTS contradicted each other in one process, so this drives both through create_router_with_config on one real 0.8B state: compute_class and backend_loaded equal /health's compute_mode, quantization and parameter_count are measured, build_features lists gpu/cuda only when the launcher can dispatch them, and build_commit is APR_GIT_SHA with no +no-git suffix. Ported from PMAT-4254-effective-config-cpu@4264ee2cc (the test only; its fix duplicates cbc571a, which is already in this batch). Refs #4254 Pmat-Ticket: PMAT-4254 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit 031c2bc)
…very member again (#4352) The stamp heredoc, the versions() JSON and the mutant builder are bash + POSIX awk + jq. Parity with the python: 121 cases (real workspace manifests + 9 edge fixtures x 11 tags, incl. CRLF/CR, no final newline, dup members, exclude ".", source on a block's first line, consecutive [[package]], quoted tags) -- stdout, rc and the rewritten tree byte-identical. ONE deliberate difference, and it is a live rc-cut defect: the python cut `members = [...]` at the first `]`, even inside a comment. #4219's "# ... every [[bin]] ..." (batch/0.70.0 only) ended the list after 3 entries, so apr-cli's `aprender = { path, version = "0.69.0" }` was never stamped: batch tree + python stamp -> `cargo metadata` rc 101 "failed to select aprender-core ^0.69.0" (unstamped control rc 0). The port strips `#` comments first: 81 manifests, 116 pins, cargo metadata --locked rc 0, all 80 workspace crates at the rc. Self-test: scratch `members` now carries the #4219 shape (RED on the old stamper: LOCKED-FAIL), and a 5th mutant (comment strip dropped) must be refused by cargo. PASS with python3 stubbed. Refs #4352 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit d20f3a3) (cherry picked from commit 51d97f4)
…275 fragment only) Pmat-Ticket: PMAT-4275 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit aeeda2a) backport note: PRESENT: roadmap.yaml union with car adds nothing (aggregate regenerate; car's aggregate is regenerated from its own fragments) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
… the host's ~/.cargo/bin — it overwrote the gx10 fleet-bins release binary every beat (Refs #4346) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit c4fd00a) (cherry picked from commit 793ff9ef364f6796fa4419da283756e01b85b2af) (cherry picked from commit 05c813d) backport note: WITHHELD for 36: workflow-only (.github/workflows/qwen-story-daily.yml); this session may not edit .github/workflows Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…4283) Measured 2026-09-24: `--version v0.69.3-rc.1` already installed a working rc, but only for a dev who already knew the tag. /releases/latest skips prereleases, --help said --version took a "stable tag", and there was no way to ask for the newest rc. --channel stable|rc|nightly. rc = the first vX.Y.Z or vX.Y.Z-rc.N tag in /releases (newest-first, prereleases included), so it never picks the rolling nightly and never an rc older than the latest stable. --channel nightly is --nightly. --channel rc + --version is refused. install_test.sh: +4 rows. Row 12 derives the expected tag itself and asserts the installer chose it. Mutant `get_rc_version -> get_latest_version` turns row 12 red (resolved v0.69.1, wanted v0.69.3-rc.1). Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> (cherry picked from commit 6205968) (cherry picked from commit 14cc35903468ab87929655a73efa2de87fe9eafa) (cherry picked from commit 9595b70)
…htly + release gate (#3546) scripts/tests/install_e2e_container.sh runs the README one-liner in debian:bookworm-slim as a fresh non-root user and checks, against facts resolved on the host (not install.sh's own output): channel -> tag (and --expect-tag for the release gate), published .sha256 == asset, install.sh reported that digest, apr on PATH byte-identical to the asset's, and `apr --version` names the tag version AND a prefix of the tag's commit (`+no-git` fails). --self-test: 8-row case table. Measured 2026-09-24 on lambda: stable v0.69.1 and rc v0.69.3-rc.2 pass 5/6; both FAIL the version check — the published binaries print `+no-git` (APR_GIT_SHA_OVERRIDE is on B2, not main). install.sh: bashrs 0 errors / 0 warnings (colour vars defaulted, quoted help heredoc, `..` in --install-dir rejected). install-script.yml: e2e-container job (stable, rc) nightly + on path changes; lint step requires 0 warnings for install.sh and the e2e. binary-release.yml: install-e2e job after verify-apr-assets, wired into summary. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit 2fe8a74) (cherry picked from commit 4a58f9fe045b62540625058ffe2831c60a63cfb9) (cherry picked from commit 0a0c597) backport note: non-workflow part applied; WITHHELD for 36: .github/workflows/binary-release.yml,.github/workflows/install-script.yml hunks (this session may not edit .github/workflows) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…nreachable) Unsigned; the CI signer signs. Guard accepts a throwaway-key-signed copy. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit 01c71e6)
/#4275 parity) Batch quorum lane 1 FAIL: build-apr-darwin built the unstamped tag tree, so a vX.Y.Z-rc.N darwin asset printed bare X.Y.Z, and no smoke job checks darwin. Now stamps like the linux lanes (stamper + checker from the workflow commit) and runs asset_version_check.sh on the built binary before packaging. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit 3c4e192) backport note: SUPERSEDED on car: build-apr-darwin already stamps via stamp_rc_version.sh from the workflow commit and checks it (binary-release.yml ~650) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
… CI signer) (cherry picked from commit cef1b12)
…s sha (#4318) - install_e2e_container.sh: resolve apr by walking PATH inside the container and assert the resolved path is tester's ~/.local/bin/apr (guard-cargo flagged 3 unpinned apr refs). - binary-release.yml build-apr-darwin: pass APR_GIT_SHA_OVERRIDE (9-hex HEAD), as the linux cuda/cpu lanes already do, so the asset is refused by asset_version_check.sh if +no-git. The e2e red on 'apr 0.69.3 (v0.69.3+no-git)' is the published v0.69.3 asset (built before 59135e4 added the override); it clears when an override-built tag is published. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit 696576e) backport note: non-workflow part applied; WITHHELD for 36: .github/workflows/binary-release.yml hunks (this session may not edit .github/workflows) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
SC2106 misread the { ...; break; } brace group as a subshell: exit 0/exit 1 from the $(...)
instead, which also makes a missing apr a non-zero status. SC2154: APR is assigned in the
su shell, invisible to the linter.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
(cherry picked from commit 6281d9f)
…xity, roadmap aggregate - kernel-fusion-v1.yaml: FUSION-006/007/010 call_site pins moved +10 lines (381/384/387) after da38177 (#4273) grew kernels_generate_gemm_cuda.rs; fusion_call_site_guard_3985 read a match arm, not the kernel call. - session.rs: Session::generate's prompt refusals move to admit_prompt, so generate drops from cognitive 26 (NEW over 25 vs main after #4325's cancel poll) to 20. Same errors, same order. - roadmap.yaml regenerated with make roadmap-aggregate (16 lines of drift). Pmat-Ticket: PMAT-4318 Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> (cherry picked from commit b02807d) backport note: PARTIAL: session.rs generate complexity split + roadmap applied; FUSION call_site pins kept as car's (they pin car's own kernels_generate_gemm_cuda.rs lines) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…an overflowing size (#4313) saturating_add let usize::MAX bytes "fit" in usize::MAX free; the test caught it. Budget: cosine 0.999995, logits 7.5e-3, state 1.25e-2 (worst of n=64/600 with splits on 0.8B: 0.9999995, 7.32e-4, 1.215e-3). Pmat-Ticket: GH-4313 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit ebdac12) (cherry picked from commit 3ca08df)
guard-cargo's check_no_claim_literals refused 3 new measured-reality literals brought in by the 0.69.5 fold; the comments now cite aprender#4273 instead. Pmat-Ticket: PMAT-4318 Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> (cherry picked from commit 5660b08) backport note: SUPERSEDED on car: #4273's car port (8d22414) already words these comments without tok/s literals Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…ateConfig The #4334 union (83bb714) made st_cpu_generate take the generate config the handlers build. The car-only #4269 witness test still passed (max_tokens, temperature). apr_cpu_generate_config(2, 0.0, None, []) is the old config exactly: stop on token 0, the default top_p. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…, 05c813d) Cop ruling (aprender-77): the workflow edits of the backport set land as ONE commit, with no runner or secret changes, reviewed by a 3/3 quorum. - 86d3775 cuda-nightly: run #4250's serve batched-prefill test on gx10; a SKIP there is RED (#4251) - f6da455 cuda-nightly: #4251's --exact filter needs the full test path, or 0 tests run. The path api::openai_handlers::qwen35_serve_tests::… is still right on car: the tests file is include!'d into openai_handlers.rs (measured with `--list` on the car lib test binary). - 05c813d qwen-story-daily: build apr in place instead of cargo-installing into the host's ~/.cargo/bin Not here: 2f9fe27 (wire check_cascade_converges.sh). Its ci.yml anchor, the guard-cargo steps, moved to ci/sections.yml on car, which this ruling does not cover. (cherry picked from commit 86d3775, f6da455 and 05c813d, .github/workflows hunks only) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…t split the git chain d96c2ad3e (#4254) rewrote `run_git(&head).or_else(|| trusted_git_retry(&head))` across three lines and appended `.or_else(read_head_sha_from_dot_git)`, so the plant's one-line anchor matched 0 times and planted_regressions_turn_red failed on its own precondition. The plant now deletes the retry link alone; the foreign-owner row still turns RED (7/7, cargo test -p aprender-build-sha --lib). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
noahgift
added a commit
that referenced
this pull request
Sep 28, 2026
…key) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
noahgift
added a commit
that referenced
this pull request
Sep 28, 2026
…ire); generated ontology files stay main's Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
noahgift
added a commit
that referenced
this pull request
Sep 28, 2026
noahgift
added a commit
that referenced
this pull request
Sep 28, 2026
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Member
Author
|
quorum-review (AD-04): NOT agreed (auto_merge: checked=true was_armed=false disarmed=false) {
"ticket": "PMAT-4502,PMAT-4431,PMAT-4438,PMAT-4072,PMAT-3847,PMAT-4199,PMAT-4200,PMAT-4347,PMAT-4445",
"head": "0be7c0037284cf5bf04b158a1a71036d60096bb8",
"width": 3,
"executor": "agy",
"agreed": false,
"auto_merge": {
"checked": true,
"was_armed": false,
"disarmed": false,
"note": "auto-merge not armed"
},
"lanes": [
{
"lane": 1,
"verdict": "FAIL",
"findings": 3
},
{
"lane": 2,
"verdict": "FAIL",
"findings": 0
},
{
"lane": 3,
"verdict": "FAIL",
"findings": 2
}
]
} |
noahgift
added a commit
that referenced
this pull request
Sep 28, 2026
bashrs SEC001 flagged `eval "$*"` in the self-test's row() (#4502 defects row 10). row() now runs "$@" directly; the four mutations that need a redirect or two steps become named functions. All 15 rows keep their wanted rc and needle, so each RED row still proves its mutation bit. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
noahgift
added a commit
that referenced
this pull request
Sep 28, 2026
… row 5, clippy -D warnings) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit afb58c0)
noahgift
added a commit
that referenced
this pull request
Sep 28, 2026
…ames #4502 renamed ten shipped [[bin]]s to aprender-* but the ledger kept the old names, so check_binary_debt.sh read each rename as one NEW plus one STALE (10+10, #4502 defects row 4). Re-key the bin field only; class, target and the legacy_names counters are unchanged (a rename is not a merge/delete decision, and counters move only through the ratchet). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit 50ab758)
noahgift
added a commit
that referenced
this pull request
Sep 28, 2026
bashrs SEC001 flagged `eval "$*"` in the self-test's row() (#4502 defects row 10). row() now runs "$@" directly; the four mutations that need a redirect or two steps become named functions. All 15 rows keep their wanted rc and needle, so each RED row still proves its mutation bit. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit 8a003ed)
noahgift
added a commit
that referenced
this pull request
Sep 28, 2026
…w files — m52 refused 458 and 481 apr-format-golden-fixtures 458 -> 459 and aprender-facade-e2e 481 -> 483, both free ordinals, so the ont8-evidence-gate (458) and pvl-comparator (481) fragments #4502 added keep their slots and pvl-comparator still runs before pvl-challenge (482). Command content unchanged. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> (cherry picked from commit 1630322)
noahgift
added a commit
that referenced
this pull request
Sep 28, 2026
… builds --locked) L0 exception: fixes running job --locked (run 36388075483, mac-check job 108817628569). The build.rs env_key build-dep lines added provable-contracts edges without re-resolving the lock; cargo check --locked failed on the stale lock at b4ebf93. Edges only, no new packages. Receipt: intel cargo check --workspace --all-targets --locked rc=0, 0 errors. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
noahgift
added a commit
that referenced
this pull request
Sep 28, 2026
… — elan 4 refuses to reinstall Row 13, run 36353187094 x86-main: step 1 failed in 0.1s with "error: 'leanprover/lean4:v4.29.0-rc4' is already installed". No judged file was missing. elan >= 4 (4.2.0 on the fleet) exits 1 on `toolchain install` of an installed toolchain, so the step passes once per host and fails every run after that. Install only when the toolchain is missing from `elan toolchain list` (first column, so a "(default)" suffix still matches), and decide that inside the same flock, so two jobs cannot both see it missing and race. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit de92271)
noahgift
added a commit
that referenced
this pull request
Sep 28, 2026
…(tree-reader drift)
workspace-test-shard step 7 (check_tree_reader_tests) failed in run 36388847584:
PR-1 adds aprender-contracts{,-cli} test targets without their registry rows.
Moves #4502's 12 contracts-cli .cmd wiring files into PR-1 instead of parking
11 targets in the unwired baseline (baseline unchanged), carries fb's m52
ordinal renames (457->459 golden-fixtures, 480->483 facade-e2e) that the moved
457/480 files need, and regenerates scripts/tree_reader_tests.txt.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Agent: aprender-88
noahgift
added a commit
that referenced
this pull request
Sep 28, 2026
…-debt re-key) Local run of the wired contracts-cli targets on PR-1 failed (ont4b_shapes_gate, shapes_vacuity, ont_entity_properties, ...: 'entity.type has no Σ entity_type_target_class mapping'): PR-1 carried #4502's contracts code but not its tests/fixtures/ont updates (164 files). Moves them in. Row 4: cherry-picked 50ab758 (contracts/binary-debt-v1.yaml re-key). (cherry picked from commit 50ab758) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent: aprender-88
noahgift
added a commit
that referenced
this pull request
Sep 28, 2026
…usal-ok fixture follows the contract - pv_surface_gate: #4502 carries VS-COUNT-001 (#2648) in validator.rs with no row in the decision table, so every_rule_in_the_validator_source_appears_in_the_table failed. Add a Case: one obligation, stated total 2, Error. - ont_refusal_receipt: #4502 changed contracts/refusal-receipt-v1.yaml but not its byte-for-byte copy in tests/fixtures/ont/refusal-ok; copy it. Agent: aprender-fb Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
noahgift
added a commit
that referenced
this pull request
Sep 28, 2026
noahgift
added a commit
that referenced
this pull request
Sep 28, 2026
… gate runs (#4502 row 2) lint_passes_on_real_contracts requires the refinement gate to RUN whenever CI is set; the gate compares HEAD with merge-base(HEAD, origin/main), else the origin/main tip. The workspace-test-shard section clones at depth 1 and only the tier step's pull_request arm fetched a main, so workflow_dispatch / merge_group / push shards had no base and shard 3 panicked "the refinement gate declined" (run 36384418038). Per the operator ruling (2026-09-28 09:41): pin the EVENT's base commit -- pull_request.base.sha, merge_group.base_sha, push's `before` -- as refs/remotes/origin/main; the origin/main tip only for workflow_dispatch. A missing or zero base fails the step with ::error (no silent skip). Placed after the tier step, whose pull_request arm would otherwise overwrite origin/main with main's tip. ci.yml is unchanged: it has no generator; fat_driver reads ci/sections.yml at run time. Agent: aprender-41 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit cd587d6a5b317467b4eac3b5c10b65ecf2656f8a) Agent: aprender-88
noahgift
added a commit
that referenced
this pull request
Sep 28, 2026
…ontracts tests read shard 1 of run 36391942814 failed ont2c_tracked_ofn_and_report_are_fresh: contracts/ontology.ofn and tbox-report.json are required-tracked writer outputs of contracts/ontology.yaml. PR-1's ontology.yaml is byte-identical to #4502 integ (f7858ca), so integ's outputs are the fresh ones; taken with the PVL discharge summary + Lean axioms the discharge-check target reads. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Agent: aprender-88
noahgift
added a commit
that referenced
this pull request
Sep 28, 2026
…e ordinals 761/762 (#4502 row 3) Shard1 step 13 REFUSEd two ordinals that each named two fragments: 458 (apr-format-golden-fixtures + contracts-cli ont8_evidence_gate) and 481 (facade-e2e + contracts-cli pvl_comparator). A shared ordinal leaves the run order ambiguous. The two contracts-cli fragments move to 761 and 762. Both commands are byte-identical; nothing else names these files. check_explicit_test_commands.sh --check: base origin/fb/4502-a-diag rc=1 (2 REFUSE), this commit rc=0 (101 fragments, no shared ordinal). Refs #4502 Agent: aprender-cb Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> (cherry picked from commit 59d6911a04a30b5f609fb67c11ca1aba32ad4381) Agent: aprender-88
noahgift
added a commit
that referenced
this pull request
Sep 28, 2026
…usal-ok fixture follows the contract - pv_surface_gate: #4502 carries VS-COUNT-001 (#2648) in validator.rs with no row in the decision table, so every_rule_in_the_validator_source_appears_in_the_table failed. Add a Case: one obligation, stated total 2, Error. - ont_refusal_receipt: #4502 changed contracts/refusal-receipt-v1.yaml but not its byte-for-byte copy in tests/fixtures/ont/refusal-ok; copy it. Agent: aprender-fb Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> (cherry picked from commit 036beee) Agent: aprender-88
This was referenced Sep 28, 2026
Member
Author
|
K37 re-scoped into this ticket (la-0.72; checked at The strict gate K37 asked for already exists: What K37 adds here:
Branch |
Member
Author
|
Closed with a pointer (0.70.2 PR backlog, C297 §5; aprender-6a for la-a). The content was split into #4587 (contracts layer) and #4588 (the rest). The docs-only rule files (scripts/ci_docs_only.sh, check_ci_gate_docs_only_rule.sh, PMAT-3668.yaml) are carried by #4639. The 599->617 CB-200 re-baseline was dropped by cop ruling (331e486). The renumbered explicit-test .cmd files are in the splits. Branch batch/ont-10 is kept, not deleted. |
This was referenced Oct 5, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
ONT-10 batch carrier. Per operator ("obviously batch merge as much as possible"), every remaining ONT-10 row commits to this one branch: one draft PR, one quorum, one CI run. No per-row PRs. #4481 stays separate because it is already armed.
Seeded at #4431's head (
fold/b3-onto-mainb1cd8c0), so the diff includes #4431 until that merges. Rows, estimates and sessions are in cop-inboxgates/ont10-sizing.md. Chain: #4479 → #4431 (ONT-4d) → ONT-4g → 20 binary slices → ONT-10.Refs #4079 #3269
🤖 Generated with Claude Code
Conflict resolution
Merge
5c54a8e964(car/0.70.07d0d403ad0into6a80c6acd2; untangle aprender-66, 2026-09-27 18:39Z). Hunks: 3, all in the generated set — 0 by hand.check_census_derived.shPASS); the train is its only writer. Freshpv censuson the merged tree: 1890 files, invariants PASSpv extract contracts, pv built from the merged tree;pv extract --check= 0contract_countscripts/readme_sync.sh --write→ 1837 (derived from the tracked census);readme_sync --check= 0,check_readme_claims.sh= 0Conflict resolution
Merge
c7169cfc0f(origin/main761d6247deinto08fe9a794a; authored by aprender-79, reused by the untangler) has 23 files and 42 hunks. The merge commit holds only the resolution. Generated files were regenerated in later commits:a69d0905a8(census 1888, README count, contracts.nt, shapes.ttl) andcabe264b89(shapes.ttl after the cherry-picks). The pv-sat witness came out fresh, with no change. This git (2.34) has no--remerge-diff, so to re-review, replay it withgit merge --no-commit -c rerere.enabled=false 761d6247deon08fe9a794aand diff againstc7169cfc0f.--versionrowspv censusviascripts/readme_sync.sh --write(1888)setsid --wait--versionrows, coverage 144/894 (16.1%)build_sha(APR_GIT_SHA)build_shabuild_sha, nowgit rev-parse --short=9so pv/pv-sat print a sha9--versionshabuild_sha. ont-10's PTOP_GIT_SHA renamed onto itbuild_shabuild_sha. SIMULAR_GIT_SHA renamed onto it816973c5bf, main @aca6f2d7f6)Cherry-picked onto the merge: aprender-88's
a3def2970a241f94554f03f24fc6c3(one roadmap hunk, a union of PMAT-4469 and PMAT-4502) and aprender-79's ONT-9c8695dcf602383872e04557e33402a(clean). Local pre-check (not a gate):cargo check --workspacegreen. Known and unchanged:check_no_claim_literalshas 3 FAILs from ont-10-only spec docs, pre-existing on ont-10.Conflict resolution
Merge
8810fdf4a9(#4429 @08bde31912 into ONT-10). Hunks: 139 — pure copy→ontology 45, by hand 63, generated→regen 31.CLAUDE.md:1Cargo.toml:1Makefile:1README.md:1README.md:2README.md:3README.md:4ci/sections.yml:1contracts/alibi-slopes-v1.yaml:1contracts/alibi-slopes-v1.yaml:2contracts/alibi-slopes-v1.yaml:3contracts/census.json:1contracts/census.json:2contracts/census.json:3contracts/census.json:4contracts/census.json:5contracts/census.json:6contracts/contracts.nt:1contracts/contracts.nt:2contracts/contracts.nt:3contracts/contracts.nt:4contracts/contracts.nt:5contracts/contracts.nt:6contracts/contracts.nt:7contracts/contracts.nt:8contracts/contracts.nt:9contracts/contracts.nt:10contracts/contracts.nt:11contracts/contracts.nt:12contracts/contracts.nt:13contracts/contracts.nt:14contracts/lint-baseline.json:1contracts/lint-baseline.json:2contracts/lint-baseline.json:3contracts/lint-baseline.json:4contracts/ontology.yaml:1contracts/ontology.yaml:2contracts/ontology.yaml:3contracts/ontology.yaml:4contracts/ontology.yaml:5contracts/ontology.yaml:6contracts/shapes.ttl:1contracts/shapes.ttl:2crates/apr-cli/src/extended_commands.rs:1crates/aprender-build-sha/src/lib.rs:1crates/aprender-build-sha/src/lib.rs:2crates/aprender-build-sha/src/lib.rs:3crates/aprender-build-sha/src/lib.rs:4crates/aprender-contracts-cli/src/commands/lint.rs:1crates/aprender-contracts-cli/tests/ont4b_shapes_gate.rs:1crates/aprender-contracts-cli/tests/ont4b_shapes_gate.rs:2crates/aprender-contracts-cli/tests/ont4b_shapes_gate.rs:3crates/aprender-contracts-cli/tests/ont4b_shapes_gate.rs:4crates/aprender-contracts-cli/tests/ont6_lint_verdict.rs:1crates/aprender-contracts-staging/lean/ProvableContracts/Defs/Rand.lean:1crates/aprender-contracts-staging/lean/ProvableContracts/Theorems/Alibi/Slopes.lean:1crates/aprender-contracts-staging/lean/ProvableContracts/Theorems/Alibi/Slopes.lean:2crates/aprender-contracts-staging/lean/ProvableContracts/Theorems/Alibi/Slopes.lean:3crates/aprender-contracts-staging/lean/ProvableContracts/Theorems/FFT/Fft2d.lean:1crates/aprender-contracts-staging/lean/ProvableContracts/Theorems/FFT/Fft3d.lean:1crates/aprender-contracts-staging/lean/ProvableContracts/Theorems/FFT/FftBatched.lean:1crates/aprender-contracts/src/lint/mod.rs:1crates/aprender-contracts/src/lint/mod.rs:2crates/aprender-contracts/src/lint/mod.rs:3crates/aprender-contracts/src/lint/mod.rs:4crates/aprender-contracts/src/lint/mod.rs:5crates/aprender-contracts/src/lint/mod.rs:6crates/aprender-contracts/src/lint/mod_tests.rs:1crates/aprender-contracts/src/lint/mod_tests.rs:2crates/aprender-contracts/src/lint/mod_tests.rs:3crates/aprender-contracts/src/lint/mod_tests.rs:4crates/aprender-contracts/src/lint/mod_tests.rs:5crates/aprender-contracts/src/lint/ratchet_gates.rs:1crates/aprender-contracts/src/lint/ratchet_gates_tests.rs:1crates/aprender-contracts/src/lint/shapes_gate.rs:1crates/aprender-contracts/src/lint/shapes_gate.rs:2crates/aprender-contracts/src/lint/shapes_gate.rs:3crates/aprender-contracts/src/lint/shapes_gate.rs:4crates/aprender-contracts/src/lint/shapes_gate.rs:5crates/aprender-contracts/src/lint/shapes_gate.rs:6crates/aprender-contracts/src/lint/shapes_gate.rs:7crates/aprender-contracts/src/lint/shapes_gate.rs:8crates/aprender-contracts/src/ontology/extract/claims.rs:1crates/aprender-contracts/src/ontology/extract/code.rs:1crates/aprender-contracts/src/ontology/extract/mod.rs:1crates/aprender-contracts/src/ontology/extract/mod.rs:2crates/aprender-contracts/src/ontology/extract/mod.rs:3crates/aprender-contracts/src/ontology/extract/mod.rs:4crates/aprender-contracts/src/ontology/extract/mod.rs:5crates/aprender-contracts/src/ontology/extract/mod.rs:6crates/aprender-contracts/src/ontology/mod.rs:1crates/aprender-contracts/src/ontology/sigma.rs:1crates/aprender-contracts/src/ontology/sigma.rs:2crates/aprender-contracts/src/ontology/sigma.rs:3crates/aprender-contracts/src/ontology/sigma.rs:4crates/aprender-contracts/src/ontology/sigma.rs:5crates/aprender-contracts/src/ontology/sigma.rs:6crates/aprender-contracts/src/ontology/sigma.rs:7crates/aprender-contracts/src/ontology/sigma.rs:8crates/aprender-contracts/src/ontology/sigma.rs:9crates/aprender-contracts/src/proof_status_tests.rs:1crates/aprender-contracts/src/tree.rs:1crates/aprender-orchestrate/tests/integration_test.rs:1crates/aprender-present-terminal/src/bin/ptop.rs:1crates/aprender-ptx-debug/src/bin/main.rs:1crates/aprender-simulate/src/cli/args.rs:1docs/audits/impl-GH-4175-receipt.md:1docs/audits/surface_audit.csv:1docs/audits/surface_audit.csv:2docs/audits/surface_audit.csv:3docs/audits/surface_audit.csv:4docs/audits/surface_audit.csv:5docs/audits/surface_audit.csv:6docs/roadmaps/roadmap.yaml:1docs/roadmaps/roadmap.yaml:2docs/roadmaps/roadmap.yaml:3docs/roadmaps/roadmap.yaml:4docs/roadmaps/roadmap.yaml:5scripts/check_ont_ratchet.sh:1scripts/check_ont_ratchet.sh:2scripts/check_ont_ratchet.sh:3scripts/check_ont_ratchet.sh:4scripts/check_ont_ratchet.sh:5scripts/lib/model_ladder_cells.py:1scripts/lib/model_ladder_cells.py:2scripts/tree_reader_tests.txt:1scripts/tree_reader_tests.txt:2scripts/tree_reader_tests.txt:3scripts/tree_reader_tests.txt:4tests/fixtures/ont/valid-under-appendix-b/ontology.yaml:1tests/fixtures/ont/valid-under-bad-qualifier/ontology.yaml:1tests/fixtures/ont/valid-under-empty/ontology.yaml:1tests/fixtures/ont/valid-under-no-kernels/ontology.yaml:1tests/fixtures/ont/valid-under-nonkernel-bad/ontology.yaml:1tests/fixtures/ont/valid-under-ok/ontology.yaml:1tests/fixtures/ont/valid-under-ratchet-rise/ontology.yaml:1tests/fixtures/ont/valid-under-undeclared-key/ontology.yaml:1tests/fixtures/ont/valid-under-unknown-world/ontology.yaml:1tests/fixtures/ont/valid-under-world-not-a-string/ontology.yaml:1Follow-up commits after the merge (each its own commit, listed by hand)
crates/aprender-ptx-debug/src/bin/main.rs(d3fa68a)crates/aprender-ptx-debug/src/run.rs(d3fa68a).versionguard moved into the car's libraryanalyze_ptx_filescripts/dogfood_surfaces.sh(faff3aa)bin-aprender-simulate--aprender-simulate-v1.yaml,surface_audit.csv,binary-contracts/README.mdunserved:rows ported onto the renamed filescrates/aprender-contracts/src/ontology/extract/claims.rs(47d1982)guard-cargo-stepsmanifest: anX-stepsjob counts when X is selected and runsci_guards.sh X(case test added); otherwise the car'ssetsid --waitREADME/CLAUDE commands trip PV-ONT-012crates/aprender-core/build.rs(30c8283)evidence/binary/snapshot.jsonlREADME.md