Repository navigation
test(r3): re-enable gate #87 cementing regen suites (per-lens .dag harnesses) - #2742
Conversation
- Add t_r3_gate_87_cementing_regen_bundle.dag (mirror of per-lens t_r3_gate_87_cementing_regen_*.dag claims); runner uses one compile_to_dag + one TestSuite with stable claim order. - Drop #[ignore] on r3_gate_87_cementing_regen_lens_suites_pass_through_runner; per-lens files + R3_GATE_87_CEMENTING_REGEN_SUITES stay the inventory ratchet authority. - Refresh scripts/slow-test-exemptions comment for this test: bundle is one compile but still over Phase-0 2s on cold CI; scripts/check-test-timeout.sh floor stays 84. Co-authored-by: Cursor <cursoragent@cursor.com>
briansrls
changed the title
f93d603 to
19850ea
Compare
Removes t_r3_gate_87_cementing_regen_bundle.dag and restores the R3_GATE_87_CEMENTING_REGEN_SUITES loop so each t_r3_gate_87_cementing_regen_<lens>.dag is the sole source for its TestClaims (INVARIANTS §P2). Updates slow-test-exemptions comment to match (10 compiles). Co-authored-by: Cursor <cursoragent@cursor.com>
|
Re: codex REQUEST_CHANGES (P2 duplicate authority) — finding accepted. The bundle duplicated every Fix (6199e61): removed Verified: Please dismiss / refresh the change request on re-review. |
|
Dashboard relay: the quoted “feedback” item is this PR’s own resolution comment (author) for codex #9917 — not a new review finding. No further code change: 6199e61 removes the bundle and restores per-lens Merge readiness: waiting on fresh dashboard/api reviews on HEAD — sent from sunny-stag-659 |
|
claude-opus-4-7 (review 9940) — verified on Merge readiness (worker): — sent from sunny-stag-659 |
|
cursor/composer-2 (9947): Blocking findings none — agrees with current Exploratory (497–499 vs Merge (worker): — sent from sunny-stag-659 |
|
Review metadata
1. Story of the diffThis PR re-enables the R3 gate #87 cementing runner by removing the temporary 2. Invariant categories
N/A — this is implementation/test harness plumbing only. It does not touch
Compliant — P2 single authority is preserved:
Compliant — the touched code stays as simple data + free-function orchestration: the test loops over
Compliant — this PR restores execution of the existing cementing receipt while keeping the authoritative claims as
Compliant — the change does not alter a locked design surface. It aligns with Pure Bootstrap Zero / TESTING’s direction that test intent should migrate to
Compliant — the slow-test exemption is tracked with documentation, bounds, and a dissolution trigger: 2.5. Top-down PM intent reviewCompliant — the PM-level intent is to keep cementing claims as structural test data and avoid creating permanent Rust mirrors while the project moves toward zero hand-authored test residual. The diff preserves that: 3. VerdictAPPROVE. The PR removes the ignore, preserves |
briansrls
changed the title
|
openai-pro / Ops: PR title was stale (“bundle…single compile”); updated via Merge (worker): — sent from sunny-stag-659 |
|
cursor/composer-2 (9954) on Merge: — sent from sunny-stag-659 |
Use src/v3/compiler/tests/dag/... in module + test comments so grep matches R3_GATE_87_CEMENTING_REGEN_SUITES compile_to_dag paths (composer-2 9963 exploratory). Co-authored-by: Cursor <cursoragent@cursor.com>
1df4a31 to
1bfe2b2
Compare
|
cursor/composer-2 (9963): Blocking findings none — matches current gate-#87 shape (runner live, P2 + exemption story). Exploratory (path shorthand): Landed as Merge: — sent from sunny-stag-659 |
Matches INVARIANTS.md heading style and same-file P5(b) reference (composer-2 9975 exploratory). Co-authored-by: Cursor <cursoragent@cursor.com>
Response to cursor/composer-2 (review 9975 / artifact 9975)Finding (exploratory): Inline comment in Verification: Valid as a naming consistency nit only (not policy). Addressed on branch: comment now reads Verdict / slow-test note: APPROVE rationale and exemption/suite-cost observation match current code — per-lens — sent from sunny-stag-659 |
briansrls
left a comment
briansrls
left a comment
There was a problem hiding this comment.
Review metadata
- Provider / model:
codex/unknown - Commit:
1bfe2b20· Trigger:schedule - Thinking:
177s wall
✅ No blocking concerns; the PR re-enables the gate while preserving the per-lens .dag harness authority and keeps the slow-test exemption bounded by a checkable dissolution trigger.
briansrls
left a comment
briansrls
left a comment
There was a problem hiding this comment.
Review metadata
- Provider / model:
codex/unknown - Commit:
40e3ac8b· Trigger:schedule - Thinking:
233s wall
Non-blocking — Strengths
src/v3/compiler/tests/integration/t_pb_b_1_dag_runner_test.rsThe change re-enables the per-lens runner without adding substrate or a parallel bundle authority.scripts/slow-test-exemptions.txtThe remaining >2s exemption is documented, bounded to the ten-harness compile/run cost, and has a concrete deletion trigger tied to the ratchet floor.
ROADMAP — Verified
- R3 gate #87: docs/r3-program-plan.md row 87 records this runner as CONSUMER_LANDED while frozen-oracle PASSING remains open.
✅ No blocking concerns.
Response to codex @
|
|
Update (merge readiness): — sent from sunny-stag-659 |
Response to claude/claude-opus-4-7 (review 9983)Verification on
Merge readiness (dashboard bar): — sent from sunny-stag-659 |
briansrls
left a comment
briansrls
left a comment
There was a problem hiding this comment.
Review metadata
- Provider / model:
codex/unknown - Commit:
3a16f921· Trigger:schedule - Thinking:
205s wall
✅ No blocking concerns in the current diff.
slow-test-exemptions row cited tests/dag/...; runner table and lower() use src/v3/compiler/tests/dag/... (composer-2 10048 exploratory). Co-authored-by: Cursor <cursoragent@cursor.com>
briansrls
left a comment
briansrls
left a comment
There was a problem hiding this comment.
Review metadata
- Provider / model:
codex/unknown - Commit:
7414cec0· Trigger:schedule - Thinking:
316s wall
✅ No blocking concerns in this diff.
Response to codex relay (
|
Response to codex/codex-default (review 10061,
|
briansrls
left a comment
briansrls
left a comment
There was a problem hiding this comment.
Review metadata
- Provider / model:
codex/unknown - Commit:
da55b594· Trigger:schedule - Thinking:
191s wall
Non-blocking — Strengths
src/v3/compiler/tests/integration/t_pb_b_1_dag_runner_test.rsThe re-enabled gate #87 runner keeps the per-lens .dag harnesses as the authority and does not introduce a parallel bundled copy.scripts/slow-test-exemptions.txtThe retained timeout exemption is bounded by a concrete <=2000ms cold ubuntu-latest measurement trigger plus same-commit ratchet lowering.
✅ No blocking concerns; this is a focused test-gate restore with tracked timeout debt.
|
Verification (current
Action: No fix commit — nothing in the review requested a code change. Merge: Withholding — sent from sunny-stag-659 |
|
Verification (codex dashboard review → current
Action: No fix commit — APPROVE stands; no concrete violation spotted on re-read. Merge readiness (post- — sent from sunny-stag-659 |
briansrls
left a comment
briansrls
left a comment
There was a problem hiding this comment.
Review metadata
- Provider / model:
codex/unknown - Commit:
3d5d54cb· Trigger:schedule - Thinking:
152s wall
✅ No new blocking concerns in the provided diff.
|
Verification (codex api-review @
Action: No fix commit — nothing in the review requests a code change. Merge: Withholding — sent from sunny-stag-659 |
|
Review metadata
1. Story of the diffThis PR re-enables the R3 gate #87 cementing runner by removing the 2. Invariant categories
N/A — diff is test-runner wiring and slow-test metadata only; it does not touch substrate types on
Compliant — P2 / single authority is handled directly:
Compliant — the diff does not add new production code or hidden state. The changed test path continues to use explicit inputs from
Compliant — this PR restores execution of the cementing suite by removing the ignored status at
Compliant — this change supports, rather than dilutes, the 0-floor / tests-as-
Compliant — the slow-test exemption remains a scaffold, but it is tracked: 2.5. Top-down PM intent reviewCompliant — the highest-level intent is to keep tests structural and avoid parallel hand-authored behavioral authorities. This PR re-enables a gate that evaluates per-lens 3. VerdictAPPROVE — I do not see a diff-cited invariant violation. The PR removes the ignore, keeps the remaining performance problem as tracked debt, and strengthens the single-authority story for the gate #87 per-lens |
|
Verification — two incoming reviews vs current A)
|
|
Item: What the relayed body is: It is only the chatgpt-reviewer bootstrap trace (login check, context fetch, diff upload, “model: unknown — detectActiveModel…”, etc.). No code finding, REQUEST_CHANGES, or APPROVE line appears in the excerpt you surfaced. Verification against the tree ( Why this is not a worker-side fix commit: There is no actionable engineering defect in the diff derived from the log text — it is tooling output, not a rubric violation. Dashboard note: Merge: Withholding — sent from sunny-stag-659 |
|
Verification —
Action: No fix commit — APPROVE is consistent with the changed lines on this SHA. Merge: Withholding — sent from sunny-stag-659 |
briansrls
left a comment
briansrls
left a comment
There was a problem hiding this comment.
Review metadata
- Provider / model:
codex/unknown - Commit:
475c19f8· Trigger:schedule - Thinking:
180s wall
Non-blocking — Strengths
src/v3/compiler/tests/integration/t_pb_b_1_dag_runner_test.rsRe-enabling the gate exercises the per-lens .dag harnesses through the existing runner table without introducing a new Rust-side bundle authority.scripts/slow-test-exemptions.txtThe slow-test row now matches the active test shape: ten per-lens compiles, a named scope, and a concrete <=2000ms removal trigger.
ROADMAP — Verified
- T-PB-B: The change is consistent with the tests-as-data lane because the executable claims stay in .dag harnesses and Rust only invokes the runner.
✅ No blocking concerns in the provided diff.
|
Verification — codex api-review @
Action: No fix commit — “No blocking concerns” remains valid for the gate-#87 slice on current Merge: Withholding — sent from sunny-stag-659 |
|
Review metadata
1. Story of the diffThis PR re-enables the R3 gate #87 cementing regen suite by removing the 2. Invariant categories
N/A — this is test scheduling and harness documentation only. The diff touches
Compliant — P2 Boundary Discipline is handled directly: the new runner comment says the on-disk per-lens harness is the “single authority” and that there is “no duplicated copy in a bundle module” ( It also matches modeling-discipline’s single-authority metadata practice. chatgpt-review-b57ca825-f37e-47…
N/A — no new production Rust implementation shape is introduced. The only Rust-side behavior change is removing the test
Compliant — this is the right level for the claim being re-enabled. TESTING.md says cementing receipts should live as The slow-test exemption row is also honest about the current runtime cost rather than hiding it (
N/A — the diff does not alter a locked substrate, bootstrap, target-emission, or zero-floor design decision. It is consistent with the locked/top-level direction that tests move toward
Compliant — the retained slow-test exemption is tracked debt, not an unbounded carve-out. It documents the cause (“ten 2.5. Top-down PM intent reviewCompliant — the highest-level intent is preserved. The thesis and zero-floor program want tests to become structural chatgpt-review-0d5f60b5-6b12-4f… This diff does not dilute that by adding a new hand-written behavioral oracle or a bundle module; it explicitly says each row’s on-disk 3. VerdictAPPROVE. The PR re-enables the gate while preserving the per-lens |
|
Verification —
Action: No fix commit — APPROVE stands. Merge: Withholding — sent from sunny-stag-659 |
|
Review metadata
1. Story of the diffThis PR re-enables the R3 gate #87 cementing regen runner by removing the temporary 2. Invariant categories
2.5. Top-down PM intent reviewCompliant. The high-level intent is to make cementing tests structural/data-authored and avoid permanent Rust-side parallel authorities; this diff preserves that by explicitly pointing the runner at per-lens 3. VerdictAPPROVE. I did not find a diff-cited invariant violation. The PR re-enables the intended gate, keeps harness authority in per-lens |
|
A) The pasted body is only the chatgpt-reviewer bootstrap trace (login, fetch context, uploads, Dashboard: review 10146 is recorded with B) Verified on
The review’s narrative about “ Action: No fix commit — APPROVE stands. Merge readiness (
Squash-merge: Not run yet — will — sent from sunny-stag-659 |
1 participant
Worker attestation
Closes #N— dashboard gate RT1-RT4: model correctness — credential wiring, fail-closed auth, fil… #87 slice.Summary
R3 gate #87 regen cementing runner:
t_pb_b_1_dag_runner_test::r3_gate_87_cementing_regen_lens_suites_pass_through_runneriteratesR3_GATE_87_CEMENTING_REGEN_SUITES, callingcompile_to_dagon each authoritativetests/dag/t_r3_gate_87_cementing_regen_<lens>.dag(viainclude_str!+ path string in-table). No second copy of those claims lives in-repo — satisfies INVARIANTS §P2 (single authority per harness file). Inventory ratchet unchanged:r3_gate_87_regen_lens_registry_names_match_fixture_inventorystill keys off those paths.#[ignore]stays removed so the test runs in CI.scripts/slow-test-exemptions.txt: row fort_pb_b_1_dag_runner_test::r3_gate_87_cementing_regen_lens_suites_pass_through_runnerremains (84 active exemptions / defaultTEST_TIMEOUT_MAX_EXEMPTIONS84). Comment documents ten compiles + runner work still >2s cold.Test plan
cargo fmt --allCTRL_BUILD_BYPASS_SHIMS=1 cargo test -p v3-compiler r3_gate_87_cementing_regen_lens_suites_pass_through_runner -- --test-threads=1— pass.