Repair Phase 20 fixed-point regeneration and E4 ownership - #30
Repair Phase 20 fixed-point regeneration and E4 ownership#30kmccleary3301 wants to merge 54 commits into
Conversation
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 626aa26218
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".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
Exact-head Standards and Spec reviews passed with zero P0/P1/P2/Q. All GitHub checks passed; no activation or score claim.
NS01: canonical Harness Definition model and validation
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 2599e33b81
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".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| if candidate_code != 0: | ||
| return candidate_code, results |
There was a problem hiding this comment.
When a scratch candidate reaches a placeholder-bearing stage and then fails, expanded_argv() reloads the readiness module against the candidate workspace, but this early return leaves that module cached after the temporary directory is removed. _fixed_point() then snapshots a watch set whose SCORE_AUTHORITY_PATH still points into the deleted candidate, so the authoritative live score file can be omitted from the failure report. Despite the earlier review reply, fresh evidence in c998661 shows this return still occurs before any reload under the restored BB_WORKSPACE_ROOT; restore the readiness module in a finally before returning.
Useful? React with 👍 / 👎.
NS02: effective Harness Lock and explanation compiler
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: e659e69f98
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".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| env_gates[gate_ids[0]] = {"env_name": str(env_name), "gate_id": gate_ids[0], | ||
| "required": True, | ||
| "satisfied": bool(metadata.get("env_satisfied", False))} |
There was a problem hiding this comment.
When multiple effective values reference the same env_name, each iteration overwrites the same env.<name> entry, so conflicting env_satisfied metadata is resolved solely by lexicographic path order. For example, an unsatisfied a value and a satisfied z value for the same variable produce a satisfied gate, while renaming the paths can reverse the result; reject inconsistent states or combine them conservatively so the lock cannot incorrectly report a required credential as available.
Useful? React with 👍 / 👎.
…odules NS03: composable Harness authoring modules
1 participant
Summary
Follow-up to #28. Repairs the Phase 20 clean-checkout fixed point without weakening gates:
Review corrections
Two-axis Standards/Spec review found and this stack corrects: external-history deletion bypass, duplicated Phase 11 request parameters, incomplete clean-checkout CI, leaked internal SDK schema exports, todo alias drift, and split todo-presentation ownership. Final Standards and Spec re-reviews are clean. All bb-0bv tasks are closed after green product-spine, full E4 battery, and dual clean-checkout fixed-point evidence;
mainbranch protection now requires the product-spine and independent fixed-point comparison contexts.Compat reviewed: non-breaking
Stack
Verification
pytest -q tests/e4_parity: 716 passed, 1 skipped, 1 xfailednpx tsc --noEmitintui_skeleton: passnpm testinsdk/ts: 3 passed against the default servernpm run build && npm testinsdk/ts-kernel-contracts: 9 passedMerge gate
Green.
mainrequiresproduct-spineandCompare independent fixed-point checkouts; strict up-to-date checks and conversation resolution are enabled.