Skip to content

Adopt behavior coverage and independent risk review - #20

Draft
david-hoffman wants to merge 2 commits into
codex-mainfrom
codex/adopt-asd-coverage-policy
Draft

david-hoffman wants to merge 2 commits into
codex-mainfrom
codex/adopt-asd-coverage-policy

Conversation

@david-hoffman

@david-hoffman david-hoffman commented Oct 7, 2026 •

Copy link
Copy Markdown
Owner

Adopt the owner-selected behavior coverage and independent risk review policy, with no numerical coverage threshold. Setup/intake settle or inherit the policy before test design; reviewers audit concrete behavior/risk omissions and classify findings before correction. Future explicitly selected targets remain binding, and other active tasks retain their pinned gates and allowances. This adapts dstack PR 7 to the installed agentic software delivery framework.

Valid reports with uncovered statements or branches now pass full verification and platform aggregation. Percentages remain diagnostic. Failed/skipped tests, tool failures, omitted sources, exclusions, malformed or contradictory details/counters, and invalid shard evidence still fail. Validation preserves negative branch exits and checks missing-detail uniqueness and executed/missing disjointness.

Dependent CI jobs also use one shared exact Python version map: Linux 3.10.22; macOS/Windows 3.10.11. A floating 3.10 selector allowed collection/shards to use 3.10.21 while Linux aggregation used 3.10.22 in the original run. The pins repair that environment drift while preserving exact receipt identity, job structure, the platform/shard matrix, permissions, and release controls.

Candidate: 9979d62be06859e1e8cc733b4cbc8af1f14a736e.

Validation:

  • Coverage implementation: 177 focused verifier cases passed on 47bd585; runtime, tests, lock, and coverage configuration are unchanged by the interpreter-pin follow-up.
  • CI repair: two existing real aggregation/foreign-input rejection tests passed in 8.78 seconds; YAML parsing and semantic comparison passed.
  • Canonical fast, whitespace checks, and local documentation-link checks passed. Installed commit/push hooks also run fast.
  • Fresh independent coverage-application and interpreter-pin reviews accepted their scoped candidates with no findings. The task's Current state is refreshed afterward only to record outcomes.

Full-product/hosted verification on the updated candidate remains pending; no merge/publication approval or verified branch-protection claim is supplied. Prior defects, checks, and owner authorization remain in docs/tasks/ASD-COVERAGE-POLICY-001.md; complete logs/review reports and final candidate identity stay outside the tracked tree.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant