Repository navigation
Description
User Story
As a Python repository maintainer, I want SpecFact to discover and prepare my project's runtime automatically, so capsule review analyzes my actual code and dependencies without requiring SpecFact-specific packaging or repository patches.
Problem and verified scope
Addresses #472. Code Review 0.49.85 launches its capsule but ordinary reviews have no consumer runtime handoff; the current reserved import catalog also conflicts with common project packages and pytest policy rejects ordinary plugin controls. The original customer PR remains inaccessible to the implementation identity; customer logs are reported evidence, not an independently reproduced result.
Delivery contract
- Automatically discover and prepare pip/pip-tools, Hatch, uv and Poetry project runtimes, preserving constraints, locks, environment selection, source roots and test configuration.
- Add
specfact code review runtime inspect --json,runtime prepare --json, and review--project-config/--project-runtimeoptions. - Add a local, content-addressed project-runtime-layer-v2 descriptor; preserve v1 and protected PR authority boundaries.
- Separate sealed control code from target dependency and plugin execution; diagnose genuine incompatibility and native-library failures without altering project pins.
- Keep independent static findings when preparation fails, with one root diagnostic and incomplete dependent evidence instead of cascading import findings.
- Validate real collection/execution and normal consumer GitHub Actions without clearing GITHUB_ACTIONS.
Acceptance
- Pinned Requests (pip), Hatch, Flask (uv), and Poetry checkouts complete applicable analysis and real test slices on non-root Ubuntu 24.04, Python 3.11/3.12/3.13.
- A clearly labelled reconstruction covers [Bug] Capsule review fails closed on external Hatch projects without project-runtime handoff #472's detached Hatch src layout, third-party/native imports and pytest plugin configuration.
- Cold preparation, offline warm reuse, descriptor attachment, source immutability, genuine defect detection and corrupt/incompatible runtime rejection are proven.
- Relevant PRs require the external corpus; the published signed release repeats the same acceptance.
- Required quality, signature, Requirements, and SpecFact review gates pass; actionable PR findings are fixed.
- [Bug] Capsule review fails closed on external Hatch projects without project-runtime handoff #472 stays open until its original customer reproduction is verified. A reconstructed fixture does not substitute for that evidence.
OpenSpec and governance
Change: code-review-16-portable-project-runtime.
Branch: codex/feature-code-review-16-portable-project-runtime; PR targets dev.
Parent: Feature #163 under Epic #162. Blocks resolution of #472.
Related baseline only: #466, #459, #175. No open prerequisite has been identified.
Labels: enhancement, module, codebase, openspec, change-proposal. Assignee: djm81. Project: SpecFact CLI #1. Priority: High.
Implementation explicitly authorized on 2026-09-14 (Europe/Berlin). Follow spec → tests → recorded failure → production implementation → passing evidence. Initial platforms are Linux x86-64, CPython 3.11–3.13. Use 90-minute corpus job ceilings and record actual resource cost. Rollback uses reviewed revert and new signed publication; prior release remains affected by #472.
Metadata
Metadata
Assignees
Labels
Type
Projects
- StatusShow more project fieldsDone
Activity
fix(code-review): isolate native project environments and execution (#…
fix(code-review): preserve isolated member execution and source bound…
fix(code-review): verify portable runtime boundaries and offline reuse (
test(code-review): provide private devices for offline corpus (#473)
djm81 commentedon Sep 14, 2026
Protected delivery now continues in #475 (supersedes #474 while preserving its complete history). The exact final CI-signed source tree is carried in a single GitHub-verified commit from dev; fresh required checks and all 15 external corpus combinations are running in https://github.com/nold-ai/specfact-cli-modules/actions/runs/34907939755. Final local smart/full suites each passed 2,593 tests; independent review found no new defects. The stronger corpus guard must confirm correction of the demonstrated Poetry/Pylint namespace false findings. Core follow-up nold-ai/specfact-cli#730 has verified retained RED/final proof and awaits the final signed/merged module fixture. Public 0.50.0 acceptance remains pending; #472 still requires the original customer reproduction.
feat(code-review): deliver portable project runtimes (#473) (#475)
docs(code-review): deliver authenticated runtime parity with retained…
fix(code-review): preserve portable runtime and native pytest coverage (
release: publish Code Review 0.50.1 portable project runtimes (#477)
feat(code-review): add project-driven native macOS ARM64 runtime (#489)