Skip to content

test(r5): orchestrate isolated native feature companions - #7102

Draft
antuz123 wants to merge 306 commits into
sprintfrom
codex/r5-native-feature-orchestration-20261022
Draft

antuz123 wants to merge 306 commits into
sprintfrom
codex/r5-native-feature-orchestration-20261022

Conversation

@antuz123

Copy link
Copy Markdown
Collaborator

Wires the separately reviewed native companions in order: existing unused-key/PIN → controlled malformed credential → controlled stale/empty history → genuine Vault references. Fixed Metro configs, owned process groups and fresh successful stopApp/JUnit boundaries prevent fixture code running alongside the ordinary app. Distinct host receipt ports and final ordinary-bundle restoration preserve fixture ownership. Earlier raw product/Retry receipts remain separate.

Exact head 196bfeded03705315871eb5c8d1b693540110581, branch codex/r5-native-feature-orchestration-20261022. Orchestration-only review. Approximately11–13k changed code/test tokens for one orchestration area, after splitting producer features. Preserve #7088 dd5fca6, #7093 b8abebb and #7097 b6cb88c as merge ancestors, with their patch-only links. No producer rewrite.

Independent review found nested companion final-save/cleanup failures could hide the original failure and skip disposal. Separate follow-up196bf preserves primary exceptions, visits every cleanup stage, disposes private UI output before final evidence and fails closed on cleanup errors. Actual-companion negative cases cover UI+save/server/purge failures, successful UI followed by cleanup failure, and initial-save failure. No app/acceptance failure was waived.

Exact source: 124Node /125Python controls PASS, actionlint3/whitespacePASS. Root independently reran9orchestration+3actual-companion cleanup tests and worker actionlintPASS. Source/runtime review completed; no formal/provider approval claimed. Maestro runScript metrics execute on the CLI host, confirmed against the pinned archive; they are not native app HTTP requests.

Native execution NOT RUN. Current app/native execution is separate immutable3eb on #7082, not this source. Same standard macos-15/ubuntu-24.04 runners and one build/device per platform; global240min fail-closed ceiling allows the expanded suite. Planning range100–150min/platform is unmeasured; not all nested maxima fit. No larger runner, paid activation, binary upload/distribution or job dispatch. New source needs exact-source native execution and combined checks. This is not 5.2 and does not identify/interrupt it.

Limits remain explicit: malformed JSON is not OS/hardware denial; controlled history is not authentic historical migration; Vault archive deletion is logical archival with reinstatement, not secret-byte inspection; helper/loopback scopes are not backend/physical/release acceptance. Full parent matrix remains blocking.

Codex local handoff

State: NEEDS_LOCAL
Base: sprint
Head branch: codex/r5-native-feature-orchestration-20261022
Head commit: 196bfed
Goal and acceptance criteria: Run separately scoped native companions with trustworthy fixture/process/JUnit ownership and preserved failures.
Changes made: Fixed owned Metro orchestration, companion order, port isolation, partial-receipt retention and robust cleanup; producer commits retained.
Cloud checks:124Node/125Python/actionlint3/whitespacePASS; root9+3targeted controlsPASS. IOS/ANDROID NOT RUN.
Required local checks: IOS, ANDROID, HARDWARE
Local checks requested: Exact combined candidate on hosted macOS/Xcode and disposable Android emulator; actual authorized devices for physical checks. No personal Mac dependency.
Risks / known gaps: New native flows/timings unverified; current3eb native failures/retry belong to that SHA; full matrix/5.2/physical/installed trust and release readiness remain open. No merge/shared promotion/release signing/upload/distribution/deployment/publication.

This branch has not been deployed

No deployments
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.

3 participants