Repository navigation
docs: map fork behavior for V2 convergence - #49
nullStack65 wants to merge 3 commits into
Conversation
STARTPOLSYNC:BUILDER — START / CHECKPOINT (publishing the existing local pending receipt) Kameron approved the bounded managed-policy sync on November 2, 2026. Scope: merge current pingdotgg/t3code main into branch Exact inputs: fork HEAD The local merge initially had 141 conflicted paths. The first native Go resolution attempt was stopped before commit because bulk upstream selection did not establish semantic fork preservation. Its candidate is not accepted. A fresh read-only Go review found 12 repair groups: continuation/refusal integrity, governed Codex wiring/catalog/readiness, release provenance/workflow compatibility, owner-only reviewer tests, usage attribution/cache correctness, title preference, post-start receipts, launch preflight, Windows service and macOS profile wiring, CI mirror/runner consistency, and fixture output drain. Owner-only reviewer code, fork release helper defaults, relay owner admission, worktree defaults and attribution persistence were found retained; these must not be weakened to make tests pass. Observed verification: A single bounded native OpenCode Go repair worker owns this worktree; the parent T3 session retains foreground collection, verification and publication. Existing V2 adaptation commit Policy feasibility: the upstream schema can set/lock a Codex model-selection default, restrict provider drivers, allow remote/Connect, set a full-access default, and direct desktop updates to a generic HTTPS feed. It cannot select a ChatGPT account/plan/workspace, directly name a GitHub release repository, pin CLI/npm updates, or force a minimum permission mode for every existing thread. The fork's latest release currently lacks generic-updater manifests. Exact limitations will appear in the policy README and implementation PR body; no fields or credentials will be invented. Next owned action: finish faithful V2 source repair and focused tests, independently verify the exact candidate, then create/link the sync and policy PRs with required provenance. No PR merge, deployment, key rotation, purchase, other-human review request, or installed-policy change occurred. Overall task remains IN PROGRESS. Model: codex/gpt-6.1-sol |
The requested backlog cleanup needs V2 source on fork main without losing the fork's existing behavior, release isolation, or production admission. The held V2 tree currently diverges from main and does not preserve several of those protections.
This draft is a planning carrier only. It records exact refs, a grouped path/commit preservation matrix, independently owned source stacks, minimal sequencing, and concrete missing evidence. The temporary draft file must be removed or replaced by reviewed implementation before landing. No code convergence, test pass, deployment, release, or installed acceptance is claimed.
The triggering request is to resolve the whole PR backlog against each PR's original intent and quality goals. This draft does not authorize closing any useful PR and does not change existing ownership. The Git fetch needed to build and test a combined tree is blocked by the cloud environment's failed VPN enrollment (HTTP 503); connector readback remains available.
Model/harness: GPT-6.1, Codex cloud.