fix(sync): route question/permission replies by the request's own session directory
Answering a question tool (or a permission prompt) could leave the session permanently stuck on "asking question": resolveDirectoryForBlockingRequest returned the containing child-store key, which only proves containment. For a worktree session (or any session whose record is grouped under a parent project store), the reply was addressed to the parent directory's OpenCode instance, where the pending request does not exist - the server answered QuestionNotFoundError, the local request was removed, and the trailing question-tool part stayed running with no recovery until Stop. Resolve the directory from the request's own session record (server- confirmed ownership: session.directory, then project.worktree) before falling back to the containing store key. When a reply/reject comes back not-found, also enqueue the settled-running-tool tail materialization so the tool part converges to the server's actual state instead of leaving the UI stuck. Refs OPE-236
This commit is contained in:
@@ -255,6 +255,10 @@ Examples of global-store updates performed in `session-actions.ts`:
|
||||
- `deleteSession()` / `deleteSessions()` -> wait for server confirmation or `404`, then remove the session and its persisted state
|
||||
- `moveSessionToDirectory()` -> move the session between directory stores and update the global directory index
|
||||
|
||||
### Blocking-request (question/permission) reply routing
|
||||
|
||||
`respondToQuestion`, `rejectQuestion`, `respondToPermission`, and `dismissPermission` route the reply through `resolveDirectoryForBlockingRequest`. The directory chosen decides which OpenCode instance resolves the pending request, so it must be the **session record's own server-confirmed directory** (ownership), never the containing child-store key (containment): a project store legitimately holds its worktree sessions, and a reply addressed to the parent instance makes the server answer `QuestionNotFoundError` while the question stays pending in the worktree instance — the session is then stuck on the running question tool with no recovery. When a reply/reject comes back not-found, the stale request is removed locally and a `settled-running-tool` tail materialization is enqueued so the trailing tool part converges to the server's actual state instead of leaving the UI on "asking question" forever.
|
||||
|
||||
### Restore (unarchive) contract
|
||||
|
||||
The OpenCode server cannot clear `time.archived` over HTTP: `session.update`
|
||||
|
||||
Reference in New Issue
Block a user