* fix(workboard): make ownership updates atomic
Persist launch intent before worker acceptance, reconcile accepted identities,
fence cross-host card mutations, and deduplicate session capture in SQLite.
Closes#125932
* test(workboard): keep race fixtures plugin-local
Use a plugin-local synchronization helper and update the sessions-page capture
expectation for the new store-owned RPC.
* fix(workboard): fence lifecycle and restore races
Apply terminal lifecycle state only while the matched association is current,
and converge concurrent archived session restores on the active winner.
* feat(ui): collapse automation sessions in activity feed
* feat(ui): show live run status on activity feed rows
* feat(ui): make run inspector deep-link only
* fix(ui): only link inspect run to a current active run
Addresses ClawSweeper P1: a stale observer digest could deep-link a live
row to an ended run's audit record; trust digest runId only when it is
still in activeRunIds.
runConfigUnset scoped its missing-path error to --dry-run, so the real
invocation printed "No change" and exited 0 while --dry-run on the same
input exited 1 with a specific message. Dry-run exists to predict the
real run, and a typo'd unset in a script reported success.
Restore the failure for both modes, keeping the strict-config assertion
introduced alongside the regression and the JSON5 byte preservation it
was added to guarantee.
* fix(ui): never restore a stored draft over typed new-session composer text
After a reload the new-session composer renders before the gateway
recovery scope arrives; selectRoute then zeroes the draft revision, so
text typed in that window lost to the stored draft's higher revision and
the async restore overwrote (or, interleaved with input, appended to)
what the user had typed. Seen on PR #125690 CI (checks-ui-e2e 5/12, run
32122284238) as a doubled composer value.
NewSessionDraftPersistence now records a pristine baseline at each
programmatic content replacement (page reset, navigation handoff,
restore apply) and restores only when no user mutation happened past it;
otherwise the typed text wins and persists above the stored revision,
including minting a revision for text typed before route activation so
that draft still lands. The chat composer sibling is unaffected: its
revision lineage persists synchronously to localStorage.
* test(ui): run draft-persistence regression in the isolated ui lane
The regression test mocks the statically imported
durable-composer-persistence helper (the store runtime is only
dynamically imported, so under the shared isolate:false module graph its
mock alone is packing-dependent). Isolating the file makes the mocks
deterministic and keeps the shared-lane file set unchanged so sibling
packing does not shift.
Recheck the request signal after an async response hook settles so a concurrent abort cannot reach response body handling. Related to #125802 and #125834.
Co-authored-by: Onur Solmaz <2453968+osolmaz@users.noreply.github.com>
Keep drain-owned recovery state authoritative when the pre-restart run reaches a normal terminal event, so startup continues the original request without requiring a manual retry.\n\nRelated: #57425
`openclaw approvals allowlist add|remove` printed "Writing local approvals."
from the shared target-resolution helper, before the mutation decision was
made. Both idempotent paths ("Already allowlisted.", "Pattern not found.")
returned without saving, so the CLI announced a write that never happened.
`approvals set` had the same problem: it announced the write and then rejected
unparseable input.
Move the announcement from `loadWritableSnapshotTarget` into the local branch
of `saveSnapshotTargeted`, the function that owns the write. Every caller of
the shared seam is fixed at once and exit codes are unchanged: idempotent add
and remove still leave the requested end state satisfied and exit 0.
* test(ui): seed mock activity sessions across owners
* feat(ui): consolidate activity filters into toolbar
Also restores the shared .input styling contract for standalone form controls.
* test(ui): update activity feed e2e for toolbar people popover
The Gateway owns start-or-steer at admission (6515f6a255) and no
client produces expectedRunId anymore (d84a910fc8). The field shipped
only in v2026.8.1-beta.2 - never a stable tag - so it is removed rather
than deprecated. Steer sends resolve the selected session's current
operation; the exact-match branch, the operation|run target identity
discriminator, run_mismatch rejection, and the suggestion producers'
active-run-id selection (with its ambiguity failure) are deleted.
Provider-native turn fencing (Codex expectedTurnId) is unchanged:
the backend-captured runId on the injection target remains.
Preserve browser availability and profile ownership across node and extension routes, close session-owned node tabs, honor screenshot refs, and hide unavailable tab-bound actions.
The retryable-send and retryable-history tests asserted the resend side
effect immediately after `vi.advanceTimersByTimeAsync(100)`. Sinon's async
tick yields exactly one real macrotask after firing the last timer, and the
retry chain (`void scheduleStoredChatOutboxDrain` -> `chat.history` ->
`chat.send`) consumes exactly that one boundary, so the assertions had zero
slack and flaked under CI contention.
Wait for the outcome with `waitForFast` instead, and fold the outbox-empty
assertion into the wait since retirement lands after the send call.
* fix(gateway): harden hook admission and shutdown
Flush structured 413 and 408 hook responses before closing oversized or timed-out request sockets, and reject deferred wake requests that target an explicit session.
Direct Gateway close now marks startup and readiness as draining immediately and attempts every HTTP listener shutdown before reporting an incomplete close.
* test(gateway): model request destruction in fuzzing
Give the HTTP request fuzz double the IncomingMessage destroy contract exercised after 413 and 408 responses finish.
* fix(gateway): preserve response and shutdown bounds
Route installed webhook body-limit failures through the shared response-first close owner and require complete Feishu 413/408 responses before terminal connection closure.
Retain the current-main grace bound for agent-harness disposal so one stuck harness cannot block listener teardown.
* test(telegram): model response-close request cleanup
Keep the slow-body request alive through response finish, then assert cleanup when the mock response transport closes.
Patrick Erichsen