* fix(mac): move device auth tokens to shared SQLite state
* fix(mac): fail identity migration when source auth vanishes mid-read
Missing-file tolerance now applies only to the first observation; a disappearance after the auth file was observed fails the migration so the claimed identity survives for retry instead of committing without its credentials.
* fix(ios): validate auth scope UTF-8 encoding
* fix(ios): isolate SQLite auth profiles
* chore(ios): refresh native source inventory
* fix(mac): keep device auth in the shared token table
* test(mac): pin shared-key token cache semantics
* feat(chat): show live run telemetry across clients
* fix(chat): satisfy native validation gates
* fix(chat): settle advertised terminal runs
* fix(android): clear stale run usage on gaps
* chore(i18n): align native source inventory
* fix(chat): reconcile telemetry with current projections
* refactor(chat): keep run state in owner extensions
* chore(i18n): refresh native source inventory
* test(ui): expect persistent working telemetry through streaming
* fix(mac): drop stale native identity claim when key material matches
* fix(mac): acquire stale identity claim before deletion
* fix(mac): park stale identity claim instead of deleting it
Rescue path now contains no destructive unlink: the acquired claim is parked as .stale-<uuid> and validated before any continue path, closing both replacement races review found.
* fix(mac): restore parked claim when the rescue source vanishes
Review round 4: accepted the vanished-source regression (restore before continue so claim-without-source stays importable); rejected serializing the source acquisition - rescue imports nothing and post-rescue source writes follow the standard atomic claim-then-import path (invariant documented inline).
* style(apple): satisfy device identity SwiftFormat
* fix(thinking): fall back instead of erroring on unsupported adaptive level (#109351)
The adaptive level is the auto sentinel (normalizeThinkLevel maps auto ->
adaptive) meaning 'pick an appropriate level', not 'force this exact level'.
Some clients (e.g. the Apple Watch client) send it as a transport default
while the iOS app sends none, so hard-erroring on an unsupported adaptive
override rejected otherwise-valid turns from those clients. Now the auto
sentinel always falls back to a supported level; explicit concrete levels
(e.g. an explicit /think xhigh) still hard-error as before.
Closes#109351
* docs(proof): add real-behavior-proof screenshot + run log for #111301
Captured run of the real get-reply-run.media-only test (101 passed)
showing the adaptive-thinking fallback fix verified.
Co-Authored-By: yuvrajlaptop2008-byte <284412773+yuvrajlaptop2008-byte@users.noreply.github.com>
* fix(ios): inherit Watch chat thinking defaults
Co-authored-by: yuvraj thakur <284412773+yuvrajlaptop2008-byte@users.noreply.github.com>
* chore(ios): sync native i18n inventory
---------
Co-authored-by: yuvrajlaptop2008-byte <284412773+yuvrajlaptop2008-byte@users.noreply.github.com>
Co-authored-by: Peter Steinberger <steipete@gmail.com>
* fix(macos): bound websocket ping so a dropped pong cannot orphan its continuation
URLSession drops the pong handler entirely when a web socket task is cancelled or
closed mid-flight. The existing gate guarded a double resume but nothing guarded
zero resumes, so the checked continuation leaked, Swift logged CONTINUATION
MISUSE, and keepaliveLoop's await never returned - wedging the keepalive for the
lifetime of the process.
Race the ping against a deadline and reuse the gate so exactly one resume wins.
Cancellation of the deadline returns without resuming, otherwise a delivered pong
would race the gate and surface as a spurious timeout.
* fix(macos): reuse URLError(.timedOut) instead of a new public ping error
Periphery flags a public enum referenced only inside its own module as dead
code in the shared OpenClawKit intersection scan. URLError keeps the timeout
indistinguishable from a transport timeout for callers, which already handle
URLSession errors from every other path, and adds no public surface.
* fix(ios): honor speech locale for system voice
* chore(ios): refresh talk locale i18n inventory
* fix(ios): fall through unavailable system voices
* chore(ios): refresh talk locale i18n inventory
---------
Co-authored-by: Colin Johnson <colin@solvely.net>
Co-authored-by: Peter Steinberger <steipete@gmail.com>
* fix(state): name the refusing install in schema and exec-approval refusals
The newer-schema refusal told operators not to downgrade and to upgrade
OpenClaw, neither of which is actionable when two builds share one release
version string. It now names the install root that refused, both schema
versions, and warns that a linked source checkout reports its git HEAD even
when its built dist is older.
The exec-approvals gate told operators to run `openclaw doctor --fix` without
naming the state directory, so a bare invocation repaired the default root
while the scoped install stayed blocked. Both the TypeScript gate and its
Swift sibling now scope the command to the blocked state directory.
Refs #115008
* fix(gateway): name the refusing install in the startup schema refusal
* fix(mac): keep the exec-approvals gate message buildable on iOS
* fix(exec-approvals): shell-quote the state directory in the repair command
* fix(exec-approvals): state the repair directory in prose so every shell can follow it
* fix(ios): forgetting a paired gateway does nothing
Confirming "Forget Gateway" in Settings never removed the gateway. The
confirmation dialog derived its isPresented binding from the payload it
later needed and cleared that payload on dismissal, so by the time the
destructive button's action ran, forgetPendingGateway() found nil at its
opening guard and returned without calling forgetGateway(stableID:).
Nothing was logged either, because the failure branch was never reached.
Hand the entry to the action through the dialog's presenting: parameter
instead of reading it back from view state, and take it as a parameter
rather than looking it up. Everything downstream is unchanged; it simply
runs now.
* Shorten the dialog comment to one line
The three-line version repeated in prose what the next line already shows.
What is worth keeping is the trap: someone simplifying this back to reading
the entry from view state reintroduces the bug.
* fix(ios): harden forget gateway payload handoff
Co-authored-by: Cihat Gündüz <anthropic@fline.dev>
* test(ios): run gateway settings source guards
* test(ios): allow forget dialog state cleanup
* test(ci): provide rg in PR review fixture
---------
Co-authored-by: Cihat Gündüz <anthropic@fline.dev>
Co-authored-by: Peter Steinberger <steipete@gmail.com>
* fix(android): honor reduced motion reactively in mascot and talk waveform
OpenClawMascot and TalkWaveform read Settings.Global.ANIMATOR_DURATION_SCALE
with a one-shot remember(context){} that never reacts when the user toggles
"Remove animations" while the screen stays mounted. Extract the reactive
ContentObserver read that ChatWorkingIndicator already used into a shared
ui/SystemAnimations.kt helper and use it in all three surfaces. Follow-up to
#112245 (Wear avatar).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(android): resync the native i18n baseline after the animation refactor
The native-i18n check failed with "native app i18n inventory drift detected".
apps/.i18n/native-source.json pins a line number for each of its 5290 entries,
and 23 of them point into ChatWorkingIndicator.kt. Extracting the shared
rememberSystemAnimationsEnabled helper shortened that file by 26 lines, so every
recorded string below the edit moved and the inventory no longer matched.
Regenerated with `pnpm native:i18n:baseline`, as the failure message instructs.
The diff is 23 line-number updates (384 -> 358 and so on); no `source` value
changes, because this refactor adds and removes no user-facing strings.
`node --import tsx scripts/native-app-i18n.ts verify` now passes.
* fix(android): reuse Compose motion scale
Co-authored-by: Masato Hoshino <g515hoshino@gmail.com>
* fix(android): keep system motion scale authoritative
Co-authored-by: Masato Hoshino <g515hoshino@gmail.com>
* test(cli): raise root help deadlock guard
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-authored-by: Peter Steinberger <steipete@gmail.com>
CPP was added to blockedEverywhereKeys in #112560, but its Autoconf twin
CXXCPP was not. Autoconf defines the preprocessor selectors as a pair —
AC_PROG_CPP sets CPP, AC_PROG_CXXCPP sets CXXCPP — and this list already
carries the compiler pair CC/CXX for the same reason.
Measured on current main: CC, CXX and CPP all report dangerous, while
CXXCPP lands in no category at all (blockedPrefixes is only DYLD_, LD_
and BASH_FUNC_, so no prefix rule covers it either).
* perf(gateway): warm handler families and cache identity avatars
* chore: drop changelog edit (release-generation-owned)
* fix(ci): regenerate protocol models, satisfy knip, absorb sidecar drift
- commit regenerated Swift GatewayModels for the additive cron.list param
- split the avatar data-url cache into assistant-avatar-cache.ts so its
factory has a production consumer (knip production scan)
- de-export prewarm family names/type; test now drives fake families only
- postReadySidecarCount 2 -> 3: main added a post-ready sidecar in parallel
