The +1px emulated viewport changed the CSS box, so a terminal sitting one
pixel under an xterm row boundary gained a row. The pane fit observer's
two-frame stability check reads that transient grid as stable well inside
the 32ms hold, forwards a real PTY resize, then reverses it on restore —
two SIGWINCHes per reveal, measured on 3/54 window heights (~1/cellHeight).
Nudging the device scale factor instead re-runs layout with byte-identical
CSS geometry. Same reflow, 0/54 SIGWINCH, no WebGL atlas rebuild or context
loss, and webview guests stop seeing spurious native resizes too.
Co-authored-by: Orca <help@stably.ai>
* fix(window): restore the macOS 26 reflow without touching the native frame
#10253 stopped the main-thread deadlock by skipping the repaint size nudge on
macOS 26, but invalidate() repaints without reflowing, so the h-dvh root kept a
stale viewport height and the status bar stayed clipped off-screen (STA-2383) on
every Tahoe reveal, restore and wake.
Drive the reflow through device emulation instead: a +1px emulated viewport,
reverted a frame later, makes the renderer recompute layout without any NSWindow
mutation, so the FrontBoardServices re-entrancy that wedged the main thread for
109 minutes is still never triggered.
Verified against real Electron 43.1.0 on macOS 26.3.1 (Darwin 25.3.0): the
renderer sees the resize and relayouts, the native frame is untouched, the
viewport and devicePixelRatio restore exactly across zoom levels, overlapping
calls collapse to one cycle, and a 1px delta never crosses a terminal cell
boundary so no pane reports new geometry (no SIGWINCH to running shells).
Also close two gaps in the surrounding code:
- the pre-Tahoe size jiggle now clears its WeakSet latch in a finally block, so a
throwing setSize can no longer suppress every later repaint for that window
- cover powerMonitor 'resume' under the Tahoe guard, the other AppKit dispatch
context implicated in the freeze
* fix(tray): keep NSStatusItem scene updates off the AppKit callout stack
The main-window repaint was only one of the two doors into the macOS 26
FrontBoardServices deadlock. Showing or restoring the window calls
setTrayAttention(false) straight from the window event handler, and
tray.setImage/setToolTip drive an NSStatusItem scene update — the same
re-entrant scene mutation from inside AppKit's own dispatch, matching the
stackshot in openai/codex#23695.
Defer the native mutation to a fresh event-loop turn so the callout frame is
vacated first. The attention flag itself still flips synchronously: rapid
show/hide would otherwise mis-dedupe against a value that had not landed yet.
Bursts collapse to a single repaint, and because the deferred pass reads current
module state rather than a captured value, a coalesced schedule can never apply
a stale icon. applyTrayImage already no-ops on a destroyed tray, so a repaint
still queued when the tray goes away is harmless.
* fix(window): reflow maximized and fullscreen windows on macOS 26 too
The maximized/fullscreen bail-out predates the Tahoe path and exists only to
keep the size nudge from resizing a window out of those states. Emulation never
touches the frame, so that guard was suppressing the reflow for no reason — and
a maximized window strands its dvh layout exactly like a normal one.
Run the Tahoe branch before the guard. Verified on macOS 26.3.1 that the
emulated viewport reflows a maximized and a fullscreen window while leaving both
states intact.
* fix(window): retry the viewport restore instead of stranding the renderer
If disableDeviceEmulation threw while the webContents was still alive, the
previous code swallowed the error and cleared the latch anyway, leaving the
renderer pinned at the emulated 1px-taller viewport for the rest of the window's
life — and letting the next reveal stack a fresh cycle on top of it.
Retry the restore on a bounded schedule and hold the latch while a retry is
pending. A destroyed webContents still short-circuits, since the emulated
viewport dies with it, and the attempt budget keeps a permanently failing
restore from pinning the latch forever.
* fix(terminal): resume hibernated agents that reattach with no payload
When agent hibernation is on, a stopped (done) agent's PTY is killed and a
passive sleeping record is kept; returning to the worktree relies on the pane
reattaching on remount. On the daemon path (Windows/local worktrees) the daemon
can reattach the hibernation-killed session as already-live (isReattach,
isNew:false) and return no snapshot/replay/coldRestore — and, being a reattach
rather than a fresh spawn, it silently drops the --resume command passed on
connect. The renderer adopted that empty session, leaving a blank terminal with
nothing running and the sidebar history still pointing at the dead tab (the
sleeping record never cleared).
A reopened pane that owns a resumable slept session must always be re-driven
with its resume command, never left as a bare empty attach. handleReattachResult
now discards a contentless isReattach for a pane with a hibernation record and
re-drives the prepared resume. This excludes the healthy cases: a fresh session
the daemon created (isReattach falsy — it already ran the command) and a live
reattach (carries a snapshot/replay). Forward the isReattach signal the
transport was dropping so the two cases are distinguishable.
Adds a deterministic regression test that fails without the guard and passes
with it.
* fix(terminal): preserve provider ownership on resume
* chore(skills): refresh generated skill bundle manifests
Regenerate the skill bundle artifacts against the full release-tag set so
the freshness verify check passes. Append-only additions for the newer
release tags; no released snapshot history is rewritten.
* feat(codex): surface a stalled config sync instead of failing silently
Why: the mirror keeps serving the last synced settings when ~/.codex/config.toml
is missing, blank, or unreadable. That is the right call for data safety, but it
is invisible — a downed WSL distro or an unhydrated cloud-synced home leaves
"Orca ignores my config edits" with no log line and no UI to diagnose.
Status is derived on demand from the same predicates the mirror uses, so the two
cannot disagree. The stall is logged once per episode rather than on every launch
and quota poll, and the Codex account section names the file and what to do.
* fix(codex): latch an unreadable source and stop over-claiming recovery
An unreadable source throws out of the mirror, so reporting only on the success
path left that stall latch-less: it logged the raw failure on every launch and
quota poll while its reason never reached the surfaced status. Report from the
catch path too.
The clear message also claimed the source was "readable again", which is false
when the stall ended because the runtime config was removed rather than because
the source came back.
Restoring console.warn now happens in afterEach — an inline mockRestore is
skipped by a failing assertion, and the leaked spy made every later case in the
block fail spuriously.
* fix(codex): latch the stall promotion hits first, and scope it to the host
Review round 1 findings:
- The unreadable-source latch still never fired in the steady state. Once a
baseline exists, promotion reads the source before the mirror does, so it
throws first and `!promotionPlan` returned before any reporting — logging a
reasonless failure every launch and quota poll, which is exactly what the
previous commit claimed to fix. Report from that branch too. The test only
passed because its fixture had no baseline; it now seeds one first and fails
without the fix.
- The banner named the host's ~/.codex while a WSL or per-account runtime was
selected, whose real source is a different file entirely. Gate it to the host
scope, matching how the sign-in warning is already gated.
- Three new translate keys were missing from the locale catalogs, failing the
localization gate in `pnpm lint`.
- The registrar mock was never asserted, so deleting the registration left the
suite green.
- `codexConfigSyncStatus` hung off the `agentHooks` namespace despite having
nothing to do with agent hooks; moved to its own `codexConfigSync.status`
while it is still a four-file change.
* fix(codex): report sync health for the home the selection actually mirrors
Review round 2:
- The status resolved the shared runtime home, but the system default now runs
Codex directly against ~/.codex and managed accounts get their own home. So a
stalled per-account mirror showed no banner at all, while a stale shared home
could warn about a config the active lane never reads. Resolve the mirrored
home from the current selection, and report synced when the lane has no mirror
to fall behind.
- The round-1 report on the promotion failure path could clear the latch on a
pass where no mirror ran, claiming a recovery that never happened and
silencing every later pass. Only ever latch a stall there; leave clearing to
the path that actually mirrored.
* fix(codex): refetch sync status when the active Codex account changes
Review round 3:
- Resolving the status per selection made the fetch account-dependent, but the
effect was not keyed on the active account. Switching accounts left the banner
describing the previous one — and switching INTO a stalled account showed
nothing at all, which is the silence this change exists to remove.
- Pin the home resolution itself: it had no direct test, and its shared-home
path was a hand-copied literal that could drift from the real helper and
silence the banner with every other test still green.
- Narrow the handler's dependency to the one method it calls, which also drops
an `as unknown as` cast from its test.
- Skip the chmod-based test on Windows, where a read-only directory does not
block writes so the scenario cannot be constructed; matches the convention
already used in config-settings-promotion.test.ts.
* chore(codex): restore the handler docstring and isolate the resolver suite
Round 4 returned clean; these are its two non-blocking nits.
Narrowing the handler param left its JSDoc stranded above the new type, so the
function had no hover doc. The resolver suite also read the developer's real
CODEX_HOME and shell rc, so anyone exporting one would see it fail locally.
* fix(ssh): re-arm remote file watches when the provider reconnects
Remote file changes stopped being detected over SSH until the file was
reopened. The watch pipeline itself was fine — nothing ever re-established
the subscription after the transport it was made on went away.
Two paths left an editor tab permanently stale:
- A reconnect kills the relay's watch registrations, and the previous
provider's unwatch handle belongs to the dead transport.
- A connect slower than the 60s retry window made installRemoteWatcher
give up for good; first deploy to a new host far exceeds that.
Neither recovered, because installRemoteWatcher is only reachable from the
fs:watchWorktree handler, the retry timer, and the removal-restore path,
and the renderer only issues a watch for newly added targets. Reopening
the file just re-read it — the watcher stayed dead.
Give the layer that owns the transport the job of re-arming:
registerSshFilesystemProvider now notifies subscribers, which covers both
establish and reconnect since registerProviders runs on both. The watcher
keeps the intent to watch in a registry that outlives any single
connection, and on registration drops the stale entry (installRemoteWatcher
treats an existing entry as installed and would otherwise hand back a
watcher that can never fire), reinstalls, and emits overflow so consumers
resync the gap. Intent is dropped on unwatch and sender destroy so a
closed tab is never resurrected.
Verified over SSH to a Rocky Linux 10 host: after the reconnect that
previously killed it, a remote append lands in the editor in ~2s with a
live remote watcher process, and a second edit in ~1.5s.
* fix(ssh): drop watch intent when a remote worktree is removed
A removed worktree kept its entry in the intent registry, so a reconnect
landing before the renderer's unwatch would re-watch a deleted path —
60s of retries against the host and then a bogus overflow.
Also covers two reinstall cases: several senders on one connection must
collapse onto a single relay watch (and all of them resync), and a
destroyed renderer must not be reinstalled.
* fix(ssh): resync when a reconnect's watch only lands on a retry
The reinstall emitted the overflow only for listeners whose first install
returned 'installed'. If that attempt failed — relay-watcher.js still
spawning on the fresh transport, a transient fs.watch rejection — the 1s
retry restored the watch but never signalled the gap, so everything that
changed while the transport was down stayed invisible: the STA-2525
symptom reappearing inside the fix for it.
Thread the resync intent through the retry record so the overflow lands
when the retry does. All pre-existing callers default to false, so the
watch/terminal-error/restore paths are unchanged.
* test(ssh): cover the resync merge when a fresh watch claims the retry slot
The `resyncOnInstall ||=` merge was uncovered: removing it left every
existing test green. It is load-bearing — if a second renderer joins the
reinstall's failing install and reaches the retry slot first with
resync=false, the whole chain stays false and the retry restores the
watch without ever signalling the gap.
New sibling file rather than an append: filesystem-watcher.test.ts is
~10 counted lines from the 800-line lint cap, and AGENTS.md forbids a
max-lines disable.
* fix(ssh): re-arm a remote watch that died without a reconnect
The 60s fast-retry window gave up with a single overflow and no further
trigger — provider registration is the only re-arm, and a watch killed by
remote OOM/inotify exhaustion leaves the SSH link perfectly healthy. Back
off from 1min to a 30min ceiling instead, and stand down entirely when the
provider is gone so registration owns that case.
* fix(ssh): re-arm the remote file-explorer watch on reconnect
SshFilesystemProvider.dispose() stops each watch registration without
invoking its terminal callbacks, so a dropped transport left the runtime
file-explorer watch silently dead: the lease's restart only runs from the
failed-removal path, and the renderer's runtime subscription stays open
because it rides a different link. Reinstall on the connection's next
provider registration and emit overflow so clients resync.
* fix(editor): read out-of-worktree SSH paths without a stamped target id (#9743)
Route external absolute-path tabs by their resolved SSH connection instead of
the externalSshTargetId stamp, which only the terminal-link open path sets. The
stamp keeps its fail-closed role when present.
* fix(editor): keep client-local live-tail log tabs off the worktree SSH host
AI Vault "View Log" tabs are opened client-local by construction (readOnly +
liveTail, no runtime env), so inferring the external SSH owner from the
worktree connection made them read the remote host instead of the granted
client path. Restrict that inference to non-live-tail tabs; an explicit
externalSshTargetId stamp still routes remotely.
* fix(ssh): re-arm remote file watches when the provider reconnects
Remote file changes stopped being detected over SSH until the file was
reopened. The watch pipeline itself was fine — nothing ever re-established
the subscription after the transport it was made on went away.
Two paths left an editor tab permanently stale:
- A reconnect kills the relay's watch registrations, and the previous
provider's unwatch handle belongs to the dead transport.
- A connect slower than the 60s retry window made installRemoteWatcher
give up for good; first deploy to a new host far exceeds that.
Neither recovered, because installRemoteWatcher is only reachable from the
fs:watchWorktree handler, the retry timer, and the removal-restore path,
and the renderer only issues a watch for newly added targets. Reopening
the file just re-read it — the watcher stayed dead.
Give the layer that owns the transport the job of re-arming:
registerSshFilesystemProvider now notifies subscribers, which covers both
establish and reconnect since registerProviders runs on both. The watcher
keeps the intent to watch in a registry that outlives any single
connection, and on registration drops the stale entry (installRemoteWatcher
treats an existing entry as installed and would otherwise hand back a
watcher that can never fire), reinstalls, and emits overflow so consumers
resync the gap. Intent is dropped on unwatch and sender destroy so a
closed tab is never resurrected.
Verified over SSH to a Rocky Linux 10 host: after the reconnect that
previously killed it, a remote append lands in the editor in ~2s with a
live remote watcher process, and a second edit in ~1.5s.
* fix(ssh): drop watch intent when a remote worktree is removed
A removed worktree kept its entry in the intent registry, so a reconnect
landing before the renderer's unwatch would re-watch a deleted path —
60s of retries against the host and then a bogus overflow.
Also covers two reinstall cases: several senders on one connection must
collapse onto a single relay watch (and all of them resync), and a
destroyed renderer must not be reinstalled.
* fix(ssh): resync when a reconnect's watch only lands on a retry
The reinstall emitted the overflow only for listeners whose first install
returned 'installed'. If that attempt failed — relay-watcher.js still
spawning on the fresh transport, a transient fs.watch rejection — the 1s
retry restored the watch but never signalled the gap, so everything that
changed while the transport was down stayed invisible: the STA-2525
symptom reappearing inside the fix for it.
Thread the resync intent through the retry record so the overflow lands
when the retry does. All pre-existing callers default to false, so the
watch/terminal-error/restore paths are unchanged.
* test(ssh): cover the resync merge when a fresh watch claims the retry slot
The `resyncOnInstall ||=` merge was uncovered: removing it left every
existing test green. It is load-bearing — if a second renderer joins the
reinstall's failing install and reaches the retry slot first with
resync=false, the whole chain stays false and the retry restores the
watch without ever signalling the gap.
New sibling file rather than an append: filesystem-watcher.test.ts is
~10 counted lines from the 800-line lint cap, and AGENTS.md forbids a
max-lines disable.
Add closeBeforePress flag to Rename, Browser, and Refresh actions to
defer modal opening until the action sheet closes. Eliminates the race
condition that caused the mobile app to freeze when opening these modals.
#10340 added a ledger-advance step to the release cut, which violates the
contract test #9119 added: the cut must not run generate-skill-bundle-manifest
or stage resources/skills. Both PRs were green on their own branches and only
conflicted once merged, so nothing failed until main had both — main and every
open PR have been red since.
Revert the two workflow lines. The script's --release implementation stays: it
is correct and harmless when unused, and the root fix in #10340 — verify no
longer walking git tags — does not depend on the cut step.
This leaves #10340 semantically incomplete and that must not be dropped. With
the registry seeded from the committed ledger instead of a tag walk, nothing
advances the ledger at cut, so each new skill change re-uses the same unreleased
tail revision for different bytes; older installs then match no known snapshot
and degrade to unrecognized, which reads in the UI as a skill that needs
attention and cannot be updated. Follow-up is to reintroduce the advance
narrowly — stage only resources/skills/release-mapping.json and narrow the
assertion to forbid mutating the content-addressed artifacts while permitting
the provenance row.
* fix(rate-limits): keep Codex PTY reset text for weekly-only plans
The PTY /status fallback parses '5h limit' and 'Weekly limit' lines by
label, but the extracted reset text was only ever attached to the
session window. Codex plans without a 5h session bucket (e.g. current
Pro) produce a weekly-only parse, so the reset time the CLI printed was
silently dropped. Fall back to the weekly window when no session window
exists.
* review: parse Codex PTY reset text per window into resetsAt
* review: make Codex PTY status fallback work on codex >=0.145
* review: harden PTY status parse against model-scoped rows and styled output
* fix(rate-limits): strip private PTY control sequences
---------
Co-authored-by: Brennan Benson <79079362+brennanb2025@users.noreply.github.com>
* fix(skills): stop promising a skill update the command cannot deliver
An "Update available" badge could never clear: pressing Update ran
`npx skills update <name> --global`, which reported "All global skills are
up to date" and wrote nothing, while the badge stayed on.
Freshness marked a name updatable whenever ANY placement was outdated,
including a standalone duplicate in an agent home. The global command only
converges the canonical copy and its symlink aliases, so a stale duplicate
kept the badge lit with no command that could clear it.
- Count only reliably-convergent placements toward the update promise, so a
stale duplicate no longer advertises an update that cannot land.
- Give the duplicate chip a real skipped-reason instead of falling through to
the generic sentence, naming the copy and how to resolve it.
- Surface the existing freshness review dialog from the setup rails via a
Details link, so a blocked or duplicate copy is explainable where the user
actually sees the badge. Covers orchestration, Computer Use, Ephemeral VMs,
Linear, the CLI section, the floating orchestration modal, the Browser Use
card, and the Mobile Emulator row.
* fix(skills): say when a skill copy needs attention instead of reading as all-clear
A skill with an out-of-date copy the update command cannot reach rendered as a
green "Installed" pill. That is honest about the main copy but reads as
all-clear, so real drift in an agent home stayed invisible — the user had no
reason to suspect there was anything to click.
- Add a needs-attention display status (amber) for a placement that is not
current but has no eligible update: stale duplicates, edited copies,
read-only, inaccessible, and broken or external links. Presence-only stays
green, and an unscanned inventory stays quiet so nothing flashes amber on
launch.
- State the reason inline on the setup rails, so the cause is readable without
opening the review dialog; Details still opens the full per-location list.
- Extract the skipped-reason sentence into its own module so the rails and the
dialog share one source and can never drift apart.
- Carry the state through the settings sidebar badge so the nav and the card
cannot disagree.
* fix(skills): give the inline skill warning something to point at
The shared reason sentences are deictic — "this copy", "the copy here" — because
they were written for the review dialog, where the location rows they describe sit
directly beneath them. On the setup rails there are no rows, so "this" referred to
nothing and the sentence read as if it were about the skill itself.
Name the offending paths above the sentence on the rails, so the referent is
present before the wording that depends on it. Every copy sharing the blocking
reason is listed, not just the first, so resolving one does not leave the badge
unexplained. The dialog keeps its existing wording and rows unchanged.
* fix(skills): stop withholding the update over copies it cannot reach
Eligibility is now decided purely over the placements the global command actually
converges — the canonical copy and its symlink aliases.
A project skill, plugin cache, standalone duplicate, or unreadable copy in another
agent's home used to withhold the update from the whole name. `skills update
--global` provably never writes any of them, so that refused work the command could
have done over a copy that was never at stake. A blocked *convergent* copy still
withholds it: that is the placement the command writes to, and overwriting it is the
real data-loss case.
Also drop the inline reason from the setup rails. The reasons are written to sit
beside the location rows they describe, so on a card they had nothing to point at and
could only ever name one cause. The rails now mark Details with a warning icon when a
copy needs the user's own hands, and the dialog does the explaining with every
location and cause it knows about.
* fix(skills): make the skill Details affordance read as a control
The review link rendered as bare text, so nothing but hover said it could be
clicked — the badge told the user something was wrong and then gave them no visible
way in.
Use the ghost variant so it carries a hover/focus background and a real hit target
instead of a zero-padding text run, and add a chevron so it reads as a control at
rest. The chevron points right, not down: this opens the review dialog, while a down
chevron already means the in-place expander inside that dialog.
* feat(worktrees): copy project-level .worktreeinclude paths into new worktrees
Read .worktreeinclude at the repo root (gitignore syntax) and copy matching
gitignored paths from the primary checkout into each newly created local
worktree, so .env and other local config carry over with zero per-user setup.
- Literal patterns resolve by direct stat; globs match against
ls-files --others --ignored --exclude-standard --directory (collapsed
dirs keep huge repos fast); every candidate is re-verified with
check-ignore so tracked or unignored files are never copied.
- Copy semantics, never symlink: APFS clone-copy on macOS, real copy
elsewhere, so each worktree owns its files (unlike repo.symlinkPaths,
which it merges with rather than replaces).
- Failures never block worktree creation.
- Remote (SSH) creation skips it, same as symlinkPaths.
- Split APFS clone helpers into worktree-apfs-clone.ts (max-lines).
Closes#7549
* fix(worktrees): harden worktree include copying
* fix(worktrees): support nested includes on Git 2.25
* fix(worktrees): bound include copy costs
* fix(worktrees): close include correctness and perf gaps
* fix(worktrees): preserve included copy semantics
* fix(worktrees): harden include resolution
* fix(worktrees): preserve bounded include resolution
* fix(worktrees): bound include filesystem resolution
* fix(worktrees): harden include matching
* fix(worktrees): tighten include matching and scan bounds
* fix(types): use concrete filesystem stat types
* fix(worktrees): harden included path materialization
* perf(worktrees): stop include parsing at resolver budgets
* chore(skills): refresh bundled skill manifests
* refactor(worktrees): reduce .worktreeinclude to focused literal-only scope
The reviewed implementation grew well past the ticket (#7549), which asks for a
size-M feature that reuses existing worktree machinery. Trim back to the minimal
change that solves the reported problem safely:
- Resolver now supports literal files and directories only. Glob/negation lines
are skipped with a warning (documented follow-up), which removes the entire
user-controlled-regex ReDoS surface, the CPU/byte budgets, the git enumeration
scan, and the case-sensitivity engine. The filesystem + git check-ignore
handle existence and case for free.
- Copy layer folded back into worktree-symlinks.ts (link/copy modes share one
loop); dropped worktree-path-copy.ts, worktree-target-safety.ts, the
descendant-dedup/realpath/target-parent machinery, and the per-materialization
APFS filesystem cache. Kept the df/diskutil probe timeout.
- Reverted unrelated changes: check-ignored-paths timeout param and the
git-binary-compatibility enumeration tests.
Net: -1903/+172 across the include+copy code. Behavior for the ticket's cases
(.env, .env.local, .vscode/, node_modules, config/secrets.json) is unchanged;
gitignored-only + copy-not-symlink semantics preserved.
Closes#7549
* fix(worktrees): dereference symlinked .worktreeinclude entries + cache APFS volume probe
Two issues found by review + perf audit of the copy path:
- Correctness (HIGH): a listed entry that is itself a gitignored symlink was
copied AS a symlink (fs.cp dereference:false), and the darwin APFS branch was
skipped for all symlink sources. Editing the worktree's copy then wrote through
to the shared/primary target — inverting copy-mode's 'each worktree owns its
files' guarantee, and escaping the worktree entirely if the link pointed
outside it. Now resolve realpath for a top-level symlink in copy mode so we
copy content; nested symlinks inside a copied dir stay as-is (cp -R semantics).
- Perf: assertSameApfsVolume ran df+diskutil per copied path (4 subprocesses
each), so an N-entry include spawned ~4N short-lived processes on the macOS
create hot path, all re-probing one volume. Add a per-materialization
device-keyed cache: one probe per distinct volume (4N -> ~4).
Tests: symlinked-file and symlinked-dir dereference regressions (no leak to
primary); APFS volume probed once regardless of copied-path count.
* fix(win): taskkill plain-shell PTY trees on immediate teardown
Deleting a Windows worktree that still has a live process in a terminal
tab (a `pnpm i`, or any command that spawns a child tree) failed with
"Failed to physically stop every PTY for worktree" after the full 10s
teardown deadline.
Root cause: on Windows, closing a plain shell's ConPTY does not reap its
orphaned children — node-pty's `useConptyDll` skips the console-process
reap. A live `pnpm i`/`node` child survives the shell's exit, keeps the
ConPTY console non-empty (so the daemon still reports the session alive),
and holds the worktree cwd handle. The destructive-removal physical-stop
check then fails closed. #10100 fixed this for agent sessions
(killWithDescendantSweep -> taskkill /T /F) but plain shells were never
swept.
Fix: extend the Windows taskkill /T /F descendant tree-kill to non-agent
shells on the immediate/destructive teardown path, in both the daemon
(TerminalSessionTeardown) and local provider. Gated to win32 + immediate
with the same ownsRoot guard so a naturally-exited/recycled PID is never
signalled; POSIX shells keep reaching their child pgroup via forceKill
and are unchanged.
Reproduced and verified end to end via the Electron dev build: a worktree
with a live node child previously failed to delete after 10s; with the
fix the child tree is taskkilled, the directory is removed, and the
delete succeeds in ~1.8s.
* fix(win): claim plain-shell termination before the taskkill sweep
forceKillAndWaitForExit sets _isTerminating in its synchronous prologue.
Awaiting the Windows descendant sweep ahead of it left createOrAttach's
doomed-session guard open for the taskkill's duration, so a concurrent
attach could bind a pane to a session about to be tree-killed.
* fix(skills): source released history from the committed ledger, not a tag walk
verify:skill-bundle-manifest rebuilt the entire released-skill history by
walking every local refs/tags/v* on each run and demanded byte-equality with
the committed artifacts. Output was therefore a function of (skill bytes x
local tag set x release timing), so any clone holding stray, deleted, or fork
tags the committed artifacts predate rebuilt a divergent registry and failed
lint. This was the 4th instance of one failure class (#8637 -> #9119 version
bumps -> #9778 new tags -> local tag drift), each patched with a new tolerance
rather than removing the tag coupling.
Fix: the committed snapshot-registry + release-mapping ARE the released history;
trust them instead of re-deriving from tags.
- releasedHistoryFromCommitted() seeds generation from the committed ledger,
dropping the floating unreleased tail (entries beyond what the mapping names).
verify and --write are now pure functions of working-tree bytes with zero tag
access. The tag walk survives only behind --rebuild-from-tags (disaster
recovery), off the everyday path.
- --release <version> + appendReleaseRow() perform the O(1) append of one
mapping row at release cut (dedupes vs the last row, strips the v-prefix) --
the single authoritative point where working-tree bytes become an immutable
released revision.
- release-cut.yml runs generate --release "$VERSION" before the release commit
(Node built-ins only, no install needed); pr.yml drops fetch-depth: 0 from the
lint job since verify no longer needs tag history.
Recognition is unaffected: the runtime uses knownSnapshots = registry.skills
(all entries, incl. the tail committed at PR-merge time), so a missing mapping
row only loses a version label, never recognition or the update nudge.
Trade-off: lint no longer cross-checks committed historical snapshots against
tags. A hand-edit to an old released entry is still caught by the runtime
manifest<->registry consistency check when the current manifest points at it,
and can be audited anytime with --rebuild-from-tags.
Verified: verify passes committed-sourced; --write is zero-diff (byte parity);
a planted stray v-tag no longer changes output; edit-stub -> --write -> --release
appends the correct single row; double --release is idempotent;
--rebuild-from-tags reproduces the committed artifacts. Generator tests 14 pass/
1 skip; runtime skill-bundle-artifacts + freshness-inventory 14 pass; bundled
skill guides verify passes.
* fix(skills): keep one release-mapping row per version on a re-cut
A cut that pushed the version bump to main but died before pushing the
tag is re-cut at the same version. If skills changed in between, the
second --release appended a duplicate row, and the stale one named
revisions that tag never ships — which verify-skill-update-roundtrip
then pairs with the tag's real bytes.
Overwrite the trailing row instead (the tag is absent, so that version
was never published). Refuse only when an earlier row claims the
version, which the cut workflow already rejects upstream, so this
cannot wedge a recovering cut.
#9343 broke two contracts at once and #10437 fixed both, but neither is
asserted: #10429 repaired the failing tests by deleting the poll assertion and
by handing the retirement store a live repo, so a revert would land silently.
Restore `expect(getRepos).not.toHaveBeenCalled()` on the floating-tab poll, and
add a hydrate case for a store that cannot report repos. Verified both fail
against the pre-#10437 code: the poll assertion reports getRepos "called 2
times", and the hydrate case returns [] instead of the persisted tab.
* fix(codex): preserve runtime config without system source
* fix(codex): retain baseline when mirror is skipped
* refactor(codex): extract deprecated hook-flag normalization
Why: codex-config-mirror.ts sat at the 300-line cap, so the missing-source
guard could not land without a max-lines disable.
* fix(codex): bootstrap a baseline when the mirror is skipped
Why: a runtime home seeded outside the mirror (WSL, per-account) never got a
baseline while the source was missing, so promotion stayed inert and silently
reverted the in-Codex change once the source returned.
* fix(codex): stop a synthesized source config from wiping runtime settings
Two routes still reached the #9073 data loss after the missing-source guard:
- Promotion runs before the guard and, with no ~/.codex/config.toml, created
one holding only the promoted keys. The next mirror treated that skeleton as
authoritative and deleted every other runtime setting. It needs no missing
file: `codex mcp add` inside an Orca-launched Codex plus /model was enough to
drop the MCP server for good. Promotion now seeds a brand-new system config
from the runtime's ordinary settings, so the mirror round-trips them.
- A 0-byte source (half-written, or an unhydrated cloud-synced home) still read
as an authoritative empty config and advanced the baseline, making the loss
unrecoverable. A blank source is now treated like a missing one.
Moves the TOML section model out of codex-config-mirror.ts so promotion can
share it without a cycle.
---------
Co-authored-by: Brennan Benson <79079362+brennanb2025@users.noreply.github.com>
* fix(runtime): stop the headless hydration repo gate from dropping every tab
#9343 gated headless mobile-session hydration on the live repo list, but read
it as `this.store?.getRepos?.() ?? []`. A store that cannot report repos then
yields an empty set, which reads as "every repo is gone" and skips every
parseable session key — so no tabs hydrate at all. It also called getRepos on
every hydrate, including the hot floating-tab poll path that is contractually
free of repo/provider inventory work.
Resolve the inventory lazily and only for keys that parse to a repoId, and keep
`null` (unavailable) distinct from an empty list (all repos really gone). An
absent list now fails open; a known list still prunes as #9343 intended.
Fixes 3 tests that have been failing on main since #9343 landed:
orca-runtime-terminal-retirement (2) and orca-runtime (1).
* refactor(editor): extract the pending-focus effect to clear the max-lines cap
#8083 pushed RichMarkdownEditor.tsx to 404 counted lines against the 400-line
.tsx cap, failing lint for every PR that merges current main. The Explorer
find-focus request is self-contained, so it moves to its own hook.
* Fix fork PR/MR worktree creation race via durable review-head refs
When creating a fork PR/MR worktree, concurrent `git fetch origin` operations
clobber the shared FETCH_HEAD, causing the wrong commit to be checked out.
Fetch PR/MR heads into dedicated per-review refs (`refs/orca/pull/<N>`,
`refs/orca/merge-requests/<N>`) that persist and isolate each head from other
fetches. Gracefully keep the compare-base when the fetch fails but the local
ref already exists, avoiding silent fallback to the wrong branch on transient
network errors.
* Bound PR/MR head fetches with 60s timeout
Prevent PR/MR creation from hanging when a remote is stalled or
unreachable. Both GitHub and GitLab head fetches now enforce a
60-second timeout, matching the bound used in the create-path
fetch. Durable refs (refs/orca/pull/*, refs/orca/merge-requests/*)
decouple the ref from FETCH_HEAD, preserving legacy client semantics.
* test: align CI expectations with main PowerShell/sparse regressions
PR checks merge into main, which recently changed PowerShell launch args
(cwd restore after profiles) and sparse-checkout detection (require
core.sparseCheckout). Derive PowerShell spawn args from the production
resolver, mock the sparse config flag, reset shared worktree list scan
cache between tests, and stop requiring floating polls to avoid getRepos
hydration.
* Address review follow-ups on durable review-head refs
- Unify PR review-head remote selection: local and SSH GitHub paths share
resolveGitHubReviewHeadRemote, which prefers the remote mapping to the
hosting GitHub project (upstream before origin, matching work-item/API
candidate order) so contributor clones fetch refs/pull from the repo
that actually hosts the PR.
- Soft-keep durable review heads: when the PR/MR head fetch fails but
refs/orca/pull/<N> / refs/orca/merge-requests/<iid> still resolves,
keep the pinned SHA (warn) instead of failing resolve, mirroring the
compare-base fallback. Extracted shared compare-base soft-keep into
compare-base-ref-fetch.ts.
- Extract fetchGitLabMergeRequestHeadRef (local + SSH) parallel to the
GitHub helper; bound its local fetch with the shared 60s timeout.
- Share relay-style fetch validation (positive safe-integer id, remote
not starting with "-") between relay and local helpers via
review-head-tracking-ref.ts; move REVIEW_HEAD_FETCH_TIMEOUT_MS there.
- Drop the githubPullRequestHeadLocalRef re-export; resolve head SHAs via
rev-parse --verify <ref>^{commit}.
- Add GitLab anti-FETCH_HEAD regression test plus durable-head soft-keep
and remote-selection unit tests.
Co-authored-by: Orca <help@stably.ai>
* test: supply live getRepos for terminal-retirement hydrates
Main's headless tab hydrate (#9343) skips worktree keys whose repo is not
in getRepos. Retirement tests that rebuild mobile tabs from a persisted
session now advertise the fixture repo as live so PR Checks merge stays green.
* fix(editor): extract RichMarkdownEditor props to stay under max-lines
Main's SSH external-image wiring (#10323) pushed RichMarkdownEditor.tsx over
the 400-line tsx budget, failing PR Checks lint on every merge into main.
Move the props type into a sibling module so the component stays under the
limit without disabling max-lines.
* Make durable review-head refs remote-identity scoped
Embed remote name + URL hash into refs/orca/pull|merge-requests refs to prevent soft-keep from serving wrong project's PR/MR when FETCH_HEAD is clobbered by concurrent fetch. Fetch functions now return the written ref path (writer-authoritative) so callers rev-parse exactly what was fetched, not re-derive identity. Soft-keep only applies to transient errors (timeout, network); fails hard on missing refs, auth failures, and stale relay. Relay returns localRef so client avoids re-hashing (URL normalization can disagree).
---------
Co-authored-by: Orca <help@stably.ai>
* feat(agent-dashboard): choose in-window screen popover or pop-out window
The experimental Agent Dashboard opened only as a separate pop-out
window. Add an "Open as" mode under the experimental toggle so it can
open as an in-window screen popover (new default) or a pop-out window
(prior behavior). The mode row appears only when the feature is on.
- New setting `experimentalAgentDashboardMode: 'in-window' | 'popout'`
(default in-window); sidebar entry branches on it.
- In-window: AgentDashboardOverlay renders the shared AgentKanbanBoard
in a near-fullscreen dialog, snapshot built locally via
useLiveDashboardSnapshot (the pop-out relays over IPC; in-window has
no relay). Ack/reveal act on the local store — the pop-out IPC
handlers are gated to the pop-out renderer.
- AgentKanbanBoard gains containerClassName/onAckAgent/onRevealAgent/
onClose props; defaults preserve the pop-out behavior.
- Extracted AgentDashboardExperimentalSetting to keep ExperimentalPane
under the max-lines cap.
* fix(agent-dashboard): admit main renderer to terminal-preview IPC for in-window dialog
The terminalPreview:* handlers gated every channel to the pop-out
renderer, so the in-window overlay's terminal dialog (running in the
main renderer) got { snapshot: null } from connect and falsely showed
"No live terminal — this agent's pane has closed." for live agents.
Accept the trusted UI renderer too — it already has full PTY access
through the regular terminal channels, so this adds no reach.
* fix(agent-dashboard): sync locale catalogs for new mode/close keys
verify:localization-catalog (part of lint CI) fails when en.json keys are
missing from the other locale catalogs; run sync:localization-catalog so
the six new agent-dashboard keys exist everywhere (English fallback text;
translated copy remains the documented follow-up).
* feat(agent-dashboard): present in-window mode as a companion board sheet
The in-window dashboard now uses the same non-modal left sheet as the
workspace kanban board — anchored to the sidebar edge, chrome/status-bar
bounds, sidebar stays interactive — instead of a near-fullscreen modal
dialog. Both companion boards are mutually exclusive; the sidebar entry
toggles the drawer. Removes the modal focus-restore timing coupling on
reveal.
* fix(agent-dashboard): ignore Radix dismiss requests like the workspace board
Non-modal Radix layers also request dismissal for interactions the drawer's
outside guards cannot classify — focus moving outside carries no pointer
coordinates, and clicks in the status bar / top chrome fall outside the
right-side dismiss band. Forward only open requests from the Sheet, matching
WorkspaceKanbanDrawer, so only the drawer's own escape/outside/close paths
close it.
* fix(agent-dashboard): guard reveal relay and refresh stale mode copy
CodeRabbit review: revealAgent lacked the ?. HMR-skew guard its sibling
ackAgent has (both channels shipped together, so a stale dev preload
lacks both). The es/ja/ko/zh catalogs also still described the dashboard
as pop-out-only in stale English, contradicting the new in-window
default; refreshed to the current English source.
* feat(agent-dashboard): add board settings menu to the in-window header
Mirrors the workspace board's settings gear: an Open as segmented control
in the board header so the mode is changeable without opening Settings.
Switching to pop-out hands the surface over (closes the drawer, opens the
window) instead of leaving a board the setting says should be a window.
In-window only via an optional headerActions slot - the pop-out renderer
has no store to drive it.
* fix(agent-dashboard): reset the settings-menu flag when the drawer closes
The pop-out hand-off closes the sheet while the menu is still open, so the
menu unmounts without Radix reporting onOpenChange(false). The stale
menuOpen=true then blocked outside-dismiss permanently on the next
in-window open. Mirror closeWorkspaceBoard by resetting the flag in close.
* fix(agent-dashboard): reset the menu flag on store-driven drawer closes
Cmd+B sidebar collapse and the workspace-board exclusivity effect close
the drawer via setAgentDashboardDrawerOpen directly, bypassing close();
a settings menu open at that moment unmounted without Radix reporting
onOpenChange(false), leaving menuOpen stuck true and outside-dismiss
disabled on the next open. Sync the flag to the open state so every
close path resets it.
Stacked layout (head / → base) lets long branch names fit narrow sidebars
without truncating either line. Also show head-only identity when compare base
isn't configured, and improve upstream vs compare target distinction in stats.
Move link-tooltip chrome into terminal.css so offsets cannot drift via
inline styles, and square the bottom-left corner so the hover preview
sits flush against the pane edge (Ghostty-style).
An image send whose text+Enter RPC ended 'unknown' (ack loss / path
cutover) collapsed to accepted=true, so the terminal was never marked
stale. When the Enter truly never landed, the already-pasted image path
sat on the input line and glued onto the next plain-text message.
Propagate the send outcome through handleNativeChatSendWithOutcome and
mark the terminal input stale on any non-accepted outcome; the next send
heals with Ctrl+U (a no-op when the message did land). Chips still clear
on 'unknown' to avoid a double-send on retry.
* fix(terminal): reveal Markdown links at target lines
* fix(terminal): scope line reveals to opened tabs
---------
Co-authored-by: kaynan <kaynan.camargo@terceiro-sky.com.br>
* Normalize branch prefixes and flag invalid ones in settings
A custom branch prefix ending in a slash (e.g. "team/") produced a
double-slashed branch name like "team//feature" that git rejects, and
the raw check-ref-format error gave no hint that the prefix caused it.
- Normalize the configured prefix (trim whitespace, strip leading/
trailing and duplicate slashes) in the shared branch-name builder so
the common trailing-slash case just works, for local and SSH worktrees.
- Validate the prefix on the worktree-create path (computeValidatedBranchName)
so a genuinely invalid prefix fails fast with a clear
"update it in Settings -> Git" message instead of an opaque git error.
- Add a live BranchPrefixFeedback under the Branch Prefix setting: previews
the resulting branch name, warns on invalid characters, and notes when a
prefix collapses to none.
- Keep the background first-work rename on the non-throwing builder since the
prefix is already validated at create time.
* Keep caret in place when editing the branch prefix
The custom branch prefix input was directly controlled by settings, but
updateSettings persists through an async IPC round-trip, so the value
updated a tick late and React re-assigned it, snapping the caret to the
end on mid-string edits. Drive the input from a local draft and only
adopt genuine external settings changes so the caret stays put (and fast
typing survives slow SSH round-trips).
Co-authored-by: Cursor <cursoragent@cursor.com>
* Return ReactNode from BranchPrefixFeedback
JSX.Element needlessly excludes null/string/number returns; ReactNode
keeps the component's return type from over-constraining future changes.
Co-authored-by: Cursor <cursoragent@cursor.com>
---------
Co-authored-by: Cursor <cursoragent@cursor.com>
* test(session): guard against workspace-reopen tab fork-bomb (STA-1111)
Relates to STA-1111 (already fixed on main by #6945).
The runaway-tab-on-reopen bug was fixed by #6945 (activeOrQueuedResumeClaimsProviderSession dedup guard) hours before the ticket was filed. Verified: disabling that guard makes tab count climb 1->2->3->4 across reopens; restoring it holds at 1. This adds a revert-sensitive regression test covering the created-agent and sleeping-resume reopen paths so it cannot regress.
Test plan: 2 new tests pass; fail if the #6945 guard is removed.
* test(session): isolate automatic resume replay coverage
* test(session): assert resumed tab identity stays stable
* fix(window): reflow on macOS occlusion-reveal so the bottom bar is not clipped
Fixes STA-2383.
On macOS the window is background-throttled while hidden; on occlusion-uncover only 'focus' fires and its handler runs webContents.invalidate() (the setSize jiggle is skipped to avoid SIGWINCH-ing terminals). invalidate() repaints but does not reflow, so the app-shell h-dvh root keeps a stale dynamic-viewport height and the StatusBar is clipped below the viewport ('no bottom bar'); a manual resize restores it.
Fix: renderer relays a genuine hidden->visible reveal (visibilitychange) to main via new ui.notifyWindowRevealed IPC; main runs the same proven forceRepaint. Occlusion-gated (no per-Cmd+Tab SIGWINCH regression), darwin-scoped, sender-guarded, cleaned up on close.
Test plan: vitest createMainWindow + web-preload-api green. Recommend on-device macOS occlusion QA before merge.
* fix(window): preserve user resizes during reveal repaint
createPtySubprocess asserted assertSafeAgentStartupCwd on the raw
opts.cwd before applying opts.cwd || getDefaultCwd(). An omitted cwd
was treated as root-like and threw even though the post-fallback home
default is safe — while LocalPtyProvider already gates the effective
cwd after fallback.
Compute requestedCwd first, then gate. Update the daemon unit test to
assert omitted-cwd agent launch uses the safe default (#9578).
The New Workspace flow built a bare launch command client-side and sent it as
`startupCommand`, so the host ran it verbatim and never applied the default
launch args. The first Claude session therefore started in manual mode, while
opening another Claude via the "+" tab (which sends the agent id and lets the
host resolve args) started with `--dangerously-skip-permissions`.
Send `startupAgent` from every client-built create path (blank, reuse-branch,
and new-branch) so the host resolves the launch command, args, env, and
host-shell quoting through the same path the "+" new-tab and CLI use. The
work-item path already delegated via `startupDraft`. Custom `agentDefaultArgs`
are now honored on all paths.
Adds a shared `agentLaunchCreateFields` helper and removes the now-unused
client-side command map, which had also drifted from the canonical launch
commands for continue, hermes, command-code, kiro, and mistral-vibe.
Claude-Session: https://claude.ai/code/session_014iufZnQwPD2obYuvdahjaE
Co-authored-by: kaynan <kaynan.camargo@terceiro-sky.com.br>
The claude/codex/openCode usage store slices read `scanState.enabled`
directly off `window.api.<provider>Usage.getScanState()`. In the web
client that usage IPC is not bridged, so the preload fallback proxy
resolves those calls to `undefined`, and enabling usage tracking from
Settings -> Stats & Usage throws
`TypeError: Cannot read properties of undefined (reading 'enabled')`
(reproduced live against `orca serve` v1.4.150; present on main too).
Guard the getScanState()/setEnabled() seams in all three slices so an
absent scan state degrades to a graceful no-op instead of crashing.
Desktop behavior is unchanged (a real ScanState is always truthy).
Adds a regression test that stubs the web-client fallback (every call
-> undefined) and asserts fetch*/enable* no-op without throwing for all
three providers.
Co-authored-by: ECO2G Migration <cmeia.ai02@cmeia.co.kr>
* fix(win): taskkill agent PTY descendant trees on stop
Windows agent teardown previously degraded to shell-only kill because
descendant snapshot is POSIX-only. Orphaned claude/codex/MCP children
kept worktree cwd handles open and blocked worktree remove.
Use taskkill /T /F on the PTY root in killWithDescendantSweep (daemon +
local agent shutdown) so the ConPTY tree is cleared before teardown.
Fixes#10004
* fix(win): bound taskkill with timeout and windowsHide
Prevent a wedged taskkill from delaying killRoot, and hide the console
flash during agent PTY tree teardown (CodeRabbit on #10100).
During scrollToIndex/reveal, rangeStartIndex can advance before TanStack
mounts the candidate Project row. Returning that unmounted index made a
Project sticky paint over the Host card in multi-host views. Prefer a
previous mounted sticky (or none for the group tier) until geometry exists.
Closes#10088
The theme picker count row concatenates "Showing {count}" directly with
the " of {{value0}}" fragment. The ko/ja/zh translations dropped the
fragment's leading separator, so the shown and total counts fused
(e.g. Korean rendered "표시 중 3030 중" instead of "표시 중 30/30").
Restore a slash separator for the total-count fragment and the leading
space for the search-match fragment in ko/ja/zh, in both the runtime
catalogs and the key-override sources so catalog regeneration keeps the
repaired values. Add a regression test covering both fragments.
🤖 Generated with Claude Code
`git sparse-checkout disable` restores the full working tree and sets
core.sparseCheckout=false, but deliberately leaves <gitdir>/info/sparse-checkout
in place so the checkout can be re-enabled with the same patterns.
detectSparseCheckout treated the mere presence of that pattern file as "sparse",
so a fully-populated worktree kept showing the sparse badge and the misleading
"Partial checkout. Files outside these paths are not on disk." tooltip.
Gate the fast-path fs.stat behind a config read that confirms core.sparseCheckout
is actually enabled (shared repo config or per-worktree config.worktree, honoring
git's precedence). The config read runs only when a non-empty pattern file
exists, so it does not reintroduce the per-poll subprocess fan-out PR #1290
removed, and it reads git's config files directly (no subprocess).
Adds a real-git regression test (enable -> disable leaves file -> not sparse) and
unit tests for the git-config boolean parser.
* fix(terminal): make Ctrl+V paste work in the HTTP web client
navigator.clipboard only exists in secure contexts, so the web client
served over plain HTTP (e.g. a LAN address) read an empty clipboard and
Ctrl+V silently did nothing: the keydown handler preventDefault-ed the
chord and suppressed the native paste event whose clipboardData is the
only clipboard access available there.
When the async clipboard reader is unavailable in the web client, let
the chord's native paste event fire and feed its clipboardData text
through the existing paste pipeline (size limits, bracketed-paste
handling and error toasts included).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01RTjQWLg4wy1CgmcZKmGAuN
* fix: address review finding (bug-bash takeover)
Co-authored-by: Orca <help@stably.ai>
---------
Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: Neil <4138956+nwparker@users.noreply.github.com>
Co-authored-by: Orca <help@stably.ai>
NeilandOrca
OrcaWin
github-actions[bot]
Brennan Benson
Jinjing
moseohandBrennan Benson
kazu-42andBrennan Benson
kin001andBrennan Benson
Jinwoo HongandOrcaWin
Mark Xian
Kaynan Sampaio de Camargoandkaynan
YoelandCursor
Wooseong Kim
scokeepaandECO2G Migration
JeongUk Park
Taras Kornichuk
Tingde Liu