Commit Graph
20254 Commits
Author SHA1 Message Date
OrcaWinandOrcaWin b67159291b test: remove 14 low-value renderer tests (#39871)
Co-authored-by: OrcaWin <293788423+OrcaWin@users.noreply.github.com>
2026-11-08 06:26:47 -08:00
Neilandm4air dddc1260e5 chore(lint): oxlint 1.91, cheap perf/correctness rules, return-await fixes (#39730)
* chore(lint): bump oxlint to 1.91, add perf lint rules and perf audits

* fix(lint): await promises returned inside try/catch/finally for return-await

Prepare typescript/return-await with error-handling-correctness-only: await
returns whose catch or finally was meant to cover them, and suppress the eight
sites where the catch must not re-handle the returned promise.

* chore(lint): enable return-await (error-handling-correctness-only); exempt secondary-scan directives in the changed gate

* refactor(mobile): move file preview discard modal into its own module

Keeps MobileFilePreviewScreen under the 400-line limit after merging main.

---------

Co-authored-by: m4air <m4air@m4airs-MacBook-Air.local>
2026-11-08 06:21:54 -08:00
ae6675d67c fix(window): cancelling a quit prompt clears the quit latch so later closes aren't treated as quits (#39597)
* fix(window): cancelling a quit prompt now cancels the quit

Renderer-owned quit prompts (running-work "Close Window?", Save dialog,
close-guard and shutdown-checkpoint vetoes) never reach
will-prevent-unload, so main kept isQuitting latched after Cancel. A later
plain close was treated as a quit and renderer crashes were not recovered.

Add a window:close-cancelled IPC (preload ui.cancelWindowClose; no-op in the
web client) that the renderer sends on each cancel path. Main accepts it only
from its own renderer, only while quitting, and only once that renderer has
acknowledged the newest close request, then runs the same abort path as a
beforeunload veto.

Fixes #39596

Co-authored-by: Neil <neil@stably.ai>

* test(window): expose webContents.send on the renderer recovery harness type

---------

Co-authored-by: tomarai85 <tomarai85@users.noreply.github.com>
Co-authored-by: Neil <neil@stably.ai>
2026-11-08 06:17:53 -08:00
d464685250 fix(codex): do not count a dangling link raced into the target as bridged (#39862)
When a link attempt finds the target taken, check what is there. A
dangling symlink another writer created after the directory listing is
not readable history, so the thread stays out of the bridged set, as
on main.

Co-authored-by: m4air <m4air@m4airs-MacBook-Air.local>
Co-authored-by: AmethystLiang <6427696+AmethystLiang@users.noreply.github.com>
2026-11-08 06:14:48 -08:00
OrcaWinandOrcaWin 60aee85d91 test: remove 27 low-value hook and terminal tests (#39866)
Co-authored-by: OrcaWin <293788423+OrcaWin@users.noreply.github.com>
2026-11-08 06:05:26 -08:00
OrcaWinandOrcaWin 7d195c7e50 test: remove 95 low-value renderer tests (#39859)
* test: remove 95 low-value renderer tests

* test: remove duplicate environment directives

---------

Co-authored-by: OrcaWin <293788423+OrcaWin@users.noreply.github.com>
2026-11-08 05:53:18 -08:00
OrcaWinandOrcaWin 8c1e6dfba7 test: remove four redundant terminal tests (#39858)
Co-authored-by: OrcaWin <293788423+OrcaWin@users.noreply.github.com>
2026-11-08 05:43:52 -08:00
Neil 10c01509de fix(ime): don't swallow the first Enter after toggling Japanese IME off with Hankaku/Zenkaku (#39834)
On Windows, Hankaku/Zenkaku commits the Japanese composition as a
Process/229 keydown that may never send a keyup. The composer treated any
Process/229 keydown as an IME-confirm Enter and armed the carry, so the
user's next real Enter was swallowed. Only arm the carry when the
processed key's physical code is Enter (or unknown), sharing the code
check with the terminal IME path.

Fixes #39289
2026-11-08 05:32:19 -08:00
Neil 5fb71170ff fix(hermes): tolerate a deleted cwd in the status hook so pre_tool_call isn't blocked (#39833)
Hermes Kanban cleanup can remove a worker's cwd; os.getcwd() then raised
FileNotFoundError from the orca-status plugin and Hermes failed pre_tool_call
closed, blocking the tool. The payload now omits cwd when it can't be read.

Fixes #39736.
2026-11-08 05:32:09 -08:00
OrcaWinandOrcaWin 96f5f5a9eb test: remove 128 low-value tests (#39853)
Co-authored-by: OrcaWin <293788423+OrcaWin@users.noreply.github.com>
2026-11-08 05:31:22 -08:00
Jinjing 63637e82e8 perf(codex): bridge account history with async directory diffs (#39805)
Account selection and every Codex launch walked all rollouts in every
source home and ran a synchronous existsSync per rollout on the main
process, then linked missing ones with sync mkdir/link/symlink. A CPU
profile of an installed build attributed ~6.4s of sampled existsSync time
to this bridge during typing lag.

The bridge now walks all source homes together and diffs directory
listings asynchronously, so a target directory is read once per pass and
history the target already holds costs no per-rollout synchronous call.
Small directories use per-path checks instead of listing a wide target.
Name matching yields every 256 entries.

Behavior matches the previous bridge:
- An existing target name is never overwritten.
- Earlier sources are tried first, and a later source still gets an
  attempt when an earlier one cannot be linked.
- Hardlink-only targets skip and leave no folders behind.
- Readiness, index healing and quit handling are unchanged.
2026-11-08 13:14:22 +00:00
OrcaWinandOrcaWin 22f387b6b3 test: remove 231 low-value tests (#39850)
Co-authored-by: OrcaWin <293788423+OrcaWin@users.noreply.github.com>
2026-11-08 05:06:36 -08:00
Neil 656a05de0d perf(main): overlap independent fs probes; skip per-entry copies (#39729)
* perf(main): overlap independent fs probes on git status, agent detection and orcad startup

* perf: skip per-entry copies in git change list and tab hydration; drop unused activity barrel
2026-11-08 05:05:19 -08:00
537d764f58 perf(mobile): warm WebKit's processes on iOS before the first page opens (#39851)
* perf(mobile): warm WebKit's processes on iOS before the first page opens

Builds a blank, non-persistent WKWebView a second after Home has a paired
host, so the first page open starts against WebContent, GPU and Network
processes already running. A release before the delayed start now makes
the start a no-op.

Based on #32163 by Jinwoo-H.

Co-authored-by: Jinwoo-H <73622457+Jinwoo-H@users.noreply.github.com>

* fix(mobile): guard warmUp on older binaries and release the warm view on page failure

Co-authored-by: Jinwoo-H <73622457+Jinwoo-H@users.noreply.github.com>

---------

Co-authored-by: m4air <m4air@m4airs-MacBook-Air.local>
Co-authored-by: Jinwoo-H <73622457+Jinwoo-H@users.noreply.github.com>
2026-11-08 05:01:36 -08:00
346dcae485 perf(test): default check:changed unit tests to two Vitest workers (#39849)
Plain local pnpm test keeps automatic sizing; explicit --maxWorkers still wins.

Co-authored-by: m4air <m4air@m4airs-MacBook-Air.local>
Co-authored-by: nwparker <4138956+nwparker@users.noreply.github.com>
2026-11-08 04:53:17 -08:00
8995b03646 perf(skills): read skill folders ahead of the walk, three reads at a time (#39836)
* perf(skills): read skill folders ahead of the walk, three reads at a time

Port of #35220. findSkillFiles awaited realpath then readdir for every
folder, so a busy main process paid about two event-loop turns per folder.
A read-ahead now lists folders in walk order while the walk claims them
in its own depth-first order, so results, order, alias winners and the
cycle guard are unchanged.

Changes from #35220: up to 3 read-ahead reads (was 64) plus at most one
demand read from the serial walk, so at most 4 directory reads in flight
on libuv's 4-thread pool; a sorted array replaces the min-heap; tests
assert walk results only.

Co-authored-by: thefrikidude <100475418+thefrikidude@users.noreply.github.com>

* test(skills): drop walk tests that cannot detect the read-ahead

Both pass on the old walk, so they do not guard this perf-only change.
Note the in-flight bound applies per walk.

Co-authored-by: thefrikidude <100475418+thefrikidude@users.noreply.github.com>

---------

Co-authored-by: m4air <m4air@m4airs-MacBook-Air.local>
Co-authored-by: thefrikidude <100475418+thefrikidude@users.noreply.github.com>
2026-11-08 04:41:21 -08:00
0f6432319d perf(orchestration): back off failing federation relays instead of polling every second (#39734)
Each federated dispatch relay now reschedules itself and doubles its wait
after consecutive failures (capped at 30 s) instead of syncing every second
forever. A server turning verified, or a sync completed by another caller,
returns the relay to the base cadence. A coordinator blocked in
`check --wait` re-arms its run's relays every 15 s, so a recovered worker is
seen within that window. Relays for a server removed from this host stop
instead of retrying; the next relay arm brings them back if it is re-added.
Stale in-flight syncs no longer record acknowledgments, epochs or home
auto-release after the relay stopped.

Replaces #39283.

Co-authored-by: m4air <m4air@m4airs-MacBook-Air.local>
Co-authored-by: brennanb2025 <79079362+brennanb2025@users.noreply.github.com>
2026-11-08 04:38:49 -08:00
Neilandm4air 2a5a939a0c test: split GitHub and GitLab mocks out of runtime setup.spec to fix main lint (#39837)
* test: split GitHub and GitLab mocks out of runtime setup.spec

setup.spec.ts reached 801 counted lines after #39575, failing the root
oxlint max-lines rule on main. Move the GitHub/hosted-review and GitLab
mock functions and their vi.mock registrations into sibling modules that
setup.spec re-exports, so every consumer keeps the same imports and mocks.

* test: name split runtime mock modules as test harnesses

The module-mocking lint exempts *-test-harness.ts support modules shared by
split suites; the split-out GitHub and GitLab mock modules are exactly that.

---------

Co-authored-by: m4air <m4air@m4airs-MacBook-Air.local>
2026-11-08 04:22:50 -08:00
RoastFriedandNeil ba04481e42 fix(browser): keep native popup minimum content size and first Windows fixture error (#39319)
Native popups now hide the app menu bar like origin-bar popups, so the
Windows menu no longer eats into the 200px minimum content height.

The popup compatibility fixture keeps the first reported error across
later status writes, stops serializing the destroyed Event as an error,
and replaces its status file with the Windows retry helper. The native
resize check now compares against the native outer window size.

Fixes #39299

Co-authored-by: Neil <neil@stably.ai>
2026-11-08 04:01:25 -08:00
OrcaWinandOrcaWin 5fc3b26766 test: remove 712 low-value tests (#39829)
Co-authored-by: OrcaWin <293788423+OrcaWin@users.noreply.github.com>
2026-11-08 03:51:04 -08:00
WenhaoandNeil 448aaadde6 fix(antigravity): detect Orca Chat sign-in via antigravity-acp/acp_token.json (#39398)
agy-acp-server 1.4.1 keeps the personal sign-in at
<GEMINI_HOME>/antigravity-acp/acp_token.json; nothing it writes ends in
oauth-token, so a signed-in Chat read as signed out and new chats could
move to the host's ~/.gemini account.

Fixes #39395

Co-authored-by: Neil <neil@stably.ai>
2026-11-08 03:45:48 -08:00
Brennan Benson d857e09615 Git Graph: commit actions (checkout, cherry-pick, revert) (#39069)
Adds commit actions to the Git Graph: check out a commit or branch, cherry-pick, revert, and create a branch from a commit. Conflicts hand off to Source Control with continue and abort. Works on local, WSL and remote hosts.
2026-11-08 03:43:30 -08:00
Vitalii KulishandNeil 4824562f3e fix(packaging): run the terminal host PTY probe in /bin/sh (#39602)
The packaged daemon PTY check types a POSIX arithmetic command and waits
for "OK 42". The daemon picked the PTY shell from the builder's $SHELL, so
a fish builder never printed it and Linux afterPack timed out. Pin the
probe daemon's SHELL to /bin/sh so the check exercises the packaged PTY,
not the builder's login shell.

Fixes #39591

Co-authored-by: Neil <neil@stably.ai>
2026-11-08 03:35:16 -08:00
Brennan Benson 3fe350eaed Create a worktree per repository from an empty folder workspace (#39575)
A folder workspace with no worktrees now offers Create worktree for each repository, named after the folder workspace. If that branch name is already taken, the host refuses instead of picking another name, and the composer opens so you can choose.
2026-11-08 03:30:10 -08:00
YohanandNeil 56d9db5042 fix(sidebar): don't wake a sibling Hide sleeping keeps hidden when the active workspace is deleted (#39230)
With Hide sleeping on, deleting the active workspace activated the most
recently visited sibling (or main) and seeded it a terminal, so a hidden
sleeping row popped back into the sidebar. The removal successor now
skips siblings the sleeping sweep hides, using the same check the
sidebar's reveal path uses and honoring rows held through the PTY-rebind
grace window. With no eligible sibling the user lands on the start screen.

Fixes #39231

Co-authored-by: Neil <neil@stably.ai>
2026-11-08 03:29:45 -08:00
luisgonzaleznfandNeil a7d94eb86f fix(settings): line up an agent's launch fields on one type scale (#39199)
Command, Environment and Codex's session source home now use Input's
code variant, so they render at text-xs instead of md:text-sm above
768px. The Arguments label matches its siblings' text-xs style, and the
Arguments wrapper uses py-0.75 so its text is vertically centred.

Co-authored-by: Neil <neil@stably.ai>
2026-11-08 03:29:34 -08:00
Naveen ChatlapalliandNeil 58f17b3af2 fix(bitbucket): document API token scopes in connection setup (#39118)
List the Bitbucket API token scopes Orca needs (read:user, read:repository,
read:pullrequest, write:pullrequest) in the Email & API token form so users
configure a working token up front instead of hitting missing-permission
errors when creating a pull request. Strings are translated in every locale.

Fixes #38652

Co-authored-by: Neil <neil@stably.ai>
2026-11-08 03:24:14 -08:00
Neil b746eef069 fix(shell): keep a user's claude/codex function when the cwd has a matching entry (#39820)
command -v prints a shell function's bare name, so a claude/codex directory
or executable in the cwd passed the -x check and Orca replaced the user's
function. Skip the wrapper when a function with that name is already defined.

STA-13246
2026-11-08 03:23:55 -08:00
NeilandCopilot 40236ab701 Use searchable pickers for long Settings lists (#39815)
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
2026-11-08 03:20:47 -08:00
Neil 00d3e44ba5 fix(agent-hooks): let the last-status throttle apply to routine hook events (#39826)
* fix(agent-hooks): let the last-status throttle apply to routine hook events

Each hook event re-recorded the pane's authority evidence, which bumped the
revision counters behind the "urgent persist" fingerprint even when only
observedAt moved. Every event then skipped the 5s throttle and rewrote
last-status.json. Update the stored evidence without bumping the revision when
the authority itself (launch token, connection, tab, worktree) is unchanged.

* test(agent-hooks): remove timer-racy persistence count assertions
2026-11-08 03:18:10 -08:00
Brennan Benson f399a071bb Open a landing instead of a terminal when splitting a Chat UI pane (#39234)
* Open the new-tab landing when splitting a Chat UI pane

Splitting a chat with Cmd+D opened a terminal nobody asked for. The split now
opens an empty pane that shows the home-screen landing (New agent, New
Terminal, New Browser Tab, plus shortcuts); Cmd+W closes it. Chats in the
floating panel and inside terminal tabs keep their terminal split.

* Split landing review fixes: agent launch cancellation, browser policy, labels

- New Agent on the landing cancels like the keyboard shortcut when the user leaves the workspace or host
- Hide New Browser Tab where the creation policy hides it (paired web without streaming)
- Chat split actions read Split Right / Split Down; New Agent gets its own key
- Tour visibility reads the projection's existing group subscriptions

* Show the split landing only in panes a chat split opened

Other flows reserve an empty split group for content they already chose
(remote terminal splits, AI Vault resume, browser and source-control
opens). The landing, its Close collapse and the tour now apply only to
groups the chat split marked, kept in window-only state and dropped when
the group goes away.

* Move split landing marks with their groups on worktree rename

* Keep the rename test's chooser mark out of the cast fixture

* Read the landing's browser policy by workspace; keep Agent in the Japanese label

* Keep the passed split lesson in the agent sessions tour count

* Keep the workspace open on a waiting split landing when its last tab closes

* Keep a waiting split landing when its workspace is reopened; drop its close row when it is the only pane

* Let a waiting split landing satisfy automatic default seeding, not recovery

Restores main's agent-activation inventory gate (a chooser never bypasses
ownership checks). Each automatic default-surface decision (shared seeder,
deferred reseed, startup watcher, paired wake, stream bootstrap) now leaves
a waiting chooser in place, and the paired mirror keeps a marked chooser when
the host reports no tabs. Explicit launches and requested default tabs are
unchanged.

* Fill an emptied paired workspace with its waiting landings, layout and focus together

* Read landing labels through useTranslate for main's render-path ratchet

* Drop the split landing's subtitle
2026-11-08 03:08:20 -08:00
OrcaWinandOrcaWin 88aa9fa82a test: remove 154 low-value tests and redundant relay checkpoints (#39814)
* test: remove 155 low-value tests and two relay checkpoints

* test: retain schema drift protection and remove dead imports

---------

Co-authored-by: OrcaWin <293788423+OrcaWin@users.noreply.github.com>
2026-11-08 03:06:41 -08:00
OrcaWinandOrcaWin de3da487fa test: stop worker-teardown console race from failing unit shards (#39813)
Vitest rejects a console-log RPC still in flight when a worker tears
down, failing a shard whose tests all passed. Ignore only that rejection
(the lost line belongs to a finished file and passed-only shards drop
it anyway), and stop the two specs seen in CI from spawning real git and
wsl.exe in fire-and-forget background work.

Co-authored-by: OrcaWin <293788423+OrcaWin@users.noreply.github.com>
2026-11-08 02:55:36 -08:00
NeilandCopilot b4a522deeb Use Appearance's searchable picker for terminal themes (#39800)
* fix(settings): unify terminal theme pickers

Reuse the Appearance searchable selector and palette swatches while preserving imported-theme actions and independent terminal theme slots.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* fix(settings): don't replay a handled theme import after remount

Settings search unmounts the theme row, so the picker remounted with
handledImportSignal=0 and reopened at Imported without a new import.

---------

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
2026-11-08 02:50:22 -08:00
NeilandCopilot fc8dcd65eb Simplify delete workspace dialog (#39801)
* feat(sidebar): simplify delete workspace dialog

Replace the "Don't ask again" checkbox with an "Always delete" option in a
split-button dropdown on the Delete button, matching the force-delete toasts.
Shorten the description and change-loss copy so the dialog reads lighter.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* fix(sidebar): keep delete change-loss warning protected and shown when status is unknown

Default Delete forces removal, so restore the permanent-loss notice while a
target's Git status is still checking or unavailable. Mark the dirty-change
hint as protected copy so language packs cannot replace the loss warning.

---------

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
2026-11-08 02:29:48 -08:00
Brennan Benson 7a8c16caa5 Preserve the chat's history when forking (#39168)
* Preserve the chat's history when forking

* Preserve fork receipts and finished conversation history

Save native-copy receipts on the real create path, retain standalone Codex copies and Claude compaction bridges, and omit unfinished helper output. Restore rewind read-error refusals and update the affected provider fixtures.

* Keep fork coverage tied to OpenCode's copied messages

* Preserve the actionable refusal when chat creation stops

* Keep completed helper answers when copying chat history

Follow spawning tools and active helpers to their descendants without making an omitted goal row invalidate a settled helper's independent output. Reuse the public copy-error distinction so private transcript exceptions stay behind the generic create refusal while authored fork errors remain actionable.

* Keep fork history stable with OpenCode 1.x

* Exclude later continuation notices from fork snapshots

* Keep fork continuation notices tied to retained work

* Keep fork catch-up recoverable and preserve copy cancellation

* Keep OpenCode coverage timeouts on the read refusal path
2026-11-08 01:48:11 -08:00
Neil fca1d7b134 fix(mobile): prevent duplicate queued sends after chat remount (#39809) 2026-11-08 01:16:15 -08:00
Jinwoo Hong 4fabde28b1 feat(cli): one enclosing-workspace rule, --environment local, copyable addresses, and --parent-environment (#39733)
* feat(cli): one enclosing-workspace rule, --environment local, and copyable workspace addresses

current/active and cwd inference now pick the innermost workspace of either kind (worktree or
folder workspace); at equal depth the caller terminal's own workspace wins, otherwise the choice
is refused as ambiguous. --environment local names this machine's own Orca over an ambient
ORCA_ENVIRONMENT or pairing code. worktree ps/show/current print an environment: line and an
address: line (JSON environment, address, addressUnavailable, parent) that always carries its
--environment, and name a parent's environment only when it is verified.

* test: answer the folder catalog read in agent launch cwd tests

* feat(cli): worktree rm refuses current/active and names the selector to pass

Removal needs an explicit workspace, as git worktree remove needs a path. The refusal runs before
rm decides between a worktree and a folder workspace, and reports what current resolves to now
(one selector, the ambiguous choices, or none) in text and as invalid_argument data.

* fix(cli): keep the remote-environment guidance when rm refuses current

* fix(cli): surface folder-catalog failures for omitted targets; no address for a duplicated folder parent

* fix(cli): surface a worktree-list failure for omitted browser and terminal targets too

* test(cli): folder rm refuses current/active before the folder removal path

* fix(cli): refuse a cwd owner whose folder host can't be derived instead of skipping it

* fix(cli): compare cwd owners by filesystem identity, refuse every equal-path tie, surface worktree-list failures

Paths on the caller's own filesystem compare by realpath (reusing resolveRealPathSync), so /tmp
and /private/tmp aliases find the deeper owner in both the folder prefilter and ranking; other
hosts' paths stay lexical. Equal-depth owners with different repo registrations now refuse without
a caller instead of picking by catalog order. An omitted target no longer swallows worktree.list
method_not_found; only the optional folder reads fall back.

* fix(cli): read the realpath helper from shared so the packaged CLI ships it

* refactor(cli): move the catalog row type and unresolved-owner refusal out of the selection/catalog cycle

* fix(cli): resolve every caller-filesystem workspace path, not only basename matches

A workspace registered through a differently named symlink was skipped by the basename gate.
Containment now realpaths cwd and each row on the caller's own filesystem once per command,
shared by the folder prefilter and ranking; folder hosts are derived first so another host's
paths are never resolved locally.

* test: answer the folder catalog read in the ambiguous orchestration-check caller test

* feat(cli): --parent-environment names where a worktree's parent lives

worktree create and worktree set take --parent-environment <local|saved id|name>
next to --parent-worktree. The parent is resolved in that environment only, and
sent to the child as a verified cross-store ref after the child's cross-machine
capability check. Without the flag, the parent stays in the child's environment;
active/current always mean this machine's own Orca.

Preflight failures happen before any change: environment_not_paired,
environment_unavailable, LINEAGE_PARENT_NOT_FOUND / LINEAGE_PARENT_AMBIGUOUS
naming the searched environment, and incompatible_runtime with data.environment
and data.requires.

* fix(cli): parent lookup errors name and list the environment searched

LINEAGE_PARENT_AMBIGUOUS and LINEAGE_PARENT_NOT_FOUND now always carry
data.environment and a listing command for the environment the parent was
looked up in, for parents in another environment and in the child's own. Steps
inherited from the runtime listed the caller's default route instead.

* docs(cli): a local address names the Orca that printed it
2026-11-08 04:08:07 -05:00
OrcaWinandOrcaWin 0da741a22f test: remove 988 low-value tests (#39797)
Co-authored-by: OrcaWin <293788423+OrcaWin@users.noreply.github.com>
2026-11-08 00:42:37 -08:00
Jinjing 9a1c3a362d more accurate chat ui invocation count (#39784)
* rm doc

* wip

* wip

* wip

* wip

* wip

* Handle opening message receipt collisions
2026-11-08 00:40:56 -08:00
Jinjing c60bd0f2a3 fix: align sidebar activity badge with visible agents (#39783)
* fix: align sidebar activity badge with visible agents

* fix: retain activity focus read ratchet entries
2026-11-08 00:22:21 -08:00
Jinjing 5e496520c5 Allow interleaving projects and groups in the sidebar (#39768)
* Allow interleaving projects and groups in sidebar

* Fix drag refs during render
2026-11-08 00:16:23 -08:00
NeilandCopilot 178a9f27bd fix(new-workspace): keep project list closed when Add project dialog lazy-loads (#39764)
The Add project dialog focused Browse in a mount effect before its focus
trap registered, so the composer trap bounced focus back to the Project
field, which reopened the project list on top of the new dialog.

Focus Browse from the dialog's onOpenAutoFocus instead, and close the
project list when focus leaves the field and its popup.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
2026-11-08 00:09:44 -08:00
OrcaWinandOrcaWin ee6506714e test: remove 244 low-value tests (#39781)
Co-authored-by: OrcaWin <293788423+OrcaWin@users.noreply.github.com>
2026-11-08 00:06:49 -08:00
Jinwoo Hong 14492c79cd fix(runtime): keep the output of a worker whose terminal dies during startup (#38475)
* fix(runtime): keep the output of a worker whose PTY dies during startup

A PTY exit frees its recent-output buffer before the readiness failure is formatted, so a launch that killed its terminal reported only terminal_exited. The exit now keeps what the last command printed on the PTY record, and the readiness failure for that exact incarnation reports it like a launch whose shell survived. The start receipt redacts dispatch capabilities from the reason.

Builds on #37866 by @gsxdsm.

* fix(runtime): bind saved exit output to its incarnation and survive a repeated exit

* fix(runtime): redact secrets from a failed worker start's reason
2026-11-08 03:03:33 -05:00
OrcaWinandOrcaWin d058920e15 test: remove 357 low-value tests (#39743)
* test: remove 362 low-value tests

* test: retain five boundary regression cases

---------

Co-authored-by: OrcaWin <293788423+OrcaWin@users.noreply.github.com>
2026-11-07 23:40:13 -08:00
Neil 5df115db15 fix(relay): runtime edge control on the director, safer skipped exports, faster evacuation (#39773)
- Directors with an edge read director/edge.json from the relay-control bucket through the shared
  flag channel: mode on/off/down and rolloutPermille (0..1000). Boot defaults are mode on and the
  env percent; every cohort decision reads the applied flags. Off empties the cohort and hosts
  leave normally; down also skips the export at once (reason mode_down). The permille bucket keeps
  every host of P percent in at P x 10 permille. A control change re-arms the evacuation sweep
  within 5 s.
- A skipped export (120 s outage clock or mode down) removes the host's relay_devices rows before
  placing it, so revokes made on the edge are never undone; logs a count only.
- Evacuation: 100 drains per tick with 10 in flight, each pass starting at a random point of the
  keys and wrapping around; pages stay 500 rows and an idle pass reads each edge row once.
- Fence retries: 56 attempts (last about 25.5 h after the failure), 50 per sweep with 10 in
  flight; fence_abandoned now carries outcome, cell and epoch like fence_failed.
2026-11-07 23:38:33 -08:00
Neil c362cb183e relay-edge: keep valid desktops through JWKS outages, deep object probe, cohort-scaled alerts (#39772)
* relay-edge: keep valid desktops on JWKS outages, deep object probe, cohort-scaled alerts

- auth: tell 'token invalid' from 'keys unavailable'. An auth-refresh that cannot get keys keeps
  the control and its session unchanged (metric auth_refresh:keys_unavailable); the exp+grace
  close still applies. A control upgrade without keys gets 503 instead of 401 (shipped desktops
  treat both as a retried connect failure). The last good JWKS is kept in the colo's Cache API
  (24 h max age) as a fallback for cold isolates; a stored set lacking the token's key never
  reads as invalid.
- admin export ignores closing controls, like drain, so a just-drained host is not busy.
- GET /__health/object: canary HostRelay reads its SQLite (SELECT 1), no writes; per-source
  limited; 503 on failure or after 3 s. The alerts Worker probes it (object-probe, > 2 s alerts).
- alerts: MIN_VOLUME stays configurable, reconnect-surge rule (control hellos >= max(3x previous,
  RECONNECT_SURGE_MIN)), optional cloudflare-errors rule from the GraphQL Analytics API
  (CF_GRAPHQL_TOKEN). Rules moved to alerts-rules.ts.
- metrics: heartbeat events carry their 1-in-10 sampling factor as their last double.

* relay-edge: persisted JWKS trusted for 30 min, as the cells' stale window

* relay-edge alerts: opt-in reuse of the Analytics Engine read token for the GraphQL error rule
2026-11-07 23:38:28 -08:00
Jinjing e55b23c10e Prevent hover previews from interfering with context menus (#39757)
Keep workspace renames and input focus intact while dismissing hover
cards and help popovers on pointer or context menu interactions.
2026-11-07 23:37:21 -08:00
Jinwoo Hong 0d5a07fd0a fix(orchestration): report an idle worker once per coordinator message (#38484)
* fix(orchestration): report an idle worker once per coordinator message

A worker that ends turn after turn without worker_done (one that wakes on its own schedule, or one the user talks to) raised a HIGH escalation every turn. An idle finish now answers the coordinator's latest mail to the Dispatch, so it is reported once per mail; errors, prompts and exits still report every time.

Fixes the problem reported in #37868 by @gsxdsm, without a new worker command.

* fix(orchestration): key an idle report by mail that reached the worker, and make relay repeats a no-op

* fix(orchestration): key an idle report by coordinator mail the worker consumed

Delivery stamps can land before a chat admits the mail, and a bounded scan can miss the watermark; read mail is what a later turn answers. Helper reports into the same mailbox do not count.

* fix(orchestration): count consumed coordinator mail so out-of-order reads still re-arm the report

* fix(orchestration): leave heartbeats out of the consumed-mail key

* fix(orchestration): tell coordinator mail from helper reports by sender, not type

The consumed-mail key had its own type list, so a settled helper's status mail could still re-arm the idle report. A message counts unless its sender is one of the worker's own helpers, using the existing creator and assignee relationship.

* fix(orchestration): match a helper by its pane too, and only for mail it sent while helping

* fix(orchestration): leave any working peer's mail out of the consumed-mail key

A worker's own heartbeats and a sibling's broadcast also land in its mailbox. Mail from any worker of the Run sent while its Dispatch was open does not count, except the creator Dispatch that coordinates this one.

* fix(orchestration): count a peer's mail as a worker's only while its Dispatch was open
2026-11-08 02:30:03 -05:00