starred/orca
1
0
Fork 0
mirror of https://github.com/stablyai/orca.git synced 2026-09-22 16:02:32 +00:00
Code Issues Packages Projects Releases Wiki Activity
Files
c56ff58dd736bb163fa682ca995ef9ae996a2f5a
orca/.github/workflows/computer-e2e.yml
T
Neil 7da9368b78 fix(terminal): fence detached daemon endpoint ownership (#12709) 
* fix(terminal): fence daemon endpoint ownership

* fix(terminal): clean failed daemon PID claims

* fix(terminal): close daemon ownership review gaps

* test(daemon): release startup IPC in boot smoke

* test(daemon): mirror production stdio in boot smoke

* fix(daemon): exit after rpc shutdown cleanup

* fix(terminal): make the socket name the daemon endpoint authority

The reported failure was a live daemon hosting PTYs that nothing could
reach: terminals acknowledged input and never ran it, listings diverged
from reality, and restarting the app never helped because the detached
helper survived. The ownership fence added for it could not fire in the
sequence that produces the split brain.

libuv unlinks the pathname a server bound to when that server closes,
with no ownership check. A daemon that lost its endpoint name therefore
deleted whichever socket then sat at that path — including a live
replacement's — stranding a daemon that still hosted every session.
Bind a private same-directory name and hard-link it into place instead:
libuv can only ever unlink our own bind name, the exclusive link is a
kernel-enforced endpoint claim, and the canonical name is removed only
under an inode ownership check. The bind name replaces the basename
rather than extending it, so it cannot overflow sun_path.

killStaleDaemon removed the PID record unconditionally immediately
before every fork, so the exclusive PID claim was always uncontested at
bind time. It also unlinked a live daemon's endpoint whenever a connect
probe merely timed out, and treated a `ps` timeout as proof of PID
recycling. Now only positive evidence of a dead endpoint authorizes
reclaiming it, SIGKILL is confirmed rather than assumed, and a daemon
that cannot be proven stopped keeps its record and endpoint while the
launcher refuses to fork beside it.

A daemon whose endpoint was taken over now retires itself, draining
rather than killing, so an unreachable orphan stops being permanent.

A repaired PID record re-derives entryPath, appVersion and the Linux
incarnation markers from the authenticated owner instead of dropping
them; without appVersion a healthy daemon read as a permanently stale
bundle and, on Windows, went unpinned against daemon-host pruning.
Repair failure now fails open — abandoning a healthy daemon over a pid
file write cost every persistent terminal on the machine.

Also: treat only ENOENT as an unclaimed record so a Windows file lock is
not reported as an ownership conflict; settle start() before close() so
an accepted connection cannot defer it forever; sweep abandoned claim
and bind names; and type the endpoint-identity seam so a rename cannot
silently disable the fence.

Adds a real-process handover smoke that reproduces the failure with two
daemons racing one endpoint, and wires it into the native-smoke job.

* fix(daemon): retire only on proven endpoint ownership loss

The ownership watchdog read a null identity for any stat failure, so a
transient EACCES or EIO on the runtime directory would retire a daemon
that was still serving every terminal on the machine. Distinguish "the
entry is gone" from "the probe failed" and act only on the former.

Also require the loss to persist across two polls: a replacement
publishes by unlink-then-link, and a single observation can land in that
gap.

* fix(daemon): source repaired ownership metadata from the authenticated hello

Adversarial review found three defects in the previous two commits.

Re-deriving entryPath from the owner's command line truncated it at the
first space. A command line is a single space-joined string, so
`C:\Program Files\Orca\...` and `/Applications/Orca 2.app/...` came back
as `"C:\Program` and `/Applications/Orca`. getDaemonLaunchIdentity treats
a present entryPath as authoritative, so a healthy daemon read as
`different_app_path` and was killed and re-forked — worse than the
missing-metadata case the derivation was added to fix. Carry entryPath
and appVersion as optional fields on the daemon hello identity instead:
the daemon already has both from its own argv, and per
docs/reference/remote-wire-compatibility.md a new optional field is safe
because every reader falls back when it is absent. This also removes a
synchronous `ps` spawn from the Electron main thread during startup.

`start()` rolled back the PID record even when it never published one.
Losing the endpoint link now runs that path, and the ownership-checked
unlink briefly renames the incumbent's record aside — enough to strand a
live daemon's ownership. Roll back only what we actually wrote.

publishDaemonSocketPath read its identity from the canonical name after
linking, so a concurrent unlink returned null: no ownership watchdog and
no endpoint cleanup on any shutdown path. Read it from the bound name
before linking, which shares the inode.

Refusing to fork beside an unconfirmed daemon left the user with no
daemon at all and no in-app recovery, since restart re-entered the same
fence. We have just proved something answers the endpoint, so adopt it
in degraded mode: live sessions keep working, fresh terminals run
locally. SIGTERM is also individually guarded now — an EPERM fell into
the blanket catch and reported "nothing alive", authorizing the very
duplicate this fence exists to prevent.

Also reset the ownership-loss streak on an inconclusive probe so the
confirmations are consecutive, and sweep scratch names before the launch
so a failed launch still reclaims them.
2026-08-05 21:12:32 -07:00

242 lines
11 KiB
YAML
Raw Blame History

name: Computer-use e2e
on:
pull_request:
paths:
- '.github/workflows/computer-e2e.yml'
- 'config/electron-builder.config.cjs'
- 'config/scripts/build-computer-macos.mjs'
- 'config/scripts/build-windows-cli-launcher.mjs'
- 'config/scripts/build-windows-cli-launcher.test.mjs'
- 'config/scripts/computer-e2e-workflow.test.mjs'
- 'config/scripts/macos-computer-helper-owner-loss-benchmark.mjs'
- 'config/scripts/macos-computer-helper-owner-loss-group-recovery.test.mjs'
- 'config/scripts/macos-computer-helper-owner-loss-metrics.mjs'
- 'config/scripts/macos-computer-helper-owner-loss-processes.mjs'
- 'config/scripts/macos-computer-helper-owner-loss-processes.test.mjs'
- 'config/scripts/macos-computer-helper-owner-loss-trial-cleanup.mjs'
- 'config/scripts/computer-use-modifier-safety.test.mjs'
- 'config/scripts/computer-use-skill-guidance.test.mjs'
- 'config/scripts/computer-use-smoke.mjs'
- 'config/scripts/computer-use-smoke.test.mjs'
- 'config/scripts/daemon-boot-smoke.mjs'
- 'config/scripts/daemon-endpoint-handover-smoke.mjs'
- 'config/scripts/windows-daemon-workspace-close-repro.mjs'
- 'config/scripts/verify-computer-native.mjs'
# Why: the native-smoke job boots the built terminal daemon under plain
# Node, so any change to the daemon bundle graph or the main build must
# re-run it (the v1.4.129-rc.1 daemon outage shipped with green CI).
- 'electron.vite.config.ts'
- 'config/build-plugins/**'
- 'src/main/daemon/**'
- 'native/computer-use-macos/**'
- 'native/computer-use-linux/**'
- 'native/computer-use-windows/**'
- 'native/windows-cli-launcher/**'
- 'skills/computer-use/SKILL.md'
- 'src/cli/**'
- 'src/main/computer/**'
- 'src/main/runtime/rpc/dispatcher.ts'
- 'src/main/runtime/rpc/errors.ts'
- 'src/main/runtime/rpc/methods/computer*.ts'
- 'src/main/ssh/ssh-remote-cli-launcher.ts'
- 'src/main/ssh/ssh-remote-cli-launcher.test.ts'
- 'src/shared/computer-use-*.ts'
- 'tests/e2e/computer-linux.e2e.ts'
- 'tests/e2e/computer-mac.e2e.ts'
- 'tests/e2e/computer-mac-safari.e2e.ts'
- 'tests/e2e/computer-windows.e2e.ts'
- 'tests/e2e/computer-windows-store.e2e.ts'
- 'tests/e2e/helpers/computer-cli-driver.ts'
- 'tests/e2e/helpers/computer-driver.ts'
- 'tests/e2e/vitest.config.ts'
workflow_dispatch:
schedule:
- cron: '0 6 * * *'
jobs:
native-smoke:
if: github.event_name == 'pull_request'
permissions:
contents: read
strategy:
fail-fast: false
matrix:
os: [ubuntu-22.04, windows-latest]
runs-on: ${{ matrix.os }}
steps:
- uses: actions/checkout@v6
with:
persist-credentials: false
- uses: actions/setup-node@v6
with:
node-version-file: package.json
- uses: pnpm/action-setup@v6
with:
run_install: false
- if: runner.os == 'Linux'
run: sudo apt-get update && sudo apt-get install -y python3 python3-gi gir1.2-atspi-2.0 at-spi2-core gedit xvfb xclip xdotool
# Why: pnpm's bundled node-gyp can ship gyp_main.py without execute
# permission on Linux runners; node-pty's install fallback then fails
# before this smoke job can exercise the native package.
- name: Use external node-gyp to avoid pnpm's bundled copy (Linux only)
if: runner.os == 'Linux'
run: |
npm install -g node-gyp@11.5.0
echo "npm_config_node_gyp=$(npm root -g)/node-gyp/bin/node-gyp.js" >> "$GITHUB_ENV"
- run: pnpm install --frozen-lockfile
# Why: without --config, bare vitest ignores config/vitest.config.ts (there
# is no root config) and falls back to the 5s default timeout with no
# Windows worker cap, so the real csc.exe launcher-compile tests time out
# on hosted Windows. Use the shared config so this job matches pnpm test.
- run: >-
pnpm vitest run --config config/vitest.config.ts
config/scripts/build-windows-cli-launcher.test.mjs
src/main/ssh/ssh-remote-cli-launcher.test.ts
config/scripts/computer-e2e-workflow.test.mjs
config/scripts/macos-computer-helper-owner-loss-group-recovery.test.mjs
config/scripts/macos-computer-helper-owner-loss-processes.test.mjs
config/scripts/computer-use-modifier-safety.test.mjs
config/scripts/computer-use-skill-guidance.test.mjs
config/scripts/computer-use-smoke.test.mjs
src/main/computer/computer-provider-lifecycle.test.ts
src/main/computer/computer-provider-unavailable-message.test.ts
src/main/computer/sidecar-client.test.ts
src/main/computer/macos-native-provider-client.test.ts
src/main/computer/macos-native-provider-socket.test.ts
src/main/computer/macos-computer-use-permissions.test.ts
src/main/computer/macos-computer-use-permission-status.test.ts
src/main/computer/desktop-script-provider-client.test.ts
src/main/computer/desktop-script-provider-cache.test.ts
src/main/computer/desktop-script-provider-actions.test.ts
src/main/computer/desktop-script-provider-cache-lifecycle.test.ts
src/main/computer/desktop-script-provider-errors.test.ts
src/main/computer/desktop-script-provider-action-errors.test.ts
src/shared/computer-use-error-recovery.test.ts
src/shared/computer-use-key-spec.test.ts
src/cli/format.test.ts
src/cli/handlers/computer.test.ts
src/cli/handlers/computer-action-routing.test.ts
src/cli/handlers/computer-action-validation.test.ts
src/cli/handlers/computer-state-formatting.test.ts
src/cli/specs/computer.test.ts
src/cli/index.test.ts
src/main/runtime/rpc/dispatcher-computer-errors.test.ts
src/main/runtime/rpc/errors.test.ts
src/main/runtime/rpc/methods/computer.test.ts
src/main/runtime/rpc/methods/computer-actions.test.ts
src/cli/runtime/envelope-schema.test.ts
src/shared/remote-runtime-client.test.ts
- run: pnpm verify:computer-native
- run: pnpm build:cli
- run: pnpm build:electron-vite
# Why: boot the BUILT daemon-entry under plain Node the way production
# forks it. v1.4.129-rc.1 shipped a daemon that exited at module load
# (leaked electron require) while every other check passed; this fails
# the PR when the built daemon cannot start on ubuntu-22.04 / windows.
- name: Daemon boot smoke
run: node config/scripts/daemon-boot-smoke.mjs
# Why: a daemon that lost its endpoint name used to delete the replacement's
# socket on exit, stranding a live daemon that hosts PTYs nothing can reach.
# Only two real processes racing the same endpoint reproduce it.
- name: Daemon endpoint handover smoke
run: node config/scripts/daemon-endpoint-handover-smoke.mjs
# Why: workspace removal overlaps graceful renderer teardown with the
# forced main-process sweep; exercise that exact pair against real ConPTY.
- name: Windows daemon workspace-close repro
if: runner.os == 'Windows'
run: node config/scripts/windows-daemon-workspace-close-repro.mjs
mac-native-owner-smoke:
if: github.event_name == 'pull_request'
runs-on: macos-15
permissions:
contents: read
steps:
- uses: actions/checkout@v6
with:
persist-credentials: false
- uses: actions/setup-node@v6
with:
node-version-file: package.json
- uses: pnpm/action-setup@v6
with:
run_install: false
- run: pnpm install --frozen-lockfile
- name: Owner-loss benchmark process cleanup
run: >-
pnpm vitest run
config/scripts/macos-computer-helper-owner-loss-group-recovery.test.mjs
config/scripts/macos-computer-helper-owner-loss-processes.test.mjs
- name: Authenticated helper owner-loss smoke
run: pnpm bench:macos-computer-helper-owner-loss --expect reaped --trials 1
- name: Swift tests and signed universal helper verification
run: pnpm verify:computer-native
mac:
# macOS Accessibility and Screen Recording require user-granted TCC entries.
# Keep this on manual/scheduled permission-bearing runners instead of PR CI.
if: github.event_name != 'pull_request'
runs-on: macos-15
env:
ORCA_COMPUTER_E2E: '1'
steps:
- uses: actions/checkout@v6
- uses: actions/setup-node@v6
with:
node-version-file: package.json
- uses: pnpm/action-setup@v6
with:
run_install: false
- run: pnpm install --frozen-lockfile
- run: pnpm build:computer-macos
- run: pnpm verify:computer-native
- run: pnpm build:cli
- run: pnpm build:electron-vite
- run: pnpm test:e2e:computer --reporter=verbose tests/e2e/computer-mac.e2e.ts tests/e2e/computer-mac-safari.e2e.ts
linux:
if: github.event_name != 'pull_request'
runs-on: ubuntu-22.04
env:
ORCA_COMPUTER_E2E: '1'
ACCESSIBILITY_ENABLED: '1'
steps:
- uses: actions/checkout@v6
- uses: actions/setup-node@v6
with:
node-version-file: package.json
- uses: pnpm/action-setup@v6
with:
run_install: false
- run: sudo apt-get update && sudo apt-get install -y build-essential python3 python3-gi gir1.2-atspi-2.0 gedit at-spi2-core xvfb xclip xdotool
# Why: keep scheduled Linux e2e on the same native install path as PR
# smoke and pr.yml's verify job.
- name: Use external node-gyp to avoid pnpm's bundled copy (Linux only)
run: |
npm install -g node-gyp@11.5.0
echo "npm_config_node_gyp=$(npm root -g)/node-gyp/bin/node-gyp.js" >> "$GITHUB_ENV"
- run: pnpm install --frozen-lockfile
- run: pnpm verify:computer-native
- run: pnpm build:cli
- run: pnpm build:electron-vite
- run: xvfb-run --auto-servernum dbus-run-session -- pnpm test:e2e:computer --reporter=verbose tests/e2e/computer-linux.e2e.ts
windows:
if: github.event_name != 'pull_request'
runs-on: windows-latest
env:
ORCA_COMPUTER_E2E: '1'
steps:
- uses: actions/checkout@v6
- uses: actions/setup-node@v6
with:
node-version-file: package.json
- uses: pnpm/action-setup@v6
with:
run_install: false
- run: pnpm install --frozen-lockfile
- run: pnpm verify:computer-native
- run: pnpm build:cli
- run: pnpm build:electron-vite
- run: pnpm test:e2e:computer --reporter=verbose tests/e2e/computer-windows.e2e.ts tests/e2e/computer-windows-store.e2e.ts
Reference in New Issue View Git Blame Copy Permalink