Every merge to main already triggers a nightly build, making the hourly
cron redundant. The skipped job was cosmetic (just echoed a message) and
caused confusing red X statuses when cancel-in-progress kicked in.
- Grant kTCCServiceScreenCapture in system-level TCC database (sudo)
and pre-date ScreenCaptureApprovals.plist to suppress Sequoia's
private window picker dialog
- Move recording start to right before xcodebuild test (skip build time)
- Trim leading black frames from video using ffmpeg blackdetect
- Add runner input: macos-15 (Sequoia) or macos-26 (Tahoe)
* Install ffmpeg via brew for screen recording
macos-15 GitHub runners don't have ffmpeg pre-installed.
* Clean up ffmpeg device detection and add fallback
Suppress noisy device listing errors, add fallback to index 1 if
detected index fails, upgrade warning to error on total failure.
* Add E2E test workflow with video recording and issue posting
New workflow_dispatch workflow (test-e2e.yml) that runs XCUITests on
GitHub-hosted macos-15 runners, records the virtual display, uploads the
video as an artifact, and posts results as an issue on cmux-dev-artifacts.
Includes scripts/run-e2e.sh for convenient triggering from the terminal.
* Print issue URL in workflow annotation and run-e2e.sh output
- Capture gh issue create output URL, print as ::notice annotation
- Search issues by run ID instead of grabbing most recent
Build immediately on merge instead of waiting for the hourly cron.
Concurrency group cancels in-progress builds when new commits land.
Depot macos runner replaced with GitHub macos-15 (similar perf, simpler).
* Split CI: GitHub runners for tests, Depot for perf regression
Unit/UI tests move to macos-15 (no queue wait, fast enough for test
suites). Typing-lag regression stays on Depot in a new tests-depot job
(needs stronger hardware). No duplicated test work between the two.
* Fix Xcode selection: add pipefail guard, use sort|tail for consistency
Address review comments:
- tests job: add || true to ls pipeline so fallback works under pipefail
- tests-depot job: use sort | tail -n 1 instead of head -n 1
* Move XCUITests from GitHub runner to Depot
tests (macos-15) now runs unit tests only. tests-depot (Depot) runs
UI tests and the typing-lag regression, reusing the same build.
* Add workspace-churn typing lag regression and fix
* Fix CI build for debug stress split calls
* Stabilize lag regression gate for low baseline latency
* Add macOS compatibility CI: unit tests + smoke test on macos-14/15
New workflow runs on GitHub-hosted macos-14 and macos-15 runners
(matrix strategy). Each run: unit tests via cmux-unit scheme, then
a smoke test that builds the app, launches it, sends a command via
the socket, and verifies it stays alive for 15 seconds.
* Select latest Xcode on runner (fix macos-14 Swift tools version)
macos-14 runners default to Xcode 15.4, but sentry-cocoa needs
Swift tools version 6.0 (Xcode 16+). Pick the latest Xcode_*.app
instead of the default symlink.
* Launch app binary directly in smoke test for better CI compatibility
Using `open` can fail silently on CI runners. Launch the binary
directly with env vars set, capture stdout/stderr, and add process
health checks with diagnostic output (debug log tail, crash reports)
on failure.
* Migrate all workflows from self-hosted Mac Mini to Depot runners
Move CI, nightly, and release workflows to depot-macos-latest. Replace
zig GhosttyKit builds with pre-built xcframework downloads. Add virtual
display for CI UI tests. Remove concurrency groups (ephemeral VMs don't
need them).
* Add per-test timeout to CI UI tests to prevent hangs on Depot
SidebarResizeUITests hangs on headless Depot runners due to mouse drag
simulation issues. Adding -maximum-test-execution-time-allowance 120
(matching test-depot.yml) ensures individual tests timeout after 2 min
instead of blocking the entire run.
* Skip SidebarResizeUITests in CI on Depot runners
Mouse drag simulation hangs on headless Depot runners even with a
virtual display. The per-test timeout doesn't prevent the hang either.
Skip this test class in CI; it still runs fine on local machines.
* Handle XCTExpectFailure in CI UI tests (exit 65 with 0 unexpected)
xcodebuild exits 65 even when all failures use XCTExpectFailure. Add
the same expected-failure handling from the unit test step so browser
focus tests (which are expected to fail on headless runners) don't
break CI.
* Add virtual display for headless Depot runners
Depot macOS runners have no physical display, causing XCUITests to fail
with "Failed to activate application (current state: Running Background)".
This adds a small ObjC tool that creates a virtual display using the
private CGVirtualDisplay API before tests run.
* Split self-hosted concurrency groups per workflow
CI, nightly, and release all shared `self-hosted-build`, so the hourly
nightly cancelled in-progress CI runs. Now each workflow has its own
group (self-hosted-ci, self-hosted-nightly, self-hosted-release).
CI also gets cancel-in-progress: true so rapid pushes cancel stale runs.
Depot macOS runners have no physical display, causing XCUITests to fail
with "Failed to activate application (current state: Running Background)".
This adds a small ObjC tool that creates a virtual display using the
private CGVirtualDisplay API before tests run.
- Display diagnostics step to debug headless display issues
- test_filter input to run specific XCUITest classes
- test_timeout input (default 120s) to prevent test hangs
* Run all XCUITests in CI instead of just UpdatePillUITests
Removes the -only-testing:cmuxUITests/UpdatePillUITests filter so
new test classes are picked up automatically. No more workflow edits
needed when adding tests.
* Add skip_unit_tests input to test-depot workflow
After rm -rf of the SPM cache dir, recreate it as an empty directory
so binary target downloads (e.g. Sentry.xcframework.zip) don't hit
"already exists in file system" errors from stale artifacts on the
self-hosted runner.
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
build-ghosttykit shared the self-hosted-build concurrency group with
CI tests, causing one to get cancelled when both trigger on the same
push. Most runs are no-ops (xcframework already exists), so Depot is
a good fit. Eliminates the red X on pushes to main.
* Switch CI and Build GhosttyKit to Depot macOS runners
Moves the tests job (unit + UI) and build-ghosttykit job from
self-hosted to depot-macos-latest so CI doesn't block the local
Mac Mini. CI tests now download the pre-built xcframework from
the ghostty releases instead of building with zig. Nightly and
release workflows stay on self-hosted for signing/notarization.
* Add on-demand Depot test workflow, revert CI changes
Adds test-depot.yml (workflow_dispatch) for running tests on Depot
macOS runners during local dev without tying up the Mac Mini.
Reverts ci.yml and build-ghosttykit.yml back to self-hosted.
* Add `cmux <path>` to open directories and Homebrew binary stanza
CLI: `cmux .` or `cmux /path/to/dir` opens a new workspace at the
given directory. If the app isn't running, it launches first and waits
for the socket. Also adds `--cwd` flag to `new-workspace`.
Server: `workspace.create` now accepts an optional `cwd` parameter,
passed through to `TabManager.addWorkspace(workingDirectory:)`.
Homebrew: adds `binary` stanza to the cask so `cmux` CLI is globally
available after `brew install --cask cmux`. Updated both the cask file,
the CI workflow template, and the manual release script so automated
version bumps preserve the stanza.
* Address review: validate cwd type, fix socket detection, propagate errors
- looksLikePath now also matches paths containing `/` (e.g. `foo/bar`)
- openPath uses socket connection attempt instead of fileExists to detect
whether the app is running (Unix sockets may not appear on filesystem)
- launchApp/activateApp now throw instead of swallowing errors with try?
- Server validates that cwd param is a string, returns invalid_params error
if wrong type is passed
* Set cmux TestAction to Debug for UI tests
* Broaden XCTest detection for debug launch gate
* Fix AutomationSocketUITests launch hang in CI
* Stabilize CI Swift package resolution for test jobs
* Stabilize Xcode Cloud UI test focus and socket handling
* Add Xcode Cloud pre-xcodebuild submodule bootstrap
* Harden Xcode Cloud bonsplit bootstrap fallback
* Set up full test suite in CI and Xcode Cloud
Add build-ghosttykit.yml workflow to pre-build and publish
GhosttyKit.xcframework as a GitHub release on manaflow-ai/ghostty,
keyed by submodule SHA. Add ci_scripts/ci_post_clone.sh for Xcode
Cloud to download the pre-built xcframework with retry logic. Create
cmux-ci scheme that runs both cmuxTests and cmuxUITests. Switch the
CI tests job from running a single UI test class to the full suite.
* Run unit tests + single UI test class on self-hosted runner
The self-hosted runner can't launch the full app for UI tests (no GUI
session), so run all unit tests via cmux-unit scheme and keep the
original UpdatePillUITests as a smoke test. Full UI test suite runs
on Xcode Cloud which has proper macOS GUI support.
* Handle expected test failures in unit tests step
xcodebuild returns exit code 65 even for expected failures
(XCTExpectFailure). Parse the summary line to only fail the CI job
when there are unexpected failures.
XCUITest launches the app as a separate process that doesn't inherit
XCTest env vars (XCTestConfigurationFilePath, etc.), so
isRunningUnderXCTest() returns false. The app then hits
shouldBlockUntaggedDebugLaunch() and exits with code 64, causing the
test runner to hang waiting for the app to launch.
Fix: detect CMUX_UI_TEST_* env vars set via XCUIApplication.launchEnvironment
and skip the launch guard. Also revert the failed CMUX_TAG ci.yml workaround.
Debug builds refuse to launch without CMUX_TAG to prevent accidental
untagged launches. XCUITest launches the app as a separate process
without XCTest env vars, so the app's isRunningUnderXCTest() check
fails and the app calls exit(64). The test runner then waits forever
for the app to report back. Set CMUX_TAG=ci in the CI env.
The Mac Mini runner has no dev certificates, so xcodebuild produces
an ad-hoc signed app. Gatekeeper rejects it, causing XCUITest to hang
forever at app.launch(). Split build and test phases, strip the ad-hoc
signature between them so the app can launch.
Org policy now requires actions pinned to immutable SHAs instead of
mutable version tags. Pin actions/checkout, actions/github-script,
softprops/action-gh-release, and oven-sh/setup-bun across all workflows.
* Upgrade Sentry: tracing, breadcrumbs, dSYM upload
- Enhanced Sentry SDK init with performance tracing (10% sample),
explicit app hang timeout, stack trace attachment, and HTTP
failure capture
- Added breadcrumbs for key user actions: workspace switch/create/close,
split creation, command palette open/close, app focus — these give
context to hang/crash reports
- Added dSYM upload step to nightly and release CI workflows so hang
stacks are fully symbolicated (requires SENTRY_AUTH_TOKEN secret)
- Created SentryHelper.swift with lightweight breadcrumb helper
Closes https://github.com/manaflow-ai/cmux/issues/365
* Remove command palette breadcrumbs
Not useful for hang diagnosis — keep only workspace/tab/split/focus
breadcrumbs that correlate with heavy operations.
The decide job already skips when main HEAD matches the nightly tag,
so this only builds when there are actual changes. Hourly means users
get nightly updates within an hour of merging to main.
Follows the same pattern as AppIcon-Debug (orange DEV banner) but with
a purple banner and "NIGHTLY" text. The nightly CI workflow now passes
ASSETCATALOG_COMPILER_APPICON_NAME=AppIcon-Nightly to xcodebuild so
the nightly app gets its own distinct icon.
Includes scripts/generate_nightly_icon.py for regenerating the icons
from the production AppIcon source files.
The nightly build is now a distinct app called "cmux NIGHTLY" with
bundle ID com.cmuxterm.app.nightly, allowing side-by-side installation
with the stable release. The nightly appcast URL is baked into the
app's Info.plist by CI, so no in-app channel switching is needed.
- Nightly workflow: rename app to "cmux NIGHTLY", set bundle ID to
com.cmuxterm.app.nightly, hardcode nightly Sparkle feed URL, publish
DMG as cmux-nightly-macos.dmg
- Remove "Receive Nightly Builds" toggle from settings
- Remove UpdateChannelSettings enum and simplify feed URL resolution
to just use SUFeedURL from Info.plist
- Remove UpdateChannelSettingsTests (no longer applicable)
Root cause: update-homebrew.yml triggered on release:published, which fires
before softprops/action-gh-release finishes uploading assets. The workflow
downloaded a 404 page instead of the DMG and committed its SHA.
Fix:
- Change trigger from release:published to workflow_run (fires after the
release workflow completes, guaranteeing assets are uploaded)
- Add download validation with retries and file size checks
- Add SHA verification step before committing to the cask
- Add homebrew cask update to build-sign-upload.sh for local releases
- Add regression test (tests/test_homebrew_sha.sh)
- Update /release and /release-local skills with homebrew verification steps
Fixes#110
Remove -Dxcframework-target=native from CI and release workflows,
defaulting to universal (matching upstream ghostty). The native target
produces a macos-arm64 xcframework slice that causes Xcode to link the
final binary differently (~70KB more __text), resulting in menubar and
right-click lag on M1 Max. The arm64 static libraries are byte-for-byte
identical between native and universal builds - the difference is purely
in how Xcode resolves the xcframework slice.
The -Dxcframework-target flag controls platform slices (native=macOS
only, universal=macOS+iOS), not CPU microarchitecture. Removing it
caused CI to attempt iOS builds which fail due to missing Metal
iOS toolchain on the runner.
* Fix zsh git branch refresh race after cwd change
* Clarify intentional duplicate cwd check in git refresh path
* Add Metal Toolchain download step to CI and release workflows
Fixes build failure when compiling Metal shaders for iOS xcframework
targets — the self-hosted runner needs `xcodebuild -downloadComponent
MetalToolchain` installed before `xcrun metal` can run.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
The self-hosted runner (M4 Mac Mini) was building GhosttyKit with
-Dxcframework-target=native, producing M4-tuned binaries. This caused
menubar and right-click lag on M1 machines. Dropping the flag defaults
to universal, which works well across all Apple Silicon chips.
Hardened runtime's library validation was verifying every dylib on load,
causing noticeable UI lag. Add entitlements file with
disable-library-validation to fix while keeping notarization support.
