hermes-node now builds and publishes the prebuilt Hermes and embed library
itself (utils/build-apple-prebuilt.sh, released as prebuilt-<commit> by its
own CI), since they depend only on its commit. Drop Orbit's build script and
publishing workflow; hermes-prebuilt.sh downloads the release for the pinned
HERMES_NODE_REF. Bump hermes-node to f156186.
Without the flag the app builds exactly as before: React Native's stock Hermes,
React Native's own JSI and the pkg-built CLI binaries (Release builds only),
so main's Podfile.lock, pkg config, entitlements, EAS and e2e workflows are
unchanged.
pod install with ORBIT_HERMES_CLI=1 (or "hermesCli": "true" in
Podfile.properties.json) opts into Hermes V1 from hermes-node's pin, its JSI,
and the orbit-cli launcher + orbit-cli.hbc in every configuration (yarn
archive:hermes / codesign:hermes). pod install records the choice in
Pods/.orbit-hermes-cli for the build phases, and without the flag puts React
Native's JSI back.
The app bundles the CLI as Contents/Resources/cli/ either way and runs
cli/orbit-cli when present, otherwise the pkg binary for the architecture.
React Native has to compile against the JSI its Hermes was built with, and
react-native-macos 0.83 ships an older one. Instead of carrying the newer JSI
as a 4.3k-line patch-package diff, pod install copies it from the pinned Hermes
(the prebuilt tarballs now include it; the source fallback reads it from the
Hermes checkout) with scripts/vendor-jsi.sh. A jsi/hermes.h forwarding header
covers React Native still including the old name, so ReactInstance.cpp needs
no edit and patches/react-native-macos+0.83.0.patch is back to main's.
The copy can't drift from Hermes, and a build phase on React-jsi fails with
"run pod install" if a later yarn install restored React Native's own JSI.
Prebuilt key 52d8a44a42ecab19 is published.
scripts/build-hermes-prebuilt.sh builds hermes-engine pod tarballs (Debug with
the debugger, Release; universal, hermesc + headers, using RN macOS's own
configure_apple_framework) and the hermes-node embed library, keyed by a hash
of everything they're built from (pinned hermes-node, react-native-macos and
our patch, the build script). They're published as hermes-prebuilt-<key>
releases of gabrieldonadel/hermes-node, by the new Hermes prebuilts workflow.
pod install downloads and verifies them (scripts/hermes-prebuilt.sh) and points
HERMES_ENGINE_TARBALL_PATH at the Debug tarball; the Select Hermes flavor build
phase swaps in the Release framework for Release builds; hermes-node.sh build
uses the prebuilt embed library. Without a published prebuilt for the key,
everything falls back to building from source as before.
Also bump hermes-node to f1a9210 (fixes the embed library growing on rebuilds).
hermes-node now provides fetch/Headers/Request/Response and AbortSignal
globals like Node, so the CLI runs the same code on both. The
macos-export-certificate-and-key addon ships in its normal layout with
bindings and file-uri-to-path instead of a patched loader. Bump hermes-node
to 5c1c723.
ipa-resign is obfuscated with a self-defending check that regex-tests
Function.prototype.toString() and spins forever on Hermes, which returns
"[bytecode]", and it builds a dynamic-import helper with
new Function('u', 'return import(u)'), which Hermes can't compile. A bundler
plugin prepends Hermes's 'show source' directive and a module-local Function
that turns that one shape into a require()-based import.
Also resolve ipa-resign's directory from its main entry (its exports hide
package.json) and pin esbuild as a devDependency.
- macos/scripts/hermes-node.sh fetches and builds a pinned hermes-node; the
Podfile builds Hermes from the facebook/hermes commit it pins
(RCT_HERMES_V1_ENABLED, REACT_NATIVE_OVERRIDE_HERMES_DIR) and checks that the
JSI vendored into react-native-macos matches
- Regenerate the react-native-macos patch without rename detection: the
hermes.h -> hermes-interfaces.h rename made patch-package fail on a clean
install
- apps/cli yarn archive builds dist/: universal orbit-cli launcher,
orbit-cli.hbc, the macos-export-certificate-and-key addon built for ARCHS
without bindings, and ipa-resign's helpers; codesign signs every Mach-O
with a timestamp. Drop pkg and the V8-only JIT entitlements
- Xcode's Build CLI phase (all configurations) builds it against the app's
hermesvm.framework; the app bundles it as Contents/Resources/cli and runs
cli/orbit-cli
- orbit shell-command recognizes the launcher (was process.pkg)
- Install node-fetch as global fetch when the runtime has none
- EAS: the CLI is built by the Xcode archive; install cmake/ninja. E2E: cache
hermes-node, run on CLI changes
- Remove the in-process spike
Attach HermesNodeEmbedder to RN's runtime on init, drive its libuv loop
through RN's RuntimeScheduler, and expose __nodeRequire. The menu bar's JS
uses fs/os/child_process/https and runs the CLI's listDevicesAsync
in-process. Links /tmp/hn-embed-out/lib/libhermes-node-embed.a built by
hermes-node's utils/build-apple-embed-lib.sh (spike-only paths).
* [cli] Start remote sessions without a maximum duration
Orbit set a maximum duration of 30 minutes on each Expo Simulators session.
With no value, the session gets the maximum duration of the account's plan,
as it does from EAS CLI and expo.dev. --max-run-time still sets a shorter one.
* [cli] Add changelog entry for the session duration
Adds Expo Simulators sessions to Orbit as remote devices, and reworks the popover around accounts, projects and device actions.
- CLI: boot-device and install-and-launch accept an eas: device ID. New commands list-remote-simulators, list-remote-simulator-options, stop-remote-simulator and stop-device.
- Popover: an account row, a row of project tiles that can launch the latest app, a default device for each platform, buttons on device rows, and a Remote devices section with running and ended sessions and a default controller.
- Windows: New simulator session starts a session. The Settings pane Menu sets the sections, the order of devices and the hidden devices.
- Quit and log out ask first when sessions still run.
- Includes the fixes from the Windows and Linux tests (#391).
* [menu-bar] Fix deep links being opened as files on Windows and Linux
Co-authored-by: gabrieldonadel <[email protected]>
* Add PR number to changelog
Co-authored-by: gabrieldonadel <[email protected]>
---------
Co-authored-by: expo-tuft[bot] <288127324+expo-tuft[bot]@users.noreply.github.com>
Co-authored-by: gabrieldonadel <[email protected]>
* [cli][menu-bar] Add option to install the orbit command in PATH
* Apply batched suggestions from code review
Co-authored-by: Tomasz Sapeta <[email protected]>
* [cli][menu-bar] Address review feedback on the orbit shell command
* Bump ipa-resign to 0.0.20
---------
Co-authored-by: Tomasz Sapeta <[email protected]>
* [cli][menu-bar] Auto-renew resigned apps every 7 days
Free Apple ID provisioning profiles expire after 7 days, so a resigned build
stops opening weekly. Track each resigned app and re-sign it automatically
before it expires; also add the App ID quota tooling repeated signing needs.
- Record store (ResignedApps): every resign is saved with its device, Apple
ID, managed original/resigned ipa copies, and profile expiry. The CLI gains
--managed-dir (keep a durable copy of the original ipa) and a
cleanup-resigned-apps command.
- Renewal engine (useResignedAppRenewals, mounted in Core): re-signs records
due within 48h on app start, a 6-hour tick, popover focus, and after
sign-in; installs the renewed build when the device is connected, else on
reconnect. A popover attention row prompts re-auth when the session expired.
- Settings "Resigned apps" section: expiry countdowns, per-app and global
auto-renew toggles, Renew now / Remove; DebugMenu dev actions.
- App ID management: list-app-ids / delete-app-id CLI, the "Manage Apple App
IDs" window, and the resign quota-recovery flow that opens it when the
10-App-IDs-per-7-days limit is hit.
Stacked on the base re-signing PR.
* [cli][menu-bar][eas-shared] Add 'Keep me signed 'in and app icons for resigned apps
* [cli][menu-bar][eas-shared] Handle the free Apple ID three-app limit and show a placeholder icon
* Bump ipa-resign to 0.0.18
* [cli][menu-bar][eas-shared] Check the free Apple ID app limit before installing and show app names
* [cli][menu-bar] Address review feedback on resigned app renewals
* [cli][menu-bar] Add iOS re-signing with a free Apple ID
Re-sign an installed build that fails code-signing so it runs on a personal
device, using a free Apple ID. When an install fails with
APPLE_APP_VERIFICATION_FAILED, Orbit offers to re-sign and install.
- CLI: `resign-ipa` and `apple-id-auth` (sign-in / verify-2fa with SMS
fallback / sign-out) commands, driven by ipa-resign.
- Menu-bar: the resign offer + determinate progress in the popover, an Apple
ID sign-in window (6-digit 2FA input, resend, SMS fallback), error-copy
mapping, first-install "Untrusted Developer" trust instructions, and an
Apple ID section in Settings (signed-in state + sign out).
- Session expiry is treated as an automatic logout: the stored Apple ID is
forgotten and every "signed in as" surface updates, with one-tap re-auth.
- Electron/macOS wiring for the ipa-resign native anisette + zsign helpers,
transient-env-var plumbing to pass the password to the CLI, and the
APPLE_APP_VERIFICATION_FAILED mapping that triggers the offer.
[email protected].
* [menu-bar] Fix Electron CI: ensure extraResource dirs exist; drop useless return
- forge.config: create ./bin and ./anisette in the generateAssets hook so
packaging (and Electron E2E, which runs 'yarn package' without
'build:helpers') does not fail with ENOENT on the missing extraResource
dirs. build:helpers still populates them for real builds.
- build-helpers: remove a no-useless-return that failed eslint --max-warnings 0.
* Bump ipa-resign to 0.0.13
* Fix CLI error handling
* Update AppleID auth UI
* Fix Apple ID persistance
* Fix Apple connection lost
* Fix opening settings menu
* Support paid developer accounts in resign copy and show the install step
* Detect builds that need resigning before attempting the install
* Add changelog entry
* [menu-bar][macOS] Use @expo/ui for the settings checkboxes and build progress bar
* [menu-bar][macOS] remove local modules
* [e2e] Skip renderer screenshot when the window is hidden
* Add changelog entry
* [e2e] Add main-process and chromedriver diagnostics for the Linux hang
* [e2e] Capture main-process stacks and Xvfb state on Linux failures
* [gh] Sign Windows builds with Azure Trusted Signing
* [gh] Rename deprecated trusted-signing-account-name input
* [electron] Set Windows version metadata to 650 Industries
* [orbit] Add Wi-Fi pairing support for Android devices
Add the ability to pair and connect a physical Android device over Wi-Fi
using ADB wireless debugging pairing codes.
- eas-shared: add pairAndroidDeviceAsync / connectAndroidDeviceAsync ADB helpers
- common-types: add PairAndroidDeviceResult command type
- cli: add `pair-android-device` command (pairing address, code, optional connect address)
- menu-bar: add command wrapper and a "Pair Android Device" section in Settings
https://claude.ai/code/session_01QFVnqBsSN9GoNBed9rAzMV
* [orbit] Add QR code pairing for Android devices over Wi-Fi
Support Android's "Pair device with QR code" wireless debugging flow:
the app generates a QR code, the device scans it and advertises an mDNS
pairing service, which is discovered via `adb mdns services` to pair and
connect automatically.
- eas-shared: add getAdbMdnsServicesAsync / waitForAdbMdnsServiceAsync helpers
- cli: add `pair-android-device-qr` command that waits for the scan, pairs, and connects
- menu-bar: add QRCode component (qrcode-generator + react-native-svg) and a
QR code mode in the Settings pairing form
https://claude.ai/code/session_01QFVnqBsSN9GoNBed9rAzMV
* [cli] improve paring with code
* [menu-bar] Move PairAndroidDevice to a sepate window
* Polish UI
* Add changelog entry
* Unify pair-android-device commands
---------
Co-authored-by: Claude <[email protected]>
* [cli] Lazy-load command modules to speed up CLI startup
Each menu-bar action spawns a fresh CLI process, and index.ts eagerly
imported every command module up front. Because eas-shared is a barrel
that pulls in the whole device/build toolchain, even a trivial command
like list-devices paid the cost of loading download/launch/GraphQL code.
Defer the eas-shared/GraphQL-heavy command modules behind dynamic
import() inside their actions (matching the existing launch-update
pattern) so an invocation only loads the command actually being run.
Also drop the eager `Env` import from utils.ts in favor of a local
EXPO_MENU_BAR check; utils.ts is loaded on every invocation, and
importing Env from the eas-shared barrel alone dragged in the entire
package, defeating the lazy-loading above.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Claude-Session: https://claude.ai/code/session_01WzvKru3pz1AwhEFBXSnMgo
* [cli] Parallelize list-devices command
* Add changelog entry
---------
Co-authored-by: Claude <[email protected]>
* [menu-bar] Fix flaky onboarding e2e by not querying the closed window
Pressing "Get Started" destroys the Onboarding window the WebDriver
session is focused on, then opens the popover. The old test issued a
bare getWindowHandles() against that destroyed target, which hangs over
the legacy IPC bridge (instead of erroring) and wedges the session until
the 60s Mocha timeout plus a 120s teardown timeout — a timing race that
flaked on CI.
Capture the onboarding handle before the click, then poll until it drops
out of the handle list and switch to a surviving window by id, so we
never query the dead target.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Claude-Session: https://claude.ai/code/session_01J6H74fvJKEwuJ3oapEpx1G
* [menu-bar] Cache CocoaPods and ccache in macOS E2E workflow
The macOS E2E job rebuilds every pod from scratch on each run, which
dominates its wall-clock time. Add two caches to the `macos-e2e` job:
- actions/cache for the CocoaPods install (Pods/ + spec cache), keyed on
Podfile.lock so a dependency bump invalidates it. `pod install` still
runs to reconcile and re-apply post_install hooks.
- hendrikmuhs/ccache-action plus a USE_CCACHE-gated Podfile hook that
routes the pod/app C/C++/ObjC compiler through ccache wrappers, so
unchanged sources (RN core, folly, hermes, fmt, …) are served from
cache instead of recompiling. Gated behind USE_CCACHE so local
`pod install` is unaffected; wrappers fall back to plain clang when
ccache is absent.
Also add the matching CHANGELOG entries.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Claude-Session: https://claude.ai/code/session_01J6H74fvJKEwuJ3oapEpx1G
* [menu-bar] Update changelog PR references to #349
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Claude-Session: https://claude.ai/code/session_01J6H74fvJKEwuJ3oapEpx1G
* [menu-bar] Revert ccache from macOS E2E workflow, keep CocoaPods cache
ccache didn't yield usable hit rates for the Xcode build, so drop it and
keep only the CocoaPods cache for now. Removes the ccache-action step,
the USE_CCACHE-gated Podfile hook, the CCACHE_* env, and the clang
wrapper scripts.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Claude-Session: https://claude.ai/code/session_01J6H74fvJKEwuJ3oapEpx1G
---------
Co-authored-by: Claude <[email protected]>
The error thrown in launchUpdateOnIOSAsync was missing the word "devices", so it read "Launching updates on iOS physical is not supported yet". The matching guard in launchUpdateOnExpoGoIosAsync already reads "...iOS physical devices is not supported yet". Align the two messages.
Co-authored-by: Patrick Wehbe <[email protected]>
* [orbit] Install apps on physical iPhones from macOS, Windows & Linux
Enable installing already-signed builds onto a USB-connected iPhone from
all three desktop platforms, without requiring a paid Apple account or Xcode.
The usbmux-based install transport already existed and is cross-platform; this
un-gates it from macOS/Xcode and surfaces it everywhere:
- eas-shared: make the developer disk image mount and app launch best-effort in
runOnDevice (installation works without Xcode; launch is skipped on
Windows/Linux). Discover devices via Promise.allSettled so a missing usbmux
service surfaces a friendly, actionable error instead of crashing. Add a
cross-platform usbmuxd availability check and per-OS helper-software guidance.
- cli: guard simulator-only paths behind process.platform === 'darwin' so
list-devices and install-and-launch work for physical iPhones off macOS; add
an install-apple-device-support command (winget on Windows, pkexec usbmuxd on
Linux).
- menu-bar: add a "Show iOS devices" preference (on by default everywhere),
list physical devices independently of simulators, and show an
install-helper-software prompt when the Apple device service isn't reachable.
- common-types: add APPLE_DEVICE_USBMUXD_NOT_RUNNING / APPLE_DEVICE_NOT_PAIRED
error codes and helper guidance on the device-list error.
https://claude.ai/code/session_01TeUZg8NkPKBJN2RRQfAh3o
* Adjust comments
* [menu-bar] Consolidate iOS gating under a single showIosSimulators preference
Drop the separate showIosDevices preference. The single showIosSimulators
toggle now controls the whole iOS section (simulators on macOS, physical
iPhones everywhere) and defaults ON / is enabled on all platforms so connected
iPhones are discoverable on Windows and Linux.
https://claude.ai/code/session_01TeUZg8NkPKBJN2RRQfAh3o
* [eas-shared] Surface APPLE_DEVICE_USBMUXD_NOT_RUNNING when devicectl returns no devices
getConnectedDevicesAsync swallowed the usbmux "service not running" error on
Linux/Windows: the custom path already wraps the socket error as an
InternalError, but the guard only matched raw socket codes, and the native
devicectl path fulfils with [] off macOS, so the error was logged and dropped.
Re-throw user-facing InternalErrors from the custom path before the other
guards, and add a regression test.
https://claude.ai/code/session_01TeUZg8NkPKBJN2RRQfAh3o
* [eas-shared] Distinguish usbmuxd not-installed vs not-running on Linux
On Linux usbmuxd is socket/udev-activated and only runs while an Apple device is
attached, so an installed-but-stopped daemon previously produced the misleading
"install the usbmuxd daemon" guidance. Detect the installed binary and, when
present, tell the user to connect/trust the device or run
`sudo systemctl start usbmuxd` instead. Adds a `startCommand` to the helper
guidance, hides the menu-bar "Install" button when there's nothing to install,
updates the README, and covers both states with unit tests.
https://claude.ai/code/session_01TeUZg8NkPKBJN2RRQfAh3o
* [eas-shared] Fix extracting tar on Windows
* [orbit] Auto-install lighter Apple Mobile Device Support on Windows
Instead of opening a download page and asking the user to manually install
Apple's software, install the official, lightweight Apple Mobile Device Support
package (USB driver + device service only, no full iTunes/Apple Devices app) via
winget directly from the in-app prompt. Vendoring Apple's binaries isn't viable
(proprietary, signed kernel driver), so this uses Apple's official distribution.
- CLI install-apple-device-support: use winget id Apple.AppleMobileDeviceSupport
with --silent; surface a clear error when winget is missing so the UI can fall
back to the manual download page.
- Menu-bar prompt: run the install first (winget/pkexec) and only open the
download page as a fallback; refresh the device list after a successful install.
- Update guidance text, README, and unit tests.
https://claude.ai/code/session_01TeUZg8NkPKBJN2RRQfAh3o
* adjust manual launch message
* [orbit] Prompt to set up iOS device support during onboarding
On Windows/Linux first run, offer an opt-in to install the helper software
needed to install apps on a physical iPhone over USB, instead of leaving users
to discover it later.
- check-tools CLI reports a cross-platform `appleDevice` entry (usbmux
availability via validateAppleDeviceRequirementsAsync) carrying the same
install guidance the popover prompt uses.
- Onboarding shows an "Install apps on a physical iPhone?" section (Windows/Linux
only, when the tooling is missing) reusing AppleDeviceHelperPrompt; after a
successful install it re-runs the checks so the prompt clears.
https://claude.ai/code/session_01TeUZg8NkPKBJN2RRQfAh3o
* Fix opening IPAs on linux
* Adjust onboarding
* Install Apple Mobile Device Support during onboarding
* [menu-bar][electron] Fix Alert custom onPress
* adjust usbmux message
* Make iOS install less noisy on Windows
* Add changelog entry
---------
Co-authored-by: Claude <[email protected]>
* Add cross-platform E2E test scaffolding
WebdriverIO-based E2E setup that runs against both the Electron build
(via Chromedriver) and the macOS native build (via Appium mac2). Shared
spec files drive both platforms; a sanitizing testID -> AX-id mapping
in helpers.ts lets the same selectors find elements through either
driver. Adds the testID props the specs depend on across the popover,
device-list, and onboarding components.
Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
* Bundle wdio-electron-service into the Electron build
@electron-forge/plugin-vite ships only the .vite/build/ output into the
asar, no node_modules. The wdio IPC-bridge hook needs to be reachable
at runtime, so:
- Vite traces a dynamic import('./wdio-hook') in main.ts and a static
import in preload.ts. The shim files re-export 'wdio-electron-service/
{main,preload}', so Vite pulls those into the bundle/chunk graph
instead of leaving them as runtime requires that would fail to
resolve.
- A define-driven `process.env.WDIO_E2E` gate keeps the main-process
hook dead-coded out of production builds; the preload-side bridge is
always exposed (it's a no-op without the main-side handler).
- packagerConfig.ignore strips wdio-electron-service from the asar in
production for tidiness.
Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
* Gate macOS native sources behind EXPO_ORBIT_E2E
Three startup paths trigger TCC permission prompts (Local Network,
Screen Recording) or modal sheets that CI runners can't dismiss
non-interactively, blocking the desktop for the whole test session:
- SwifterWrapper's HTTP server binds to non-loopback ports.
- SPUStandardUpdaterController fails on unsigned bundles and pops the
"Unable to Check For Updates" sheet.
- RSClient.configure spins up reachability monitoring that trips the
local-network prompt.
Each is now skipped when EXPO_ORBIT_E2E=1 (set via appium:environment
in wdio.macos.ts). Also gates the CLI codesign step in archive_cli.sh
on CODE_SIGNING_ALLOWED so the unsigned E2E build doesn't fail at the
Xcode 'Archive and sign CLI' script phase.
Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
* Add E2E CI workflows (Windows, Linux, macOS)
GitHub Actions workflow that runs the E2E specs on all three platforms.
Each job builds the platform's binary, creates an Android AVD via
sdkmanager/avdmanager (not booted — orbit-cli uses `emulator
-list-avds` which reads from disk, so the device appears in Orbit
without nested-virt setup), runs wdio, and uploads screenshots +
AX-tree dumps as artifacts (if: always()).
Linux uses xvfb-run + a handful of apt-get'd runtime libs that
ubuntu-latest lacks. macOS pins Xcode 26.2 and builds unsigned so the
runner doesn't need a Developer ID cert.
Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <[email protected]>
* [eas-shared] Add support for DeviceHub.app on Xcode 27
Xcode 27 replaced Simulator.app with DeviceHub.app (bundle id
com.apple.dt.Devices, located at Xcode.app/Contents/Applications),
which broke opening/activating simulators from Orbit:
- 'id of app "Simulator"' no longer resolves, so system requirements
validation failed with TOOL_CHECK_FAILED
- 'open -a Simulator' and the AppleScript calls targeting the
"Simulator" app/process no longer worked
Resolve the simulator UI app dynamically (preferring DeviceHub.app when
the selected Xcode is >= 27, Simulator.app otherwise) and use the
resolved name/bundle id everywhere instead of hardcoding Simulator.
* [menu-bar] Fix building the macOS app with Xcode 27
Three build failures when building with Xcode 27:
1. Xcode 27 only supports macOS deployment targets >= 12.0, but several
pods still declare 10.x/11.x. Bump them to 12.0 in post_install.
2. Xcode 27's clang rejects fmt 11.0.2's consteval compile-time
format-string checking ('call to consteval function ... is not a
constant expression' in format-inl.h). fmt 11.0.2 ignores externally
defined FMT_USE_CONSTEVAL/FMT_CONSTEVAL (no #ifndef guard), so patch
the vendored header in post_install to disable consteval. Format
strings are validated at runtime instead of compile time.
3. ReactNativeDelegate.customize(_ rootView: RCTRootView) fails with
'method does not override any method from its superclass' because
Xcode 27 resolves RCTRootView to the 'React' module while the
Expo.swiftmodule expects it in 'React_RCTAppDelegate'. Replace the
override by making the root view layer transparent directly in
applicationDidFinishLaunching.
* Add changelog entry
* [react-native] Customize root view background color to clear
* Fix fmt pod issue
* Revert AppDelegate changes
* Apply suggestion from @gabrieldonadel
* Improve window focusing
---------
Co-authored-by: Gabriel Donadel <[email protected]>
* [menu-bar]: invert noAudio flag to audio for boot device configuration
* [cli]: rename noAudio parameter to audio to align with intended boolean logic
* [eas-shared]: rename noAudio option to audio to align with emulator flag logic
Switch the CurrentUser fragment from UserActor.profilePhoto (gravatar
only) to primaryAccountProfileImageUrl, so the avatar in the Settings
window prefers the user's uploaded account profile image when one is
set.
Also drop the client-side gravatar default-image fallback in Avatar.tsx:
the www resolver for primaryAccountProfileImageUrl already returns a
gravatar URL with d=<default-image> baked in via getUserAvatarUrl, so
the client copy was redundant. The unused name prop is dropped along
with it.
Co-authored-by: Tomek Agent <[email protected]>