Repository navigation
Relaunch apps via open as console user - #43842
Conversation
Ensure GUI apps are relaunched in the logged-in user's session instead of root. Update relaunch_application to skip when no non-root GUI user is present (empty/"root"/"loginwindow"), and use `sudo -u "$console_user" open -b <bundle_id>` when the installer runs as root so the app appears in the user's Dock/GUI. Remove the previous osascript activation attempt and add status handling and explanatory comments.
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #43842 +/- ##
==========================================
- Coverage 66.92% 66.92% -0.01%
==========================================
Files 2603 2603
Lines 209044 209073 +29
Branches 9235 9235
==========================================
+ Hits 139905 139913 +8
- Misses 56392 56413 +21
Partials 12747 12747
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
There was a problem hiding this comment.
Claude Code Review
This repository is configured for manual code reviews. Comment @claude review to trigger a review and subscribe this PR to future pushes, or @claude review once for a one-time review.
Tip: disable this comment in your organization's Code Review settings.
WalkthroughThe embedded shell functions in the Homebrew ingester were updated. The 🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@ee/maintained-apps/ingesters/homebrew/scripts.go`:
- Around line 649-656: Replace the direct sudo -u "$console_user" open -b
"$bundle_id" call with the same launchctl asuser pattern used in send_signal so
the open runs inside the console user's Mach/GU I namespace; specifically, when
EUID is 0 invoke /bin/launchctl asuser "$console_uid" sudo -u "$console_user"
open -b "$bundle_id" (preserving the >/dev/null 2>&1 || open_status=$? handling
and the open_status variable), otherwise keep the non-root open path unchanged;
reference the existing send_signal use of /bin/launchctl asuser to mirror its
context-bootstrapping approach.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
Run ID: ab8a3fe6-3f00-475a-adfd-ed470012d327
📒 Files selected for processing (1)
ee/maintained-apps/ingesters/homebrew/scripts.go
There was a problem hiding this comment.
Pull request overview
Updates the Homebrew install-script generator so previously-running GUI apps are relaunched in the logged-in user’s session (rather than root), improving Dock/GUI visibility and reliability when installers run as root.
Changes:
- Add explanatory comments describing why relaunch must happen in the console user’s GUI session.
- Update
relaunch_applicationto skip relaunch when no non-root GUI user is present (empty/root/loginwindow). - Replace
osascript ... activaterelaunching withopen -b, usingsudo -u "$console_user"when running as root and capturing exit status.
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
There was a problem hiding this comment.
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
ee/maintained-apps/ingesters/homebrew/scripts.go (1)
663-667:⚠️ Potential issue | 🟡 Minor
open -bexit status is not a strong signal of successful GUI relaunch.
/usr/bin/opencan return 0 even whenLSOpenURLsWithRole()ultimately fails to surface the app in the Dock (e.g., stale Launch Services cache, bundle not yet re-registered after the move/install), and can return non-zero transiently right after a fresh install when the new bundle hasn't been indexed. Consider either (a) pollingpgrep -f "$bundle_id"for a few seconds afteropenas a liveness check (mirroring the quit loop), or (b) softening the failure message to something like "relaunch requested" to avoid misleading log output. Not blocking — just flagging that the success/failure log may not always reflect reality.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@ee/maintained-apps/ingesters/homebrew/scripts.go` around lines 663 - 667, The current check uses open_status from the `open -b` call to declare success/failure for relaunching `bundle_id`, which can be unreliable; update the relaunch logic in the block that inspects `open_status` to either (a) poll for the process after calling `open` by running `pgrep -f "$bundle_id"` in a short loop for a few seconds (mirror the quit loop behavior) and treat presence as the true success signal, or (b) change the echoed failure message around `bundle_id` to a non-definitive string like "relaunch requested" so we don't claim a successful GUI relaunch based solely on `open_status`; modify the code around the `open_status` check and echo statements to implement one of these approaches.
🧹 Nitpick comments (1)
ee/maintained-apps/ingesters/homebrew/scripts.go (1)
654-661: Guard against emptyconsole_uidbefore invokinglaunchctl asuser.If
id -u "$console_user"fails (e.g., directory service hiccup, user removed between checks),console_uidwill be empty and the call becomes/bin/launchctl asuser "" sudo -u "$console_user" open -b "$bundle_id", which will emit a confusinglaunchctlerror rather than the intended "Failed to relaunch" message. The upstream non-empty check onconsole_userdoesn't coveridfailure.🛡️ Suggested hardening
if [[ $EUID -eq 0 ]]; then local console_uid console_uid=$(id -u "$console_user") - /bin/launchctl asuser "$console_uid" sudo -u "$console_user" open -b "$bundle_id" >/dev/null 2>&1 || open_status=$? + if [[ -z "$console_uid" ]]; then + echo "Could not resolve UID for console user '$console_user'; skipping relaunch." + return + fi + /bin/launchctl asuser "$console_uid" sudo -u "$console_user" open -b "$bundle_id" >/dev/null 2>&1 || open_status=$? else open -b "$bundle_id" >/dev/null 2>&1 || open_status=$? fi🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@ee/maintained-apps/ingesters/homebrew/scripts.go` around lines 654 - 661, The call to `/bin/launchctl asuser` can receive an empty `console_uid` if `id -u "$console_user"` fails; update the block that computes `console_uid` and invokes `/bin/launchctl asuser` to validate `console_uid` is non-empty (and preferably numeric) before using it. If `console_uid` is empty/invalid, avoid calling `launchctl asuser` and instead set `open_status` to a failure value (or fall back to the non-root `open -b "$bundle_id"` path) and emit the existing "Failed to relaunch" handling so a clear error is produced; refer to the `console_uid`, `console_user`, and `/bin/launchctl asuser` usage in this section when making the change.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Outside diff comments:
In `@ee/maintained-apps/ingesters/homebrew/scripts.go`:
- Around line 663-667: The current check uses open_status from the `open -b`
call to declare success/failure for relaunching `bundle_id`, which can be
unreliable; update the relaunch logic in the block that inspects `open_status`
to either (a) poll for the process after calling `open` by running `pgrep -f
"$bundle_id"` in a short loop for a few seconds (mirror the quit loop behavior)
and treat presence as the true success signal, or (b) change the echoed failure
message around `bundle_id` to a non-definitive string like "relaunch requested"
so we don't claim a successful GUI relaunch based solely on `open_status`;
modify the code around the `open_status` check and echo statements to implement
one of these approaches.
---
Nitpick comments:
In `@ee/maintained-apps/ingesters/homebrew/scripts.go`:
- Around line 654-661: The call to `/bin/launchctl asuser` can receive an empty
`console_uid` if `id -u "$console_user"` fails; update the block that computes
`console_uid` and invokes `/bin/launchctl asuser` to validate `console_uid` is
non-empty (and preferably numeric) before using it. If `console_uid` is
empty/invalid, avoid calling `launchctl asuser` and instead set `open_status` to
a failure value (or fall back to the non-root `open -b "$bundle_id"` path) and
emit the existing "Failed to relaunch" handling so a clear error is produced;
refer to the `console_uid`, `console_user`, and `/bin/launchctl asuser` usage in
this section when making the change.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
Run ID: 8ce7520b-d65c-47bf-90af-484e3a00b355
📒 Files selected for processing (1)
ee/maintained-apps/ingesters/homebrew/scripts.go
…pers (#48639) The custom install script embedded pre-#42951/#43842 copies of quit_and_track_application and relaunch_application. The stale 'if ! osascript' check treats any non-erroring osascript call as "app is running" (osascript exits 0 whether it prints true or false), so the app was marked for relaunch on every install and launched after every patch, even from a fully-quit state. Replace both functions with the current scripts.go constants (output-based running check, launchctl-asuser relaunch, updated console-user guards) and regenerate the darwin manifest.
…pers (#49030) **Related issue:** Resolves #48639 # Checklist for submitter If some of the following don't apply, delete the relevant line. - [ ] Input data is properly validated, `SELECT *` is avoided, SQL injection is prevented (using placeholders for values in statements), JS inline code is prevented especially for url redirects, and untrusted data interpolated into shell scripts/commands is validated against shell metacharacters. - [x] Timeouts are implemented and retries are limited to avoid infinite loops ## Testing - [x] QA'd all new/changed functionality manually ## Details The GitHub Desktop FMA uses a custom install script (`ee/maintained-apps/inputs/homebrew/scripts/github-desktop-install.sh`) that embeds its own copies of `quit_and_track_application` and `relaunch_application`. Those copies were frozen before two fixes landed in the generated helpers in `ee/maintained-apps/ingesters/homebrew/scripts.go`: - #42951 — check osascript **output** instead of exit status. `osascript -e '... is running'` exits 0 whether it prints `true` or `false`, so the stale `if ! osascript ...` guard never fired. The app was marked `APP_WAS_RUNNING=1` on **every** install with a GUI user logged in and relaunched after every patch — even from a fully-quit state. This is the root cause of #48639. - #43842 — relaunch via `launchctl asuser ... open -b` as the console user instead of `osascript ... to activate` (which is unreliable from a root context), plus the updated empty/root/loginwindow console-user guards. This PR replaces both embedded functions with the current scripts.go constants (verified byte-for-byte identical) and regenerates `ee/maintained-apps/outputs/github/darwin.json` via `go run ./cmd/maintained-apps -slug github`. The manifest diff is script-ref-only (`98ab6ed8` → `c91ea2b5`); version and uninstall script are unchanged. The other five custom scripts (Docker Desktop, OpenVPN Connect, Webex, Max, Pd) already carry the updated helpers — GitHub Desktop was the only one missed. ## Manual QA Tested the updated `quit_and_track_application` / `relaunch_application` functions on macOS against GitHub Desktop itself (`com.github.GitHubClient`): - **Fully quit (the bug scenario):** verified `is running` returns `false` and zero `GitHub Desktop.app` processes. Fixed functions set `APP_WAS_RUNNING=0` and the app stays closed. Running the old shipped check (`if ! osascript ...`) against the same state misclassifies the app as running (osascript exits 0 with output `false`) and would have relaunched it. - **Running:** quit succeeds, `APP_WAS_RUNNING=1`, app relaunches successfully afterward. - `bash -n` passes on the updated script. Note: hosts where the FMA was already added keep the baked `98ab6ed8` script until their instance refreshes the manifest. The by-design behavior "app running with dock icon but no visible window → relaunched with a window" is unchanged; window-aware relaunching would be a separate enhancement. <!-- This is an auto-generated comment: release notes by coderabbit.ai --> ## Summary by CodeRabbit * **Bug Fixes** * Improved GitHub Desktop installation behavior on macOS so the app is more reliably closed and reopened after install. * Better handles login/session edge cases, helping ensure the app relaunches in the correct user’s desktop session. * Reduces failed or missed relaunches when the installer is run with elevated permissions. <!-- end of auto-generated comment: release notes by coderabbit.ai -->
**Related issue:** Part of #54170 # Checklist for submitter - [x] Changes file added for user-visible changes in `changes/`, `orbit/changes/` or `ee/fleetd-chrome/changes`. See [Changes files](https://github.com/fleetdm/fleet/blob/main/docs/Contributing/guides/committing-changes.md#changes-files) for more information. - [x] Input data is properly validated, `SELECT *` is avoided, SQL injection is prevented (using placeholders for values in statements), JS inline code is prevented especially for url redirects, and untrusted data interpolated into shell scripts/commands is validated against shell metacharacters. - [x] Timeouts are implemented and retries are limited to avoid infinite loops ## Testing - [x] Added/updated automated tests - [ ] QA'd all new/changed functionality manually `TestInstallScriptStopsWhenAppWontQuit` runs generated install scripts (app copy and pkg) in bash with `osascript`, `sudo` and friends stubbed. It covers an app that isn't running, one that quits, one that accepts the quit event but keeps running, and one whose quit is canceled. The last two must exit 1 without copying or installing anything. The test fails against `main` and passes with this change. I also ran each edited custom install script end to end with stubs and a refused quit. All of them exit 1 before replacing the app, except Harmony SASE (see below). Docker Desktop's helper was checked under its `set -euo pipefail` for both outcomes. Not run on a real Mac. ## AI **AI:** Claude Code (claude-opus-5-5) ## Summary When a macOS app won't quit (for example, the user cancels a prompt to save changes), the install script moves it aside anyway, copies the new version in, and exits 0. fleetd then deletes `$TMPDIR`, which takes the moved-aside bundle the app is still running from. Fleet reports the update as installed while the user is still on the old version (#54170). Two problems in `quit_and_track_application` caused this: - It confirmed the quit with `pgrep -f "$bundle_id"`, which never matches because a macOS app's command line is its executable path. So it reported success as soon as the quit event was accepted. - When the app was still running at the timeout, it only logged "did not quit", and the caller carried on. Changes: - The helper now polls `osascript … is running` after each quit attempt and returns 1 at the timeout. This matches the KiCad FMA (#54613). An `is running` error doesn't count as quit. - Generated install scripts call `quit_and_track_application '<id>' || exit 1` before replacing the app, for both app-copy and pkg installs. - Custom install scripts that copy the helper get the same fix. Callers stop the install, and Pd detaches its disk image first. Docker Desktop already stopped on failure (it runs under `set -e`) and only needed the `pgrep` fix. **Harmony SASE** keeps going: its pkg force-quits the app on its own, so stopping would block an update that works today. - The `new-fma` skill now says how custom install scripts should quit the app and that helper copies must be kept in sync. Behavior change: a forced patch whose app won't quit now fails, and it's retried on the next policy run instead of being reported as installed. The yaml-files and REST API docs describe when the install is *attempted* ("whether or not the app is open"), which is still accurate, so they're unchanged. Outputs aren't regenerated here: like #43842 and #42951, the FMA bot picks up the new script refs. The 18 frozen macOS apps keep their current scripts until they're unfrozen. Follow-ups (not in this PR): - 12 custom install scripts have their own `quit_application` helper with the same `pgrep` check and no stop on failure: 1Password, Adobe Creative Cloud, Box Tools, ExpressVPN, Google Chrome, Grammarly, Gyazo, Logi Tune, Microsoft Edge, P4V, Slack, Zoom. Each has its own flow, so they need a per-script look. - The uninstall helper `quit_application` has the same `pgrep` check. It's left unchanged here to avoid churning every uninstall script ref. - #53570 changes the `var_name` line of the same helper, in a separate hunk. The Wispr Flow custom script in #54708 copies the old helper and needs this fix after whichever merges second. <!-- This is an auto-generated comment: release notes by coderabbit.ai --> ## Summary by CodeRabbit * **Bug Fixes** * On macOS, updates now stop if the app being replaced doesn’t quit within the timeout. The running app’s files are preserved, and the update is reported as failed rather than installed. <!-- end of auto-generated comment: release notes by coderabbit.ai -->
…53570) **Related issue:** NA — found while reviewing #53566 # Checklist for submitter - [x] Changes file added for user-visible changes in `changes/`. - [x] Input data is properly validated, `SELECT *` is avoided, SQL injection is prevented (using placeholders for values in statements), JS inline code is prevented especially for url redirects, and untrusted data interpolated into shell scripts/commands is validated against shell metacharacters. ## What & why `quit_and_track_application` and `relaunch_application` derive the name of the "was it running" flag from the bundle ID and interpolate it into an `eval`: ```sh local var_name="APP_WAS_RUNNING_$(echo "$bundle_id" | tr '.-' '__')" eval "export $var_name=1" ``` `tr '.-' '__'` maps only dots and dashes. Any other character survives into the variable name, and a space breaks the `eval` in two: - **Write:** `eval "export APP_WAS_RUNNING_org_mozilla_pale moon=1"` — `export` takes two arguments, so it sets `APP_WAS_RUNNING_org_mozilla_pale` and a stray `moon=1`. The intended variable is never set. - **Read:** `eval "was_running=$APP_WAS_RUNNING_org_mozilla_pale moon"` — assigns empty, then tries to run `moon` as a command, printing `moon: command not found` into the install output. - **Result:** `was_running` is empty, so `relaunch_application` returns early. **Pale Moon hits this today.** `pale-moon.json` has `"unique_identifier": "org.mozilla.pale moon"`, and the shipped script calls `quit_and_track_application 'org.mozilla.pale moon'`. The app is quit during an update and never reopened. It is the only one of the 967 shipping macOS FMAs affected — every other identifier is drawn from `[A-Za-z0-9._-]`. This replaces the derivation with bash parameter expansion that maps everything outside `[A-Za-z0-9_]`: ```sh local var_name="APP_WAS_RUNNING_${bundle_id//[^[:alnum:]_]/_}" ``` Also applied to the thirteen custom install scripts under `inputs/homebrew/scripts/` that carry their own copy of these functions, so the two don't drift. ### Reviewer notes - **Inert for every app but pale-moon.** For identifiers drawn from `[A-Za-z0-9._-]` the new mapping produces a byte-identical variable name to `tr '.-' '__'`. I compared old vs. new across all 967 homebrew `unique_identifier` values: 966 identical, 1 changed (pale-moon, the broken one). Runtime behavior for every other app is unchanged. - **Why parameter expansion instead of `tr -c '[:alnum:]_' '_'`.** Same result, but no subshell, and it sidesteps two hazards: `echo` appends a newline that `tr -c` would map to a trailing underscore, and BSD vs. GNU `tr` differences in CI. Verified on `/bin/bash` 3.2, the macOS runtime — generated scripts have used a bash shebang since #44836. - **No output regeneration here.** This changes a shared function body, so all 956 generated scripts containing it will shift by two lines when the nightly regenerates. That matches how prior changes to these same functions landed (#43842, #42951, #44836 were all generator-only). - **Not a security fix.** `unique_identifier` comes from Fleet's own input files, not the brew API, so this isn't the untrusted-input class #51324 addressed. Hardening the `eval` is still worthwhile because the name reaches a root-privileged `eval`, but the bug being fixed is the relaunch failure. ## Testing - [x] Added/updated automated tests `TestQuitRelaunchVarNameSurvivesEval` runs the real write/read round trip through bash for four shipping identifier shapes including `org.mozilla.pale moon`, and asserts the flag survives. `TestQuitRelaunchVarNameNeutralizesCommandSubstitution` asserts the derived name is a bare shell identifier and that a `$(touch ...)` payload doesn't execute. `TestQuitAndRelaunchDeriveVarNameIdentically` pins the two derivations to the same line so they can't drift, and `TestCustomInstallScriptsDoNotUseOldVarNameDerivation` fails if any custom install script still uses `tr '.-' '__'`. Confirmed these fail against the old derivation: under `set -u`, pale-moon errors with `APP_WAS_RUNNING_org_mozilla_pale: unbound variable`; without it, the read back is silently empty. - [x] QA'd all new/changed functionality manually `go test ./ee/maintained-apps/...` passes. `bash -n` clean on all thirteen patched custom scripts. `go vet` and `gofmt` clean. (`make lint-go-incremental` couldn't run locally — it needs a network clone of golangci-lint — so leaving that to CI.) ## AI **AI:** Claude Code (claude-opus-5) <!-- This is an auto-generated comment: release notes by coderabbit.ai --> ## Summary by CodeRabbit * **Bug Fixes** * Fixed macOS app updates so Fleet-maintained apps with spaces or other special characters in their bundle identifiers can relaunch successfully. * Improved tracking of whether applications were running before installation for identifiers containing special characters. * **Tests** * Added coverage for application tracking with special-character identifiers. <!-- end of auto-generated comment: release notes by coderabbit.ai -->
…one is logged in (#56457) **Related issue:** Combines #56016, #56314, #56333, #56336, #56337, #56453 and #56454 into one PR. Those PRs are closed in favor of this one. Follow-up to #55157 (#54170) and #54210. #56333 came up in review of #56085 (Logitech G HUB). ## Summary All seven change how macOS FMA scripts are generated or tested. Fleet runs these scripts as root at any time, including at the login window. ### Quit apps by bundle ID when the cask lists names (#56016) A few casks list app names in `uninstall quit:` instead of bundle IDs. `tell application id "DeskTime"` fails with -1728, so `quit_application` treated the app as not running and the script removed the bundle while the app was open. A dotless `quit:` value now becomes the app's `unique_identifier`, and repeated IDs in a stanza are quit once. A dotless value with `*` (such as `*WaveLink*`) is still a pattern. Affected: DeskTime, DisplayLink, NotesOllama and Fathom. ### Stop the uninstall when the app won't quit (#56314) The uninstall `quit_application` helper had the bug that #55157 fixed in the install helper. It confirmed the quit with `pgrep -f "$bundle_id"`, which never matches, and only logged "did not quit" before `sudo rm -rf` removed the bundle. - `quit_application` now polls `osascript … is running` after a 1s sleep, and returns 1 at the 10s timeout. `quit_matching_applications` returns 1 if any match didn't quit. - Generated scripts add `|| exit 1` only when the quit ID names the app being uninstalled: its `unique_identifier`, case-insensitively, or a wildcard that matches it. Helpers, shared updaters such as Microsoft AutoUpdate, and other vendors' apps keep Homebrew's warn-and-continue. A helper that launchd relaunches would otherwise block the uninstall for good. - With #56016, a name in `quit:` becomes the app's own ID first, so DeskTime, DisplayLink, NotesOllama and Fathom stop too. - The six custom uninstall scripts with stale copies of the helper now carry the current one: Box Tools, CleanMyMac, ForkLift, GPG Suite, Microsoft Word and P4V. **Behavior change:** if the app won't quit, the uninstall fails with "Application '<id>' did not quit." It no longer removes the app while the app is running. ### Quit the app when the cask doesn't (#56337) Generated uninstall scripts only quit the IDs in the cask's `uninstall quit:`. Most casks list none, and Excel, Outlook, PowerPoint and Teams quit only Microsoft AutoUpdate. So the app could be removed while it was still running. - The generator now writes `quit_application '<unique_identifier>' || exit 1` once, before anything removes the bundle. With an uninstall stanza, it goes at the stanza's quit step: after `launchctl` and the cask's own quits, and before `signal`, `early_script`, `script`, `pkgutil` and `delete`. Without one, it goes right before the app's `rm -rf`. A zap quit doesn't count, because zap runs after the app is removed. - It's left out when an uninstall stanza already quits or signals the app. 17 casks signal apps that don't quit cleanly, such as DBeaver, Parallels, Webex and Zoom for IT Admins. A quit that stops the script mustn't run ahead of the signal. Perimeter 81 splits its stanza across three artifacts, so every stanza is checked first. - A cask quit that lists the app's name counts as quitting the app, through #56016's mapping. So DeskTime, DisplayLink, Fathom and NotesOllama aren't quit twice. - Vendor `script` and `early_script` directives aren't exempt. A directive doesn't say whether it stops the app, and the install script already quits the same ID with `|| exit 1` on every update. - Some apps' `unique_identifier` differs from the cask's quit ID, such as KeePassXC, CleanShot X, Tower, JASP and Sequel Ace. They now also quit the `unique_identifier`, which is the ID Fleet sees installed. If LaunchServices doesn't know that ID, the helper returns 0. The new-fma skill's guidance for custom uninstall scripts moved to #56459. ### Trash cleanup with no GUI user (#56333) At the login window, `LOGGED_IN_USER` is `loginwindow`, so `trash()` tried to move every path into `/Users/loginwindow/.Trash` and failed. When the user was empty, the unquoted variable disappeared and the arguments shifted. Either way, system paths like `/Users/Shared/LGHUB` were left behind. - Generated calls quote the user: `trash "$LOGGED_IN_USER" '<path>'`. - With no real console user (empty, `root` or `loginwindow`), `~` paths are skipped with `Not logged into a non-root GUI; skipping <path>.` Other paths are deleted with `rm -rf`, and globs still expand through `compgen -G`. These are only the cask's app-specific `zap` entries. None of the absolute paths in the outputs is a shared parent folder. - Cleanup stays best-effort. A failed removal is logged, and `trash()` always returns 0. Before, the exit status depended on which path happened to be last. - Nine custom uninstall scripts carry the current `trash()` verbatim, with quoted call sites: Camtasia, CleanMyMac, Cycling '74 Max, ForkLift, GPG Suite, Homebrew, KiCad, Microsoft Word and Teleport. ### Script directives with no GUI user (#56336) Cask `script` and `early_script` directives were generated as `(cd /Users/$LOGGED_IN_USER && …)`. At the login window the `cd` failed, so vendor uninstallers never ran (Anka, Cloudflare WARP, Chrome Remote Desktop Host, Microsoft Defender, Santa, Mullvad VPN and 11 more). - Executable directives run as root, so they now start in the user's home when there is one and in `/` otherwise: `(cd "/Users/$LOGGED_IN_USER" 2>/dev/null || cd /; sudo '<exe>' …)`. `must_succeed: false` still appends `|| true`. - String directives run as the user, so they're skipped with the same "Not logged into a non-root GUI" message when there's no real user. Box Drive is the only output with one. - Box Drive's `pre_uninstall_scripts` run their per-user steps only when someone is logged in. With an empty user, the old sudoers line wasn't valid syntax. GPG Suite's custom uninstaller call uses the same `cd … || cd /` form. ### Quit 1Password with SIGTERM in the uninstall (#56453) With "Keep 1Password in the menu bar" on (the default), 1Password's Electron `before-quit` handler cancels a quit event and drops to the menu bar. Its `SIGTERM` handler quits it fully, the same as Quit in its menu bar menu. Without a custom script, the bot would regenerate 1Password's uninstall with `quit_application 'com.1password.1password' || exit 1`, and every uninstall with 1Password running would fail with "did not quit." - `1password_uninstall.sh` is this branch's generated uninstall with one change: `quit_application` is replaced by `quit_1password`. It sends `pkill -TERM -x 1Password`, polls `is running` for up to 10s, and returns 1 if 1Password is still running. The call is `quit_1password || exit 1`, right before `rm -rf`. - `quit_1password` is the install script's `quit_and_track_1password` (#56452) without the relaunch tracking. - The script was built from this branch's generator, so it already has the #56333 `trash()` and quoted call sites. ### Check every `quit_*` call in custom install scripts (#56454) `TestCustomInstallScriptsStopWhenAppWontQuit` only matched `quit_application` and `quit_and_track_application`. It now matches any `quit_*` call, so a script with its own quit function, like `quit_and_track_1password` (#56452), can't drop `|| exit 1` unnoticed. `qgis-ltr-install.sh` is allowlisted: its `quit_app_at_path` sends SIGTERM, then SIGKILL after 10 seconds. ## Merge order - #56459 (new-fma skill guidance, split out of #56337) should merge after this PR. - #55865 (`rmdir` with `~`) also changes `scripts.go` and many of the same outputs. Whichever merges second needs a rebuild of the overlapping outputs. ## Outputs 975 of 980 macOS outputs change. In each one, only the uninstall script and its ref change. Version, URL, sha256, install script and queries are byte-identical to `main`. - **Generated outputs:** they have the #56333 and #56336 changes, applied to the uninstall script that's on `main` today. The 116 outputs the bot bumped since #56333 was cut were rebuilt from `main`'s file the same way. LM Studio and Logitech G HUB were added after #56333 was cut, so they're new in this PR. The edit reproduces all 955 generated outputs in #56333 and #56336 byte for byte. - **#56314's quit change isn't applied to generated outputs.** That follows #56314 and the precedent of #55157, #43842 and #42951: the FMA bot picks up the new uninstall refs for the apps that quit something. For 35 of those apps, the bot also adds #56337's quit, for example Excel, Outlook, PowerPoint and Teams. - **#56337's quit is applied to 589 outputs.** None of them quits anything yet, and each already matched this branch's generator. So each gains only `quit_application '<unique_identifier>' || exit 1` and the helper. Eight are frozen apps, which the bot never rewrites: BrickLink Studio, Cisco Jabber, Comet, Mobirise, Twobird, Usage, Visual Paradigm and Whispering. Angry IP Scanner, InVesalius and KeyboardCleanTool weren't changed by the trash or directive edits, so they're new in this PR. - **Frozen outputs (14):** hand-edited with the trash and directive changes, since ingest never rewrites them. Logitech G HUB is newly frozen since #56333. As in #56314, frozen apps keep the old quit helper until they're unfrozen. Avast Secure Browser and Intune Company Portal also don't get #56337's quit until then. - **Custom-script outputs:** each uninstall script equals its input file. Microsoft Word was rebuilt on `main`'s newer version. 1Password's uninstall now comes from its new custom script (#56453). `go run ./cmd/maintained-apps --slug=1password/darwin` matches the committed output, and its install script is unchanged from `main`. - **DeskTime, DisplayLink, Fathom, NotesOllama:** #56016's regenerated quit line, plus the trash change. Generator cross-check: I ran the combined ingester for 20 apps, including the four from #56016, LM Studio, Logitech G HUB, Box Drive, Cloudflare WARP, Docker Desktop, the custom-script apps and a bot-bumped app. Every uninstall script is identical to this PR's, except for #56314's quit helpers and `|| exit 1` on the app's own quit, which the bot adds. #56337 cross-check: I generated all 964 non-custom macOS uninstall scripts from today's cask API with this branch's generator, before and after #56337. 624 change, each by one `quit_application '<unique_identifier>' || exit 1` line, plus the helper in the 587 that didn't have it. The next statement is always a removal: `sudo rm -rf` (553), `remove_pkg_files` (60) or the cask's uninstall script (8). DeskTime, DisplayLink, Fathom and NotesOllama don't change. For Asana, Google Chrome, Microsoft Defender and Obsidian, `go run ./cmd/maintained-apps` produces the same uninstall script as the committed output. # Checklist for submitter - [ ] Changes file added for user-visible changes in `changes/`, `orbit/changes/` or `ee/fleetd-chrome/changes`. See [Changes files](https://github.com/fleetdm/fleet/blob/main/docs/Contributing/guides/committing-changes.md#changes-files) for more information. - None. The combined PRs' changes files were removed. - [x] Input data is properly validated, `SELECT *` is avoided, SQL injection is prevented (using placeholders for values in statements), JS inline code is prevented especially for url redirects, and untrusted data interpolated into shell scripts/commands is validated against shell metacharacters. - Cask-supplied paths, executables, args and string scripts still go through `shellSingleQuote`. The only new interpolation is the quoted `"$LOGGED_IN_USER"`. - [x] Timeouts are implemented and retries are limited to avoid infinite loops ## Testing - [x] Added/updated automated tests - `TestUninstallScriptStopsWhenAppWontQuit`, `TestQuitTargetsApp`, `TestCustomUninstallScriptsUseCurrentQuitHelper` (#56314) - `TestUninstallScriptQuitsAppByBundleIDWhenCaskListsNames` (#56016) - `TestUninstallScriptTrashWithoutGUIUser`, `TestCustomUninstallScriptsUseCurrentTrash` (#56333) - `TestUninstallScriptScriptDirectiveWithoutGUIUser` (#56336) - `TestCustomInstallScriptsStopWhenAppWontQuit` checks every `quit_*` call (#56454) - `TestUninstallScriptQuitsAppCaskDoesNotName`, `TestUninstallScriptQuitsAppOnce` (#56337), and a quit-by-name case in `TestUninstallScriptQuitsAppByBundleIDWhenCaskListsNames` that fails if the app is quit twice - 1Password's uninstall (#56453): I ran it with `/bin/bash` 3.2 and `osascript`, `pkill`, `sudo`, `launchctl`, `stat`, `scutil`, `mv` and `rm` stubbed. The `osascript` stub ignores quit events, as 1Password does in menu-bar mode. Not running: no `pkill`, the app is removed, exit 0. Quits on SIGTERM: one `pkill -TERM -x 1Password`, the app is removed, exit 0. Ignores SIGTERM: "did not quit", exit 1, nothing is removed. shellcheck reports the same warnings as the generated script. The SIGTERM behavior was checked on a real Mac for #56452. - `go test ./ee/maintained-apps/... ./cmd/maintained-apps/...` passes with Homebrew bash 5.3 and with `/bin/bash` 3.2 first on `PATH`. golangci-lint (full and incremental) reports 0 issues on those packages. Two incremental findings in #56314's tests now use `errors.AsType` and `require.Contains`, like the other tests. A gosec G602 false positive in #56337's order check is restructured. - [ ] QA'd all new/changed functionality manually ## AI **AI:** Claude Code (claude-opus-5-5) --------- Co-authored-by: Claude <noreply@anthropic.com>
This pull request updates the application quit and relaunch logic in the Homebrew ingester scripts to more robustly detect valid GUI user sessions and improve how applications are relaunched after installation. The main improvements ensure that actions are only attempted when a real user is logged in, and that relaunching applications works reliably in the correct user context, especially when running as root.
User session validation:
quit_application,quit_and_track_application, andrelaunch_applicationhave been expanded to skip actions if the console user is empty,root, orloginwindow, preventing attempts to interact with the GUI when no real user is logged in. [1] [2] [3]Application relaunch improvements:
relaunch_applicationlogic now useslaunchctl asuserwithsudo -uto launch the application in the correct user's GUI session, ensuring that the app appears in the user's Dock and GUI, even when the script runs as root. This replaces the previous approach of usingosascript, which could fail in root contexts. [1] [2]Summary by CodeRabbit