Skip to content

Fail loudly when preparing the emulator for a test fails - #37

Closed
ofalvai wants to merge 1 commit into
masterfrom
push-yulnvzmsuklr
Closed

ofalvai wants to merge 1 commit into
masterfrom
push-yulnvzmsuklr

Conversation

@ofalvai

@ofalvai ofalvai commented Aug 18, 2026 •

Copy link
Copy Markdown
Collaborator

Why

The beforeEach hook in the test builder swallowed every error from prepareEmulatorForTest. A simulator or emulator in a bad state didn't fail there. It showed up later as an unrelated assertion failure somewhere else in the scenario, which sent debugging in the wrong direction.

What changed

  • The hook now logs and rethrows, so a broken emulator fails the test that hit it.
  • On iOS, endRunningApplication checks whether the app is running before calling simctl terminate, and skips terminate when it isn't.

Why iOS needs the running check

Rethrowing alone would break every scenario. simctl terminate exits non-zero when the app isn't running, and the app is never running before a scenario's first test.

Terminate's own failure can't tell "app not running" apart from "simulator broken". On Xcode 26.6 / iOS 26.5 it prints found nothing to terminate with the same exit code whether the app was never installed or is just idle. simctl appinfo, get_app_container and listapps don't help either. They only report whether the app is installed, and their output doesn't change while it runs.

So the check asks launchd instead. simctl spawn booted launchctl list shows a running app as UIKitApplication:<bundle id> with a live pid. Any failure from terminate itself still propagates.

Why Android has no equivalent

am force-stop exits 0 whether or not the app is installed or running (checked against an emulator), and pm clear already has an existence check. A guard there would only add an adb round-trip per test.

Verification

Ran end to end against a live simulator and emulator for three cases: installed but idle, never installed, and running.

🤖 Generated with Claude Code

@ofalvai
ofalvai force-pushed the push-yulnvzmsuklr branch from f68e996 to 44370bb Compare August 24, 2026 16:19
The beforeEach hook in testBuilder swallowed every error from prepareEmulatorForTest, so an emulator in a bad state surfaced later as an unrelated assertion failure somewhere else in the scenario. It now logs and rethrows.

Rethrowing on its own would have broken every scenario, because iOS "simctl terminate" exits non-zero when the app is not running, and the app is not running before the first test of a scenario. Verified on Xcode 26.6 / iOS 26.5: terminate reports "found nothing to terminate" identically whether the app was never installed or is merely idle, so the error text and the exit code cannot tell a missing app apart from a broken simulator. simctl appinfo, get_app_container and listapps do not help either, as they only report whether the app is installed, and their output is unchanged by the app running or not.

So IOSEmulatorManager.endRunningApplication now establishes the running state up front with "simctl spawn booted launchctl list", where a running app appears as UIKitApplication:<bundle id> with a live pid, and skips terminate when the app is not running. Everything else propagates.

Android needs no equivalent guard: "am force-stop" exits 0 whether or not the app is installed or running (verified against an emulator), and "pm clear" is already guarded by an existence check. Adding one there would only add an adb round-trip per test.

Verified end to end against a live simulator and emulator for the installed-but-idle, never-installed and actually-running cases.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Copilot AI balanced review requested due to automatic review settings November 10, 2026 20:23
@ofalvai ofalvai closed this Nov 10, 2026

Copilot AI left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🟡 Changes recommended

The async change breaks the declared Q.Promise contract on iOS, causing callers that use supported methods such as .done() or .fail() to throw TypeErrors.

3 open issues
What changed in this PR

This PR improves emulator-preparation diagnostics in the CodePush test framework and adds an iOS running-state check before terminating an app.

Changes:

  • Logs preparation failures in the test hook and rethrows them.
  • Uses the simulator’s launchd job list to decide whether to call simctl terminate.
File Description
code-push-plugin-testing-framework/​script/​testBuilder.js Adds logging and rethrowing for preparation failures.
code-push-plugin-testing-framework/​script/​platform.js Adds an iOS running-state guard before app termination.

Review effort: Balanced

Comment on lines +409 to 416
IOSEmulatorManager.prototype.endRunningApplication = async function (appId) {
const isAppRunning = await isAppRunningOnIOS(appId);
if (!isAppRunning) {
return null;
}
await testUtil_1.TestUtil.getProcessOutput("xcrun simctl terminate booted " + appId, undefined);
return null;
};
if (!isAppRunning) {
return null;
}
await testUtil_1.TestUtil.getProcessOutput("xcrun simctl terminate booted " + appId, undefined);
*/
async function isAppRunningOnIOS(appId) {
const output = await testUtil_1.TestUtil.getProcessOutput("xcrun simctl spawn booted launchctl list", { noLogCommand: true, noLogStdOut: true, noLogStdErr: true });
return output.includes(appId);
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants