KMP Test

This documentation was generated with the assistance of AI. Please report any inaccuracies.

Run a Kotlin Multiplatform (Android + iOS) project’s tests on each target it actually has — the shared module’s Android host (JVM) tests, jvmTest, the Android app’s testDebugUnitTest, the Kotlin/Native iOS simulator tests and the SwiftUI host app’s XCTest bundle — scoped to whatever the caller asked for, and report a compact per-target summary parsed from JUnit XML (Gradle) and .xcresult (Xcode). This is a report-only skill: it never fixes failing tests or modifies source/test code.

Purpose

iru-kmp-test discovers the KMP modules, Android app modules and iosApp/ host project at runtime, lists the real test task names from ./gradlew <module>:tasks --all (instead of assuming one plugin’s naming), and runs the matching stages. Android/JVM stages run on any host; iOS stages run only on macOS with a full Xcode and an iOS Simulator runtime and are otherwise reported "unverified", never "passed". Every stage that boots a simulator is serialized repository-wide behind the shared atomic-mkdir .integration-test.lock directory (owner file, stale after 60 minutes, released on every exit path). It also detects the "selector matched nothing" cases: Gradle fails the JVM task with "No tests found for given includes", whereas a Kotlin/Native simulator task exits 0 and writes no XML, and an unmatched -only-testing: yields zero tests. It is the KMP equivalent of Android Test and Swift Test.

A ui selector (/iru-kmp-test ui [<scope>], or a ui: yes line) runs the UI tests per the catalog’s UI contract (see Building user interfaces): Compose Multiplatform runComposeUiTest tests on the Android host (Robolectric) and the iOS simulator, ui: native Robolectric Compose tests in androidApp, iosApp ImageRenderer/XCUITest captures, and device flow tests only when an emulator is attached. Step 5a resolves UI_SCREENSHOTS_DIR (default <repo root>/.ui-screenshots, forwarded to the JVM tests and to the simulator through SIMCTL_CHILD_/TEST_RUNNER_ variables), clears it only on a full (unscoped) UI run, forces every Gradle test task to re-run with --rerun (captures are side effects that Gradle’s up-to-date checks and build cache do not track), collects .xcresult attachments and device outputs into the contract’s <kind>/<Id>/<state>--<direction>--<theme>[--<variant>…​] layout, validates the names with an embedded parser, and reports the .ui-quality.yml accessibility level with violation counts. KMP captures carry an android/ios platform variant so the same <Id> does not collide.

Purpose group

Kotlin Multiplatform Development Tooling

Plugin

Shipped by the iru-kmp plugin of the ai-catalog marketplace (install with /plugin install iru-kmp@ai-catalog; see Plugins).

Invocation

/iru-kmp-test
/iru-kmp-test <scope>
/iru-kmp-test <scope>
target: android|ios|all
module: <gradle path>
destination: <xcodebuild destination>
lock: held

Inputs

Argument Required Description Default

<scope>

No

A test class (GreeterTest), class+method (GreeterTest#greets), wildcard/package pattern, a test file path (its source set picks the targets: commonTest runs every target, androidHostTest/androidUnitTest/jvmTest run Android/JVM only, iosTest/appleTest/nativeTest run iOS only, iosApp/…​ runs the XCTest stage), a module (:shared, androidApp) or a source-set name.

None — every test task of every in-scope target.

target

No

android runs the JVM stages only; ios runs the iOS stages only (Kotlin/Native simulator tests and the iosApp XCTest bundle); all runs both.

all

module

No

Restrict the run to one Gradle module (:shared, :androidApp, :composeApp).

All modules

destination

No

An xcodebuild destination overriding the simulator picked for the XCTest stage.

First available iPhone simulator

ui

No

yes, or a leading ui word in <scope> (for example ui GreetingScreenTest): a UI run (Step 5a). With no further scope it is a full UI run that clears the capture directory first.

Not set — ordinary tests only.

lock

No

held when the caller already holds .integration-test.lock; the skill then neither acquires nor releases it.

Not held — the skill acquires it for simulator stages.

Stage-to-task mapping (Step 2):

Stage Task (first that exists)

Android host tests of a KMP module

testAndroidHostTest (or any test<Name>HostTest), else testDebugUnitTest

JVM target

jvmTest, only when a jvm() target exists

Android app module

testDebugUnitTest

Kotlin/Native iOS

iosSimulatorArm64Test on Apple silicon, iosX64Test on Intel

Outputs

  • A plain-text summary in the conversation — never a written file.

  • One line per target (Android host, JVM, Android app, iOS simulator, iosApp XCTest): passed (counts), failed (counts, each failing Class#test, assertion message, first stack line and the XML/.xcresult path), no matching tests for <pattern>, timed out, not applicable, or unverified (reason).

  • An explicit callout when a test fails on one target only, which usually means an actual implementation differs.

  • The lock is mentioned only when it was anything other than an immediate acquire.

  • For a UI run: the capture directory and whether it was cleared (full run) or kept (scoped run); screenshots per kind and per platform (android/ios); malformed names; accessibility violations (total + rule ids with counts) and the .ui-quality.yml level applied (enforce/report/off/none); device flow tests run or unverified (reason).

  • No source or test code is ever modified; raw Gradle/xcodebuild output is never pasted.

Execution flow

flowchart TD A["Start /iru-kmp-test\n(optional scope + key: value lines)"] --> B["Step 0: resolve scope,\ntarget, module, destination, lock"] B --> C["Step 1: detect KMP modules,\nAndroid app, iosApp;\ncheck JDK, Android SDK, gradlew"] C --> D{"KMP module found?"} D -- no --> Z["Stop: not a KMP project\n(point to iru-android-test / iru-java-test)"] D -- yes --> E["Step 2: list test tasks per module\n(./gradlew tasks --all)"] E --> F["Step 3: run Android/JVM stages\n(--tests repeated per task, --continue)"] F --> G{"iOS stages in scope and\nmacOS + Xcode + runtime?"} G -- no --> H["Mark iOS stages unverified\n(state what is missing)"] G -- yes --> I["Step 4: acquire\n.integration-test.lock\n(wait / break stale / time out)"] I --> J["Run Kotlin/Native\niosSimulatorArm64Test"] J --> K["Step 5: xcodebuild test on iosApp\n(XcodeGen if needed)\nparse .xcresult"] K --> L["Release lock"] L --> L2["Step 5a (ui runs): collect .xcresult attachments\nand device outputs; validate capture names;\ncount accessibility violations"] H --> M["Step 6: parse JUnit XML\nper task directory"] L2 --> M M --> N{"Build failed before\nproducing results?"} N -- yes --> O["Report the compile or\nconfiguration error"] N -- no --> P["Step 7: report per-target\npass / fail / no matching tests /\ntimed out / unverified"] O --> P

Dependencies

Invokes

None — iru-kmp-test is a leaf skill; it only runs ./gradlew, xcodebuild, xcrun, xcodegen and small inline python3 parsers, and it does not call any other skill in this catalog.

Invoked by

  • KMP Code One Task Group — Step 3.3 (scoped Android/JVM tests, target: android), Step 3.5 (full Android/JVM suite) and Step 3.7 (iOS stage, target: ios, behind the shared lock) and Step 3.8 (UI stage, with the ui selector), always through the iru-gate-runner agent.

  • KMP Generate All Tests — Step 7, to run the full suite once and confirm the generated tests cause no regressions, and Step 5a, with the ui selector, to run the newly generated UI tests.

Related agents

  • iru-gate-runner — both callers always run this skill from inside an iru-gate-runner sub-agent they spawn, not inline, so the compact summary is all that reaches their context.

Source

SKILL.md on GitHub — the file this page was generated from.