test: skip map instrumentation tests when no Maps API key is available

[kikoso]

Problem

run-instrumentation-test fails on every Dependabot pull request, and on every fork PR, for a reason unrelated to the change under test. Dependabot-triggered and fork-triggered runs cannot read secrets.ACTIONS_API_KEY, so the "Inject Maps API Key" step takes its fallback branch and writes MAPS_API_KEY=YOUR_API_KEY. hasValidApiKey is then false and every map-dependent test hits:

check(hasValidApiKey) { "Maps API key not specified" }

which throws IllegalStateException and fails the build.

On #1015 that accounted for 29 of the 41 tests. The 30th failure, StreetViewTests, has no guard at all: it just waits for a panorama that can never load and dies with ComposeTimeoutException. 19 of the last 21 instrumentation runs on dependabot/* branches failed this way, including one whose only change was a GitHub Action version bump.

Change

A test that cannot run for lack of a credential is a skipped test, not a failing one. This replaces the check with a JUnit assumption behind a shared helper in TestUtils.kt, and adds the missing guard to StreetViewTests.

fun assumeValidApiKey() {
    assumeTrue("Maps API key not specified", hasValidApiKey)
}

Nothing changes when a key is present.

Two things reviewers should know

Three tests were passing for the wrong reason. testMarkerStateCannotBeReused and its two siblings declare @Test(expected = IllegalStateException::class). Without a key they were satisfied by the missing-key IllegalStateException itself, so they went green without ever reaching the marker-reuse logic they exist to test. JUnit 4.13.2 propagates assumption failures through the expected-exception check, so they now skip when there is no key and assert for real when there is.

The report still says "failures". AGP writes assumption failures into the connected-test XML as <failure> with skipped="0", so the generated HTML reads "33 failures / 19% successful" even though the task passes. The count is cosmetic, the build result is not. Worth knowing before someone opens the artifact on a green run and panics.

Verification

Run on a local API 30 emulator, since a same-repo branch like this one gets the real secret in CI and therefore will not exercise the no-key path:

Condition	Result
MAPS_API_KEY=YOUR_API_KEY	BUILD SUCCESSFUL, all 33 non-passing results recorded as AssumptionViolatedException
Real key	39 of 41 pass

The two failures with a real key, GoogleMapViewTests.testStartingCameraPosition and MapInColumnTests.testScrollColumn_MapCameraRemainsSame, fail identically on main on the same emulator. They are pre-existing and out of scope here.

./gradlew lint and :maps-app:testDebugUnitTest both pass.

Follow-up worth considering

This makes CI honest, but it does mean dependency bumps get a green check with the map tests skipped rather than real emulator coverage. If you want coverage back on those PRs, add an API-restricted copy of the key as a Dependabot secret (Settings > Secrets and variables > Dependabot). That layers on top of this change rather than replacing it, and it is the only option of the two that gives fork PRs nothing, so it is a judgement call about how much you trust a bot-authored branch with a key.

[github-actions]

Coverage (unit tests)

No unit baseline recorded in coverage/history.csv yet, so this run only reports absolute numbers.

Module	Line %	Change	Branch %	Change
maps-compose	0.00%	new	0.00%	new
maps-compose-utils	2.04%	new	0.49%	new
maps-compose-widgets	0.00%	new	0.00%	new
TOTAL	0.42%	new	0.09%	new

_{Line and branch coverage from unit test reports. History is recorded in coverage/history.csv after each merge to main.}

[googlemaps-bot]

Code Coverage

Overall Project	24.61%	❌

There is no coverage information present for the Files changed