Reporting

Every Stove test failure ships with structured execution context. The reporter captures activity from registered systems: HTTP calls, DB ops, Kafka observations, WireMock stubs, and gRPC interactions. When an assertion fails, you get the stack trace plus the sequence of system activity that led to it.

In 30 seconds Add the Kotest or JUnit extension. Failures print a timeline by default (pretty console). Add reporting { } in Stove().with to tune output or use machine-readable JSON. Reporting pairs with Tracing and the Dashboard; it does not require either one.

Setup

KotestJUnit

dependencies {
  testImplementation("com.trendyol:stove-extensions-kotest")
}

class StoveConfig : AbstractProjectConfig() {
  override val extensions: List<Extension> = listOf(StoveKotestExtension())
  // ...
}

dependencies {
  testImplementation("com.trendyol:stove-extensions-junit")
}

@ExtendWith(StoveJUnitExtension::class)
abstract class BaseE2ETest { /* ... */ }

The extension registers an AfterTestListener that intercepts failures and prints the report.

Configure

Default is on, prints to console, and dumps only on failure. Local runs include the complete pretty report. On CI, Stove automatically uses bounded compact output for each failure report.

Stove().with {
    reporting {
        enabled()
        dumpOnFailure()
    }
    // ... your systems
}.run()

The CI-aware default recognizes common flags such as CI, GITHUB_ACTIONS, GITLAB_CI, JENKINS_URL, TF_BUILD, and BUILDKITE. Values that are blank, false, or 0 do not enable compact mode.

Compact output keeps the complete pass/fail/total counts, then:

• keeps failed entries plus the most recent timeline context, up to 50 entries;
• keeps the most recent 10 items in every rendered collection, including timeline details and snapshots;
• keeps up to 20 entries from each rendered map and up to 10 system snapshots;
• keeps 2,000 characters from each large value, split between its beginning and end;
• stops nested diagnostic traversal after eight levels and detects cyclic values;
• caps the final composed console report at 50,000 characters, including an execution trace appended by a test extension;
• shortens Kafka's inline "Messages so far" assertion dump before the test framework prints it;
• prints an explicit omission notice for every shortened section.

Tune the limits, force compact output everywhere, or force the complete console report:

reporting {
    failureRenderer(
        PrettyConsoleRenderer.compact(
            ConsoleReportLimits(
                maxTimelineEntries = 25,
                maxCollectionItems = 5,
                maxMapEntries = 10,
                maxSnapshots = 5,
                maxValueCharacters = 1_000,
                maxNestingDepth = 6,
                maxOutputCharacters = 25_000
            )
        )
    )

    // Force complete output, including on CI:
    // failureRenderer(PrettyConsoleRenderer)

    // Restore the default explicitly:
    // failureRenderer(PrettyConsoleRenderer.ciAware())
}

JsonReportRenderer is not shortened. Use it when the complete structured report should be saved as a CI artifact while console output stays compact.

Renderers

PrettyConsoleRenderer (default)

Human-friendly. Color, alignment, system snapshots inline, with compact detail on CI. Built with Mordant.

─── Stove Report ────────────
▶ http    POST /orders   201
▶ kafka   shouldBePublished
        topic=order.created.v1
        timeout=10s   (timed out)
─────────────────────────────

JsonReportRenderer

Machine-readable. Pipe into CI artifacts, MCP, or your own analyzer.

{
  "test": "OrderE2ETest.create",
  "entries": [
    { "kind": "http", "method": "POST",
      "path": "/orders", "status": 201 },
    { "kind": "kafka", "op": "shouldBePublished",
      "topic": "order.created.v1",
      "status": "timeout" }
  ]
}

What gets reported

Surface	Captured
HTTP	method, path, status, latency, request/response bodies
Kafka	producer publishes, consumer reads, topic, partition, offset, payload
Databases (SQL + NoSQL)	queries, bind args, rows affected, durations
WireMock	stub matches and misses, request body
gRPC	method, request, response, status
System snapshots	per-system state at failure time (Kafka topics, WireMock unmatched, etc.)

Snapshots make root-cause analysis faster. A WireMock snapshot, for example, can show that an "unexpected 404" was the app hitting an unmocked path.

Pairs well with

• Tracing. Reporter plus OTel = call chain inside your app, not just the test view.

• Dashboard. Same data, browseable in a local web UI; persists across sessions.

• MCP. Agents fetch the same evidence in token-efficient slices.

• When a test fails. The full failure flow as a scroll story.

Troubleshooting

Symptom	Check
No report on failure	Extension registered? StoveKotestExtension() in extensions (Kotest) or @ExtendWith(StoveJUnitExtension::class) (JUnit)
Report missing system entries	System registered before the runner block in Stove().with { }
Empty Kafka snapshot	Interceptor bean registered? See Kafka pitfalls
Need the complete console report on CI	Set failureRenderer(PrettyConsoleRenderer)
Need smaller local output too	Set failureRenderer(PrettyConsoleRenderer.compact())
JSON empty in CI	Use JsonReportRenderer; pipe System.out to a file or use dashboard JSON export