The AcceptanceTestSuite emits voluminous INFO-level logs for every test case, making CI output noisy and failures harder to diagnose.

Approach

Buffer all java.util.logging output during each individual test and replay it only when that test fails.

Changes

• CapturingLogHandler (new inner static class) — a JUL Handler that accumulates LogRecords in a synchronized ArrayList; drainRecords() atomically snapshots and clears the buffer.
• AcceptanceTest.run() — wraps each test execution with log capture:
  • Saves and sets all root-logger handler levels to OFF before the test runs.
  • Installs CapturingLogHandler on the root logger.
  • On failure: restores handler levels and replays the full buffered record set through them.
  • On success: silently discards the buffer.

No new dependencies — uses only java.util.logging APIs.

// Before test: silence existing handlers, start capturing
for (int i = 0; i < existingHandlers.length; i++) {
    originalLevels[i] = existingHandlers[i].getLevel();
    existingHandlers[i].setLevel(Level.OFF);
}
rootLogger.addHandler(capturingHandler);

// ... test runs ...

// In finally: restore, then conditionally replay
for (int i = 0; i < existingHandlers.length; i++) {
    existingHandlers[i].setLevel(originalLevels[i]);
}
if (testFailed) {
    for (LogRecord record : capturingHandler.drainRecords()) {
        for (Handler handler : existingHandlers) {
            handler.publish(record);
        }
    }
}