Add control line tests (DTR/DSR and RTS/CTS).

Author: MrDOS

src/test/java/gnu/io/SerialPortControlTest.java

@@ -0,0 +1,169 @@
+package gnu.io;
+
+import static org.junit.jupiter.api.Assertions.assertFalse;
+import static org.junit.jupiter.api.Assertions.assertTrue;
+
+import java.util.TooManyListenersException;
+import java.util.concurrent.CountDownLatch;
+import java.util.concurrent.TimeUnit;
+import java.util.logging.Logger;
+
+import org.junit.jupiter.api.Test;
+import org.junit.jupiter.api.extension.RegisterExtension;
+
+/**
+ * Test the ability of the {@link SerialPort} implementation to manage the
+ * control lines of the port (DTR/DSR and RTS/CTS).
+ * <p>
+ * This test is nonspecific as to <em>which</em> implementation; it exercises
+ * only the public interface of the Java Communications API. Ports are opened
+ * by the {@link SerialPortExtension} test extension, presumably by way of
+ * {@link CommPortIdentifier}.
+ */
+public class SerialPortControlTest
+{
+	private static final Logger log = Logger.getLogger(SerialPortControlTest.class.getName());
+
+	private static final String BUGGY_DSR_EVENTS = "DSR event propagation is buggy! DSR events are useless until this is fixed.";
+	private static final String BUGGY_CTS_EVENTS = "CTS event propagation is buggy! CTS events are useless until this is fixed.";
+
+	/**
+	 * How long to wait (in milliseconds) for asynchronous events to arrive
+	 * before failing the test.
+	 */
+	private static final long EVENT_TIMEOUT = 250;
+
+	@RegisterExtension
+	SerialPortExtension ports = new SerialPortExtension();
+
+	/**
+	 * Test that toggling DTR on a port changes the DSR state of another.
+	 */
+	@Test
+	void testDTRDSRPolling()
+	{
+		this.ports.a.setDTR(false);
+		assertFalse(this.ports.a.isDTR());
+		assertFalse(this.ports.b.isDSR());
+
+		this.ports.a.setDTR(true);
+		assertTrue(this.ports.a.isDTR());
+		assertTrue(this.ports.b.isDSR());
+
+		this.ports.a.setDTR(false);
+		assertFalse(this.ports.a.isDTR());
+		assertFalse(this.ports.b.isDSR());
+	}
+
+	/**
+	 * Test that toggling RTS on a port changes the CTS state of another.
+	 */
+	@Test
+	void testRTSCTSPolling()
+	{
+		this.ports.a.setRTS(false);
+		assertFalse(this.ports.a.isRTS());
+		assertFalse(this.ports.b.isCTS());
+
+		this.ports.a.setRTS(true);
+		assertTrue(this.ports.a.isRTS());
+		assertTrue(this.ports.b.isCTS());
+
+		this.ports.a.setRTS(false);
+		assertFalse(this.ports.a.isRTS());
+		assertFalse(this.ports.b.isCTS());
+	}
+
+	/* Use of CountDownLatch in the asynchronous event tests is based on
+	 * https://stackoverflow.com/a/1829949/640170. */
+
+	/**
+	 * Test that toggling DTR on a port changes generates a DSR event on
+	 * another.
+	 *
+	 * @throws TooManyListenersException if the port has not been properly
+	 *                                   cleaned up after previous tests
+	 * @throws InterruptedException      if the test is interrupted while waiting
+	 *                                   for the event
+	 */
+	@Test
+	void testDTRDSREvents() throws TooManyListenersException, InterruptedException
+	{
+		CountDownLatch latch = new CountDownLatch(1);
+		this.ports.b.addEventListener(ev -> {
+			if (ev.getEventType() == SerialPortEvent.DSR)
+			{
+				latch.countDown();
+			}
+		});
+		this.ports.b.notifyOnDSR(true);
+		this.ports.a.setDTR(true);
+		boolean sawEvent = false;
+		sawEvent = latch.await(SerialPortControlTest.EVENT_TIMEOUT, TimeUnit.MILLISECONDS);
+		if (!sawEvent)
+		{
+			/* FIXME: The hard part about adding tests is that sometimes you
+			 * find bugs. The DSR event _is_ generated: set a breakpoint in the
+			 * event callback, or a print statement, and it will reliably fire
+			 * – but only _after_ this await. The callback _is_ run from the
+			 * monitor thread, as you'd expect it to be, and without the await,
+			 * the listener is called within milliseconds of DTR being
+			 * asserted; but block the main thread (with this await, or with
+			 * `Thread.sleep()`), and the event will never happen.
+			 *
+			 * I'm mystified by this behaviour, because the callback
+			 * configuration here is the same as in `testRTSCTSEvents()`, where
+			 * it works as expected.
+			 *
+			 * One thing which does seem to cause the event to propagate is
+			 * de-asserting DTR. Then _that_ event gets lost, of course, but it
+			 * seems to prompt the first event to make its way through. So
+			 * that's the workaround employed here for now; but this really is
+			 * papering over a hole. The DSR event being one edge late means
+			 * it's basically impossible for any consumer to hand-roll their
+			 * own hardware flow control.
+			 *
+			 * This is broken on Windows (10), macOS (10.15), and Linux
+			 * (4.19.0). */
+			log.warning(SerialPortControlTest.BUGGY_DSR_EVENTS);
+			this.ports.a.setDTR(false);
+			sawEvent = latch.await(SerialPortControlTest.EVENT_TIMEOUT, TimeUnit.MILLISECONDS);
+		}
+		assertTrue(sawEvent);
+	}
+
+	/**
+	 * Test that toggling RTS on a port changes generates a CTS event on
+	 * another.
+	 *
+	 * @throws TooManyListenersException if the port has not been properly
+	 *                                   cleaned up after previous tests
+	 * @throws InterruptedException      if the test is interrupted while waiting
+	 *                                   for the event
+	 */
+	@Test
+	void testRTSCTSEvents() throws TooManyListenersException, InterruptedException
+	{
+		CountDownLatch latch = new CountDownLatch(1);
+		this.ports.b.addEventListener(ev -> {
+			if (ev.getEventType() == SerialPortEvent.CTS)
+			{
+				latch.countDown();
+			}
+		});
+		this.ports.b.notifyOnCTS(true);
+		this.ports.a.setRTS(true);
+		boolean sawEvent = false;
+		sawEvent = latch.await(SerialPortControlTest.EVENT_TIMEOUT, TimeUnit.MILLISECONDS);
+		if (!sawEvent)
+		{
+			/* FIXME: Same story here as with DSR events. This works correctly
+			 * on Windows (10), but fails on macOS (10.15) and Linux
+			 * (4.19.0). */
+			log.warning(SerialPortControlTest.BUGGY_CTS_EVENTS);
+			this.ports.a.setRTS(false);
+			sawEvent = latch.await(SerialPortControlTest.EVENT_TIMEOUT, TimeUnit.MILLISECONDS);
+		}
+		assertTrue(sawEvent);
+	}
+}