Test framework: Use JBoss Marshalling cloner

[dmlloyd]

Remove usage of xstream or serialization for cloning, and use the JBoss Marshalling cloner instead.

Fixes #15892.

[gsmet]

Oh, that's some nice cleanup!

[geoand]

Love it!

I wish I had thought of JBoss Marshaling when we added this

[geoand]

Also, I'm pretty sure this will close more outstanding issues

[gsmet]

@dmlloyd I was able to reproduce the problem locally with mvn clean install -Dstart-containers -Dtest-containers in the IT submodule.

In:

@QuarkusTest
public class TransactionalUniAsserterTest {

    private static final AtomicBoolean ASSERTER_USED = new AtomicBoolean();

    @RunOnVertxContext
    @Test
    public void testTransactionalUniAsserter(TransactionalUniAsserter asserter) {
        assertNotNull(asserter);
        asserter.assertThat(Panache::currentTransaction, transaction -> {
            ASSERTER_USED.set(true);
            assertNotNull(transaction);
            assertFalse(transaction.isMarkedForRollback());
            asserter.putData("tx", transaction);
        });
        asserter.assertThat(Panache::currentTransaction, transaction -> {
            assertNotNull(transaction);
            assertFalse(transaction.isMarkedForRollback());
            assertNotEquals(transaction, asserter.getData("tx"));
        });
    }

    @AfterAll
    static void afterAll() {
        assertTrue(ASSERTER_USED.get());
    }

}

we end up with ASSERTER_USED being false.

[dmlloyd]

It looks like it's having a problem because the TransactionalUniAsserter is getting cloned, and the clone, well, just doesn't work right. I'm digging into the how/why now but it might be a tricky one to solve.

[quarkus-bot]

Status for workflow Quarkus CI

This is the status report for running Quarkus CI on commit 0e113de.

✅ The latest workflow run for the pull request has completed successfully.

It should be safe to merge provided you have a look at the other checks in the summary.

You can consult the Develocity build scans.

---

Flaky tests - Develocity

⚙️ Native Tests - HTTP

📦 integration-tests/rest-client

✖ io.quarkus.it.rest.client.selfsigned.ExternalSelfSignedITCase.should_accept_self_signed_certs_java_url - History

• Read timed out - java.net.SocketTimeoutException

java.net.SocketTimeoutException: Read timed out
	at java.base/sun.nio.ch.NioSocketImpl.timedRead(NioSocketImpl.java:288)
	at java.base/sun.nio.ch.NioSocketImpl.implRead(NioSocketImpl.java:314)
	at java.base/sun.nio.ch.NioSocketImpl.read(NioSocketImpl.java:355)
	at java.base/sun.nio.ch.NioSocketImpl$1.read(NioSocketImpl.java:808)
	at java.base/java.net.Socket$SocketInputStream.read(Socket.java:966)
	at org.apache.http.impl.io.AbstractSessionInputBuffer.fillBuffer(AbstractSessionInputBuffer.java:161)
	at org.apache.http.impl.io.SocketInputBuffer.fillBuffer(SocketInputBuffer.java:82)

[holly-cummins]

Ohhhh, nice. The xstream serialization was becoming increasingly non-viable, since it doesn't work on Java 17+, and that's our minimum Java level. This possibly means I don't need to pursue my epic test classloading rewrite (#34681). I was going to pick that work up tomorrow, too (although in fairness, it's been Real Soon Now for like a year).

Here's my (partial) list of issues that I think this might resolve:

• can't get field offset on a record class using quarkus-junit5 #40148
• Deserialization Errors Caused by Avro Classes being in the Wrong ClassLoader for Tests #36812
• Better error message for @ParameterizedTest failures in a @QuarkusTest #38991

Thinking about it more, I think "load the tests with the classloader we intend to use to run them" is probably still a nice aspiration, and trying to avoid serialization could have some broader benefits. I think the following issues probably are not fixed by this change, and would need the epic test classloading rewrite (ETCR):

• Templated test contexts do not see transformed test classes (classloader/ordering issues in tests) #27821
• Quarkus CDI-managed beans not available in Pact state callbacks #22611
• Quarkus 3 - consumer tests cannot directly access pact MockServer in test methods in dev mode quarkiverse/quarkus-pact#73

[maxandersen]

Xstream goes poof? That's excellent news.

[dmlloyd]

I think that cleaning up the class loader situation is still a worthy goal.

[holly-cummins]

I'm planning to add a test which reproduces the failure @edeandrea and I saw, so we can assess how common it is. If it's a common one, we should maybe consider temporarily reverting this while we wait for the JUnit upgrade. But in the meantime, some tests which were failing because of the broken xstream serialization are now passing (yay!). So I've raised a PR to un-disable them: #40749