[fix](test) remove JUnit 4 from the fe reactor and unblock the dependency licence review

[morningman]

What problem does this PR solve?

Issue Number: close #xxx

Related PR: #xxx

Problem Summary:

Dependency License Review fails any pull request that adds a JUnit dependency to a pom — in practice, any pull request that adds a Java module with a test. It happened twice on 2026-09-01 alone, on two unrelated branches:

The following dependencies have incompatible licenses:
fe/be-java-extensions/hive-apache-shade/pom.xml » junit:junit@ – License: EPL-1.0
fe/be-java-extensions/jni-spi/pom.xml » org.junit.jupiter:junit-jupiter@ – License: LicenseRef-bad-non-standard
...

This PR fixes the immediate failure and then removes the thing that made a licence exception necessary in the first place.

---

1. The workflow exception (third_party_review.yml)

Neither reported licence is one the project has not approved:

• org.junit.jupiter:junit-jupiter is EPL-2.0, which allow-licenses already carries. GitHub's dependency graph reports it as LicenseRef-bad-non-standard, so the check rejects a licence the project already accepted — a data-quality gap on GitHub's side, not a dependency problem.
• junit:junit is EPL-1.0, an ASF Category B licence, test scope, and no release artifact ships it.

Both are excluded by purl — the same package-specific shape the existing caniuse-lite exception uses — rather than by widening allow-licenses or dropping development from fail-on-scopes. allow-dependencies-licenses excludes a package from the licence check only, so vulnerability reporting for test-scope dependencies is unchanged.

One detail not in the action's README: it matches a purl on type and name and ignores the version (purlsMatch in its src/purl.ts). The version-less entries added here therefore cover every version, and the version pin on the existing caniuse-lite entry has no effect either.

2. Removing JUnit 4 from the fe reactor

junit:junit is the one EPL-1.0 artifact in this build. Nothing declares it — junit-vintage-engine drags it in, and that engine exists only to run the JUnit 4 tests still in the tree. 450 files, 441 of them in fe-core, were what kept it alive. They are JUnit 5 now.

what moved	count
Assert → Assertions	~8600 call sites
message argument moved from first to last	407
@Test(expected = X.class) → assertThrows	94 methods
@Rule ExpectedException → assertThrows + message substring assertion	18 sites, 8 files
@Rule TemporaryFolder → @TempDir Path	4 files
Assume.assumeTrue(msg, cond) → assumeTrue(cond, msg)	12 sites
@Before/@After/@BeforeClass/@AfterClass/@Ignore	all
@RunWith(MockitoJUnitRunner) → MockitoAnnotations.openMocks	1
@FixMethodOrder → @TestMethodOrder	1
junit.framework.AssertionFailedError → java.lang.AssertionError	2 files

Two traps the compiler cannot catch, handled explicitly:

• JUnit 4 puts an assertion message FIRST, JUnit 5 puts it LAST. javac catches the swap for assertTrue/assertFalse/assertNull/assertArrayEquals, but not for assertEquals/assertSame/assertNotEquals when the last argument is itself a String. The 8 three-argument assertEquals calls that are a float delta rather than a message were identified and left alone.
• Assert.assertEquals(Object[], Object[]) compares arrays; the JUnit 5 assertEquals it would become compares references. One such call was converted to assertArrayEquals.

Three things only running the tests found

• JdbcSourceOffsetProviderAsyncSplitTest spelled its teardown @org.junit.After, fully qualified — no import, so no import scan could see it. The jupiter engine does not fail on a JUnit 4 annotation, it ignores it: the teardown stopped running, its MockedStatic never closed, and 27 tests died on static mocking is already registered. This is why the checkstyle pattern below matches anywhere on a line and not just an import.
• StatsCalculatorTest.testFilterOutofRange was annotated @org.junit.Test in a class the jupiter engine already ran — it has never executed. Spelled @Test it runs, and passes.
• CloudAuthTest extends TestWithFeService, whose setup is driven by JUnit 5 annotations the JUnit 4 engine never saw: no cluster was started and the inherited connectContext was always null, which is what every command in the class was handed. Moving the class to JUnit 5 would have activated that setup for the first time, against a class that mocks Env and ConnectContext statically. The vestigial inheritance is dropped instead, with a comment.

3. The gate

A checkstyle rule, because checkstyle runs at the validate phase with includeTestSourceDirectory: a JUnit 4 import fails a plain mvn test locally, rather than waiting for CI. The pattern also covers junit.framework.* (how a JUnit 3 import had survived here) and matches fully qualified references, for the reason above.

fe/be-java-extensions is suppressed for now and the suppression says why: #66729 is rewriting those modules, so migrating them here would only conflict. junit-vintage-engine comes out of fe/pom.xml when that lands — not before, because without it a JUnit 4 test is silently not run rather than failed.

Deliberately out of scope: extension/kettle and samples/ are standalone maven projects, outside this reactor and built by no workflow here, so their JUnit 4 cannot be verified from this build. The pkg:maven/junit/junit exception therefore stays.

Release note

None

Check List (For Author)

• Test
  • Unit Test
  • Manual test (add detailed scripts or steps below)

Workflow change, verified against the real payload rather than by inspection:

1. The dependency-graph compare API for one of the branches that hit this returns package_url pkg:maven/junit/junit and pkg:maven/org.junit.jupiter/junit-jupiter — byte-identical to the entries added here, and non-empty, which is what the action requires to exclude a change at all (groupChanges skips the exclusion list when change.package_url.length === 0).
2. Replaying the action's own filter across all 223 changes in that payload takes the result from the nine incompatible licences the workflow actually reported to zero, with no other dependency newly excluded.
3. The three purls parse under packageurl-js, the library the action validates them with; the folded YAML value parses to exactly those three comma-separated entries, which parseList trims.

Migration:

• mvn test -pl fe-common,fe-core -am over every changed test class — 444 classes, 2564 tests, 1 failure and 2 errors. Those three were separated from pre-existing failures by running the same classes against unmodified master, where PropertyAnalyzerTest, ForwardToMasterTest and FileCacheAdmissionRuleRefresherTest fail identically and are untouched by this change.
• mvn test-compile over the full 76-module reactor with the new checkstyle rule active: green.

The gate, verified from both sides: a probe file carrying org.junit.Test, org.junit.Assert, org.junit.rules.TemporaryFolder and junit.framework.AssertionFailedError is rejected with a message naming the replacement for each and fails the build; the migrated tree passes with 0 checkstyle errors.

• Behavior changed:

  • No.

• Does this need documentation?

  • No.

🤖 Generated with Claude Code

https://claude.ai/code/session_01WVPNcULjD4ieAxdLH1WhDr

[hello-stephen]

Thank you for your contribution to Apache Doris.
Don't know what should be done next? See How to process your PR.

Please clearly describe your PR:

1. What problem was fixed (it's best to include specific error reporting information). How it was fixed.
2. Which behaviors were modified. What was the previous behavior, what is it now, why was it modified, and what possible impacts might there be.
3. What features were added. Why was this function added?
4. Which code was refactored and why was this part of the code refactored?
5. Which functions were optimized and what is the difference before and after the optimization?

[morningman]

run buildall

[morningman]

run buildall