92 lines
5.6 KiB
Markdown
92 lines
5.6 KiB
Markdown
# repeated
|
|
|
|
Companion module for [JUnit @RepeatedTest (JUnit 5 → 6)](https://ankurm.com/) on ankurm.com.
|
|
|
|
## What this is
|
|
|
|
`@RepeatedTest` and `RepetitionInfo` are **byte-for-byte unchanged** between JUnit Jupiter 5.10.1
|
|
and 6.1.3 — confirmed here with `javap` against the real jars, not assumed from a changelog. The
|
|
"JUnit 5 → 6" part of this module's story is almost entirely a version-number story: JUnit 6
|
|
unified versioning across Platform, Jupiter and Vintage (they all now share one release number,
|
|
confirmed against the real `maven-metadata.xml` for all three artifacts) and raised the minimum
|
|
Java version to 17. Nothing about `@RepeatedTest` itself had to change for that.
|
|
|
|
What *did* need fixing is what the previous version of this post got wrong or left out, each
|
|
verified against a real jar, a real sources jar, or a real run rather than against prose:
|
|
|
|
- `{shortDisplayname}` is not a real placeholder. There are exactly three:
|
|
`DISPLAY_NAME_PLACEHOLDER`, `CURRENT_REPETITION_PLACEHOLDER`, `TOTAL_REPETITIONS_PLACEHOLDER`
|
|
(confirmed via `javap` on `RepeatedTest.class`). A `name()` pattern containing the bogus token
|
|
is accepted at compile time and left completely unresolved at run time — see
|
|
`docs/output/03-bogus-placeholder-scratch.txt`.
|
|
- `RepetitionInfo` carries **four** accessors, not two. `getFailureCount()` and
|
|
`getFailureThreshold()` have been on the interface since JUnit 5.10, the same release that
|
|
added `failureThreshold()` itself.
|
|
- `RepetitionInfo` can be injected into `@BeforeEach`/`@AfterEach`, not only into the
|
|
`@RepeatedTest` method.
|
|
- Once `failureThreshold` is reached, JUnit **skips** the remaining repetitions — it does not
|
|
merely "stop" them, and they are never invoked at all. See
|
|
`docs/output/07-failure-threshold-skips-remaining.txt` for the real console-tree output showing
|
|
`Failure threshold [2] exceeded` against repetitions 9 and 10.
|
|
|
|
**Self-correction made before this module was finished:** the "flaky test detection" use case
|
|
needs a reproducible failure, not real timing-dependent flakiness, so both demo scratch classes
|
|
fail deterministically on fixed repetition numbers (4 and 8 of 10) rather than relying on actual
|
|
race conditions. Both were run for real, their output captured, and then deleted — they were
|
|
never meant to be part of the permanent green suite.
|
|
|
|
## Versions
|
|
|
|
| Dependency | Version | Verified against |
|
|
|---|---|---|
|
|
| JUnit Jupiter | 6.1.3 | `maven-metadata.xml` (`latest`/`release`), matching `junit-bom` and `junit-platform-launcher` — JUnit 6's unified versioning |
|
|
| AssertJ | 3.27.7 | `maven-metadata.xml`, `<release>` |
|
|
| Maven Surefire | 3.6.0 | this module's own build |
|
|
| JDK | 25 (Temurin) | `java -version`, this module's own build |
|
|
|
|
`@RepeatedTest`/`RepetitionInfo` class files are identical (via `javap`) across 5.10.1 and 6.1.3.
|
|
`getFailureCount()`/`getFailureThreshold()` are absent in 5.9.3's `RepeatedTest.class` and present
|
|
from 5.10.1 onward — the old post's "introduced in 5.10" claim for `failureThreshold` itself is
|
|
correct; it just never documented these two accessors.
|
|
|
|
## Quickstart
|
|
|
|
```bash
|
|
mvn test # 14 tests, 0 failures
|
|
```
|
|
|
|
## Test classes
|
|
|
|
| Class | What it proves | Captured output |
|
|
|---|---|---|
|
|
| `ReliabilityTest` | The smallest `@RepeatedTest(5)`; real default display names resolved through injected `TestInfo` | `docs/output/01-basic-repetition.txt` |
|
|
| `CustomNameTest` | A custom `name()` pattern built from the three real placeholders | `docs/output/02-custom-display-name.txt` |
|
|
| `MetadataTest` | All four `RepetitionInfo` accessors, including the two the old post never covered | `docs/output/04-repetition-info-metadata.txt` |
|
|
| `LifecycleTest` | `@BeforeAll`/`@AfterAll` (once) vs `@BeforeEach`/`@AfterEach` (per repetition), plus `RepetitionInfo` injected into `@BeforeEach`/`@AfterEach` | `docs/output/05-lifecycle-order.txt` |
|
|
|
|
## Captured output
|
|
|
|
| File | Contents |
|
|
|---|---|
|
|
| `docs/output/00-full-test-run.txt` | `mvn test` — all 4 committed classes, 14 tests, 0 failures |
|
|
| `docs/output/01-basic-repetition.txt` | `ReliabilityTest` — real default display names |
|
|
| `docs/output/02-custom-display-name.txt` | `CustomNameTest` — real custom display names |
|
|
| `docs/output/03-bogus-placeholder-scratch.txt` | Scratch run proving `{shortDisplayname}` is not a real placeholder and resolves to literal text |
|
|
| `docs/output/04-repetition-info-metadata.txt` | `MetadataTest` — all four `RepetitionInfo` accessors across 3 repetitions |
|
|
| `docs/output/05-lifecycle-order.txt` | `LifecycleTest` — real call order and totals across 3 repetitions |
|
|
| `docs/output/06-flaky-detected.txt` | Scratch run: no `failureThreshold` set — all 10 repetitions run even though 2 fail |
|
|
| `docs/output/07-failure-threshold-skips-remaining.txt` | Scratch run: `failureThreshold = 2` — repetitions 9 and 10 skipped after the 2nd failure |
|
|
|
|
`docs/output/03`, `06` and `07` come from scratch classes (`ScratchPlaceholderTest`,
|
|
`FlakyDetectionDemoTest`, `FailureThresholdStopsRemainingDemoTest`) that were written, run once
|
|
for real evidence, and then deleted — they are not part of the committed suite, which stays green
|
|
at 14/14. Re-run `scripts/run-all.sh` to see exactly how each transcript was produced.
|
|
|
|
## What's sourced from documentation, not run here
|
|
|
|
The JUnit 6.0.0 release notes facts quoted in the post (minimum Java 17, Vintage deprecated,
|
|
`junit-platform-runner`/`junit-platform-jfr` removal, `MethodOrderer.Alphanumeric` removal,
|
|
`@CsvSource` moving to FastCSV) are quoted from the official JUnit 6.0.0 release notes, not
|
|
exercised in this module — none of them touch `@RepeatedTest` or `RepetitionInfo` directly.
|
|
Everything else above was compiled, run, or pulled directly from a real jar in this sitting.
|