pjfanning opened a new pull request, #3475:
URL: https://github.com/apache/pekko/pull/3475
### Motivation
Extending `AbstractPersistentActorWithTimers` from Java fails to compile
against the Scala 3 artifacts:
```
TestActor.java:[5,8] types org.apache.pekko.actor.Timers and
org.apache.pekko.persistence.Eventsourced are incompatible;
class TestActor inherits unrelated defaults for
aroundPreRestart(java.lang.Throwable,scala.Option<java.lang.Object>)
from types org.apache.pekko.actor.Timers and
org.apache.pekko.persistence.Eventsourced
```
Both `Timers` and `Eventsourced` implement `aroundReceive`,
`aroundPreRestart` and `aroundPostStop`. Scala 3 *does* emit the mixin
forwarder into the class, but flags it `ACC_BRIDGE, ACC_SYNTHETIC`:
```
public void aroundPreRestart(java.lang.Throwable, scala.Option);
flags: (0x1041) ACC_PUBLIC, ACC_BRIDGE, ACC_SYNTHETIC
```
javac ignores synthetic and bridge members when resolving inherited members,
so it falls back to the two interface defaults and rejects the subclass. Scala
2.13 emits the same forwarder without those flags, which is why the 2.12/2.13
artifacts work. This is accidental, not a deliberate limitation.
The reporter asked whether other multi-trait base classes are affected. I
compiled a javac probe (an abstract Java subclass) against the Scala 3 build
output for every Java-facing `Abstract*`/`UntypedAbstract*` class that mixes in
more than one trait — the stash, logging and timers variants,
`AbstractLoggingFSM`, `AbstractFSMWithStash`, `AbstractPersistentActor`,
`AbstractPersistentActorWithTimers`,
`AbstractPersistentActorWithAtLeastOnceDelivery`, `AbstractPersistentFSM`,
`AbstractPersistentLoggingFSM`, `journal.japi.AsyncWriteJournal`,
`AbstractInOutHandler` and `AbstractSerializationSupport`. Exactly one other
class has the same defect: `AbstractFSMWithStash`, where `FSM` and
`UnrestrictedStash` both implement `postStop`.
### Modification
Give both classes real (non-synthetic) overrides of the conflicting members
that simply delegate to `super`, which resolves exactly like the forwarders
they replace, so behaviour is unchanged on both Scala versions.
`AbstractFSMWithStash.postStop` deliberately carries no
`@throws(classOf[Exception])`: `FSM.postStop` declares no checked exceptions,
and a wider throws clause is not a valid override for javac.
Added Java sources exercising both classes, so the Scala 3 build fails if
this regresses:
- `persistence/src/test/java/.../JavaTimerPersistentActor.java`, driven by a
new case in `TimerPersistentActorSpec`
- `actor-tests/src/test/java/.../AbstractFSMWithStashActorTest.java`, a
JUnit 5 test
### Result
Java subclasses of `AbstractPersistentActorWithTimers` and
`AbstractFSMWithStash` compile against the Scala 3 artifacts. No other
Java-facing multi-trait base class in the build is affected.
### Tests
- `sbt "++3.3.8" "persistence/testOnly
org.apache.pekko.persistence.TimerPersistentActorSpec"` — 5 succeeded, 0 failed
- `sbt "++3.3.8" "actor-tests/testOnly
org.apache.pekko.actor.AbstractFSMWithStashActorTest
org.apache.pekko.actor.AbstractFSMActorTest"` — 2 succeeded, 0 failed
- `sbt "persistence/testOnly
org.apache.pekko.persistence.TimerPersistentActorSpec"` (2.13) — 5 succeeded, 0
failed
- `sbt "actor-tests/testOnly
org.apache.pekko.actor.AbstractFSMWithStashActorTest
org.apache.pekko.actor.AbstractFSMActorTest"` (2.13) — 2 succeeded, 0 failed
- `sbt headerCreateAll javafmtAll scalafmtAll scalafmtSbt` on JDK 17 —
headers generated by sbt, no formatting churn outside the changed files
- `sbt +mimaReportBinaryIssues` — Not run locally, left to the CI `Check /
Binary Compatibility` job
Both Java test sources fail to compile under Scala 3 without the
corresponding production change, and compile once it is applied.
### References
Fixes #3474
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
