tison
@codelipenghui still not found. Is there any estimate? Since 2.10.1 is an official release we must fix this issue though >
This can be related to https://github.com/apache/pulsar/issues/17468.
@Shoothzj IIUC do you suggest a replacement of powermock here?
@Shoothzj thank you. I'll consider this approach when protptyping. There're some cases we try to hijack third-party lib's field, like `eventLoopGroup` in bookkeeper code. In those situations, we should still...
I draft an incompleted patch on #16997. However, I notice that there's some code we hack in "set final" field, like: https://github.com/apache/pulsar/blob/21634edd30bbe33d231ec73e8045d7666a8ad366/pulsar-broker/src/test/java/org/apache/pulsar/client/api/SimpleProducerConsumerTest.java#L1071-L1072 .. or even "set static final" field, like:...
/pulsarbot run-failure-checks
/pulsarbot run-failure-checks
It seems CI failed only on known flaky test https://github.com/apache/pulsar/issues/17011.
@lhotari For `PulsarByteBufAllocatorDefaultTest` and `PulsarByteBufAllocatorOomThrowExceptionTest`, they assert the config and it's well verified. For `BatchMessageContainerImplTest`, you're right that OOME isn't tested anymore. I'll try to investigate a bit.
Fixed in cbda29f5d370f592c77512abfe384efa220f4a6d. Perhaps we can consider something like [failpoint](https://github.com/pingcap/failpoint) for Golang to noninvasive inject failure. I don't know whether the JVM ecosystem already has a such library, though. UPDATE:...