[test] specify surefire provider

[whhe]

fix #1416

[whhe]

There are some errors in mysql connector tests which are related to #1284 and apache/flink@bf55428, other failures have not been located and I think we can fix them in other PRs.

@leonardBang @Jiabao-Sun @GOODBOY008 PTAL.

[ruanhang1993]

@whhe Thanks for your PR.
It seems that this PR forces usage of the JUnit 4.7 provider. Maybe this change is enough to fix the problem by now.
But Flink is promoting the Junit 5 migration in FLINK-25325. I think we should use the JUnit 5 platform as well.

The problem may lie in the conflict in the dependency. And I find the junit version conflict.

[whhe]

@whhe Thanks for your PR. It seems that this PR forces usage of the JUnit 4.7 provider. Maybe this change is enough to fix the problem by now. But Flink is promoting the Junit 5 migration in FLINK-25325. I think we should use the JUnit 5 platform as well.

The problem may lie in the conflict in the dependency. And I find the junit version conflict.

Yes, JUnit provider is forced to be 4.7+ in this PR. I found that the tests of master branch were launched by junit-platform, while the JUnit version of Flink 1.14 was 4.13, which means there is something wrong in provider selection.

If we specify the JUnit provider here, we can also remove it or update to junit-platform when the Flink test utils used in Flink CDC is updated in the future.

[whhe]

#1446 would be able to resolve errors in mysql connector tests.

[ruanhang1993]

It seems like the testMetadataColumns in many connector's xxxConenctorITCase fails. I will take a look at this.

[ruanhang1993]

It seems like the testMetadataColumns in many connector's xxxConenctorITCase fails. I will take a look at this.

Because of this change in Flink(https://issues.apache.org/jira/browse/FLINK-20370), we need to add records whose type is update before when the primary keys of source and sink are different.

I will raise another PR to fix these tests.

[ruanhang1993]

#1451 would be able to resolve errors in testMetadataColumns tests.

[leonardBang]

Thanks @whhe @ruanhang1993 for the great work, but there's a failed test PolardbxSourceITCase, could you take a more look?

[whhe]

Thanks @whhe @ruanhang1993 for the great work, but there's a failed test PolardbxSourceITCase, could you take a more look?

Will have a try tomorrow, as today is the last day of my vacation.

[ruanhang1993]

Thanks @whhe @ruanhang1993 for the great work, but there's a failed test PolardbxSourceITCase, could you take a more look?

I found that this test will sometimes throw an exception like this and the sqls in initializePolardbxTables will not be executed.

If the previous error does not occur, the container may exit with a warning after a few time: [Warning] [MY-011245] [Server] Plugin mysqlx reported: 'Galaxy session 0x7f970000b850 sid 43 kill pool 0x7f9700011a08.'. And some test cases will fail because of the exiting of the polarxdb container.

I have not found the reason why this happens.

[leonardBang]

The test passed in my local env and the azure CI env after I upgrade the polardb-x docker image version, I doubt this issue is related to the docker image version instead of our codebase, I'll merge this pr now.

Thanks @whhe and @ruanhang1993 for the great work.