Hi All, I've opened a jira stating the `ArchitectureTest` is failing on our latest master. The solution in the PR [2] is obvious: re-generate the violations.
I've double checked the changes and seems like majority of the time only line numbers are different which are fine. But there are new violations where I can't make proper judgement (those class has been changed by others). If somebody has made changes in the following areas then please share thoughts on the PR: *CompactCoordinator* Accesses InProgressFileWriter$PendingFileRecoverable.getPath() and .getSize() directly → Violates layer separation by accessing internal sink details *CompactorOperator* Calls StreamingRuntimeContext.getTaskInfo() in multiple locations → Runtime context leakage into operator logic *CompactorOperatorStateHandler* Heavy use of CommittableWithLineage.getCommittable() and PendingFileRecoverable.getPath() → Breaks encapsulation of committables and file recoverables *AbstractStreamingWriter* Uses internal classes: Buckets, StreamingFileSinkHelper, and StreamRecord internals → Introduces strong coupling to legacy or internal sink/file system components *BatchFileWriter* Accesses StreamTask.getEnvironment() and TaskManagerRuntimeInfo.getConfiguration() → Direct environment access violates operator abstraction BR, G [1] https://issues.apache.org/jira/browse/FLINK-37884 [2] https://github.com/apache/flink/pull/26625
