fix(message): enumerate read queues when listing queue offsets

[zjncs]

Problem

RocketMQMessageProvider.getQueueOffsets lists the Queue Browser's queues with queueData.getWriteQueueNums(), but every other message-browse path is read-queue based:

• the broker's PullMessageProcessor rejects any pull with queueId >= topicConfig.getReadQueueNums() (SYSTEM_ERROR "queueId is illegal");
• fetchSubscribeMessageQueues — used by this provider's own queryByTopic and by the classic console's queryMessageByTopic — enumerates [0, readQueueNums).

So whenever a topic has readQueueNums != writeQueueNums (the standard expansion/shrink procedure sets the two counts independently, e.g. updateTopic -r 16 -w 8), the Queue Browser is wrong:

• read > write (shrink draining): queues in [write, read) are still readable and still hold messages, but they never appear in the browser — the user cannot browse them at all, and the queue list disagrees with what queryByTopic scans.
• write > read (queues not yet readable): queues in [read, write) are listed, but each one is a dead end — pulling from it fails with "queueId is illegal".

The writeQueueNums iteration was introduced incidentally by the admin-client pooling refactor b5a7d6c (#2544), not as a deliberate queue-selection choice.

Fix

Enumerate [0, queueData.getReadQueueNums()) in getQueueOffsets, matching the broker's pull validation and the read-queue enumeration used by the other browse paths. One-line change.

Verification

Base SHA: 3612602 (rocketmq-studio)

Fail-before / pass-after on the two new tests in RocketMQMessageProviderTest:

• getQueueOffsetsListsReadQueuesWhenReadCountExceedsWriteCount (read=4, write=2): failed with [0, 1] vs expected [0, 1, 2, 3]; passes after the fix (also asserts min/max offsets are wired per queue).
• getQueueOffsetsSkipsWriteOnlyQueuesWhenWriteCountExceedsReadCount (read=2, write=4): failed with [0, 1, 2, 3] vs expected [0, 1]; passes after the fix.

Full run after the fix:

• RocketMQMessageProviderTest: 38/38 pass (36 pre-existing + 2 new)
• whole org.apache.rocketmq.studio.provider.apache package: 256/256 pass

AI disclosure: This change was prepared with AI assistance (GitHub Copilot/Claude-style tooling guided by a human contributor).

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[RockteMQ-AI]

Summary

Correct fix — getQueueOffsets should enumerate read queues, not write queues:

• Consumers read from read queues
• Broker rejects pull requests for queueId >= readQueueNums with SYSTEM_ERROR
• When writeQueueNums > readQueueNums, old code listed non-browsable queues
• When readQueueNums > writeQueueNums (queue shrinking), old code missed queues with messages

Tests validate both scenarios (read > write and write > read).

LGTM.

---

Automated review by github-manager-bot