wang-jiahua opened a new issue, #1239:
URL: https://github.com/apache/rocketmq-client-go/issues/1239
### Describe the Bug
`QueueLock.fetchLock` in `consumer/lock.go` uses a **value receiver** on a
struct that contains a `sync.Map`:
```go
type QueueLock struct { lockTable sync.Map }
func (ql QueueLock) fetchLock(queue primitive.MessageQueue) sync.Locker {
v, _ := ql.lockTable.LoadOrStore(queue, new(sync.Mutex))
return v.(*sync.Mutex)
}
```
Every call copies the whole struct (flagged by `go vet` copylocks), so
`LoadOrStore` writes into a discarded copy and the original map stays empty
forever — **each call returns a brand-new `*sync.Mutex`**.
The only caller is `consumeMessageOrderly` (`consumer/push_consumer.go`):
`lock := pc.queueLock.fetchLock(*mq); lock.Lock()`. Since every goroutine gets
its own mutex, the per-queue mutual exclusion is completely ineffective:
multiple goroutines can process the same `MessageQueue` concurrently, breaking
the FIFO guarantee that orderly consumption exists to provide.
### Steps to Reproduce
Two deterministic tests (included in the incoming PR):
1. Identity: call `fetchLock` twice for the same queue — unfixed returns two
different mutexes.
2. Mutual exclusion: 20 goroutines fetch the lock for the same queue and
record max concurrency — unfixed observes `maxConcurrent=20` (zero
serialization); fixed observes `maxConcurrent=1`.
`go vet ./consumer/` also reports the copylocks diagnostic on this method.
### What Did You Expect to See?
Same queue → same lock; orderly consumption strictly serialized per queue.
### What Did You See Instead?
A fresh lock per call; per-queue ordering not enforced.
### Additional Context
Fix incoming: change the receiver to `*QueueLock` (one character). `go vet`
copylocks disappears; both tests flip from FAIL to PASS; the consumer suite
passes with no regressions.
--
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]
