[222twotwotwo]

What this PR does:

Prevents client shutdown from deadlocking or hanging when session teardown can trigger reconnect logic.

This PR:

• closes client sessions outside the client map lock by snapshotting ssMap during shutdown
• introduces runReconnect() to start reconnect loops only while the client is open and track them with the client WaitGroup
• routes session-triggered reconnect through runReconnect() instead of synchronously calling reConnect()
• ensures reconnect backoff is cancelled by Client.Close()
• adds regression coverage for shutdown/reconnect interleavings, backoff cancellation, and no-op reconnect after close

Which issue(s) this PR fixes:

Fixes #132

Special notes for your reviewer:

The branch has been merged with the latest origin/master

Additional tests cover:

• session error triggering reconnect while Client.Close() is running
• Client.Close() cancelling reconnect backoff
• runReconnect() becoming a no-op after client shutdown
• Client.Close() waiting for an in-flight session-triggered reconnect

Does this PR introduce a user-facing change?:

NONE

[AlexStocks]

[P1] 这里把整个重连 goroutine 计入 wg 后，NewSessionCallback 内调用公开的 Client.Close() 会等待当前 goroutine 自己，形成确定性死锁。固定 Base/Head 的公开入口探针中，Base 在回调调用 Close() 并返回错误后正常退出；当前 Head 在 500ms 内始终无法返回。回调文档没有禁止关闭所属 endpoint，这类连接校验失败后主动停机的用法不能被锁死。请避免让 Close() 等待正在执行该回调的 worker（例如把停止信号与外部 join 分离，或调整重连任务的生命周期登记），并补一个回调内关闭客户端的回归测试。

[222twotwotwo]

收到， NewSessionCallback内调用 Client.Close() 时不会再等待当前正在执行 callback 的 goroutine 自己。

[Copilot]

Pull request overview

This PR addresses a client lifecycle deadlock during shutdown by decoupling session teardown from reconnect logic and ensuring reconnect work is coordinated and cancellable via client shutdown.

Changes:

• Routes session-triggered reconnect through a new client.runReconnect() helper instead of calling reConnect() inline from session teardown.
• Updates client.stop() to snapshot and close sessions outside the client lock to avoid lock re-entrancy during shutdown.
• Adds regression tests covering shutdown/reconnect interleavings and reconnect backoff cancellation behavior.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.

File	Description
session.go	Switches session teardown from synchronous reConnect() to asynchronous, tracked runReconnect() to avoid shutdown lock re-entry.
client.go	Introduces runReconnect() (tracked via WaitGroup), waits for initial reconnect in RunEventLoop, and closes sessions outside the client lock during stop().
client_test.go	Adds regression tests for reconnect/backoff cancellation and shutdown/reconnect interleavings.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.