Fix connReaper goroutine may leak

go.mod

@@ -7,3 +7,5 @@ require (
 	golang.org/x/sync v0.3.0
 	golang.org/x/text v0.13.0
 )
+
+require go.uber.org/goleak v1.3.0 // indirect

go.sum

@@ -1,5 +1,7 @@
 github.com/go-zeromq/goczmq/v4 v4.2.2 h1:HAJN+i+3NW55ijMJJhk7oWxHKXgAuSBkoFfvr8bYj4U=
 github.com/go-zeromq/goczmq/v4 v4.2.2/go.mod h1:Sm/lxrfxP/Oxqs0tnHD6WAhwkWrx+S+1MRrKzcxoaYE=
+go.uber.org/goleak v1.3.0 h1:2K3zAYmnTNqV73imy9J1T3WC+gmCePx2hEGkimedGto=
+go.uber.org/goleak v1.3.0/go.mod h1:CoHD4mav9JJNrW/WLlf7HGZPjdw8EucARQHekz1X6bE=
 golang.org/x/sync v0.3.0 h1:ftCYgMx6zT/asHUrPw8BLLscYtGznsLAnjq5RH9P66E=
 golang.org/x/sync v0.3.0/go.mod h1:FU7BRWz2tNW+3quACPkgCx/L+uEAv1htQ0V83Z9Rj+Y=
 golang.org/x/text v0.13.0 h1:ablQoSUd0tRdKxZewP80B+BaqeKJuVhuRxj/dkrun3k=

reaper/reaper_test.go

@@ -0,0 +1,77 @@
+// Copyright 2024 The go-zeromq Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package reaper
+
+import (
+	"context"
+	"errors"
+	"sync"
+	"testing"
+	"time"
+
+	"github.com/go-zeromq/zmq4"
+	"go.uber.org/goleak"
+)
+
+// TestReaper does multiple rapid Dial/Close to check that connection reaper goroutines are not leaking.
+// In is in own package as goleak detects also threads from objects instantiated by init().
+func TestReaper(t *testing.T) {
+	defer goleak.VerifyNone(t)
+	//assert := asserts.New(t)
+
+	ctx, cancel := context.WithCancel(context.Background())
+	rep := zmq4.NewRep(ctx)
+	ep := "ipc://@test.rep.socket"
+	err := rep.Listen(ep)
+	_ = err // assert.NoError(err)
+
+	maxClients := 100
+	maxMsgs := 100
+	wgClients := &sync.WaitGroup{}
+	wgServer := &sync.WaitGroup{}
+	client := func() {
+		defer wgClients.Done()
+		for n := 0; n < maxMsgs; n++ {
+			func() {
+				ctx, cancel := context.WithTimeout(context.Background(), 1*time.Second)
+				defer cancel()
+				req := zmq4.NewReq(ctx)
+				err := req.Dial(ep)
+				_ = err // assert.NoError(err)
+
+				err = req.Close()
+				_ = err // assert.NoError(err)
+			}()
+		}
+	}
+	server := func() {
+		defer wgServer.Done()
+		pong := zmq4.NewMsgString("pong")
+		for {
+			msg, err := rep.Recv()
+			if errors.Is(err, context.Canceled) {
+				break
+			}
+			// assert.NoError(err)
+			if err != nil {
+				break
+			}
+			_ = msg // assert.Equal(string(msg.Frames[0]), "ping")
+			err = rep.Send(pong)
+			_ = err // assert.NoError(err)
+		}
+	}
+
+	wgServer.Add(1)
+	go server()
+	wgClients.Add(maxClients)
+	for n := 0; n < maxClients; n++ {
+		go client()
+	}
+	wgClients.Wait()
+	cancel()
+	wgServer.Wait()
+	rep.Close()
+}

socket.go

@@ -120,8 +120,12 @@ func (sck *socket) topics() []string {
 
 // Close closes the open Socket
 func (sck *socket) Close() error {
+	// The Lock around Signal ensures the connReaper is running
+	// and is in sck.reaperCond.Wait()
+	sck.reaperCond.L.Lock()
 	sck.cancel()
 	sck.reaperCond.Signal()
+	sck.reaperCond.L.Unlock()
 
 	if sck.listener != nil {
 		defer sck.listener.Close()
@@ -193,7 +197,11 @@ func (sck *socket) Listen(endpoint string) error {
 	sck.listener = l
 
 	go sck.accept()
-	go sck.connReaper()
+	if !sck.reaperStarted {
+		sck.reaperCond.L.Lock()
+		go sck.connReaper()
+		sck.reaperStarted = true
+	}
 
 	return nil
 }
@@ -268,6 +276,7 @@ connect:
 	}
 
 	if !sck.reaperStarted {
+		sck.reaperCond.L.Lock()
 		go sck.connReaper()
 		sck.reaperStarted = true
 	}
@@ -372,7 +381,11 @@ func (sck *socket) Timeout() time.Duration {
 }
 
 func (sck *socket) connReaper() {
-	sck.reaperCond.L.Lock()
+	// We are not locking here sck.reaperCond.L.Lock()
+	// as it should be locked prior starting connReaper as goroutine
+	// That would ensure that sck.reaperCond.Signal()
+	// would be delivered only when reaper goroutine is really started
+	// and is in sck.reaperCond.Wait()
 	defer sck.reaperCond.L.Unlock()
 
 	for {