net: deflake TestDialTimeout

This change makes TestDialTimeout work on almost all the supported
platforms.

Updates #3016.
Updates #3307.
Updates #3867.
Updates #5380.
Updates #5349.

Change-Id: Iacf0ebea23cdd8f6c0333d70c667a5a5f5eb0ed2
Reviewed-on: https://go-review.googlesource.com/8220
Reviewed-by: Ian Lance Taylor <iant@golang.org>

7 files changed, 67 insertions, 103 deletions

diff --git a/src/net/dial_gen.go b/src/net/dial_gen.go
index ada6233003..654ef9383a 100644
--- a/src/net/dial_gen.go
+++ b/src/net/dial_gen.go
@@ -6,11 +6,7 @@
 
 package net
 
-import (
-	"time"
-)
-
-var testingIssue5349 bool // used during tests
+import "time"
 
 // dialChannel is the simple pure-Go implementation of dial, still
 // used on operating systems where the deadline hasn't been pushed
@@ -31,9 +27,7 @@ func dialChannel(net string, ra Addr, dialer func(time.Time) (Conn, error), dead
 	}
 	ch := make(chan racer, 1)
 	go func() {
-		if testingIssue5349 {
-			time.Sleep(time.Millisecond)
-		}
+		testHookDialChannel()
 		c, err := dialer(noDeadline)
 		ch <- racer{c, err}
 	}()
diff --git a/src/net/dial_gen_test.go b/src/net/dial_gen_test.go
deleted file mode 100644
index c857acd06d..0000000000
--- a/src/net/dial_gen_test.go
+++ /dev/null
@@ -1,11 +0,0 @@
-// Copyright 2013 The Go Authors.  All rights reserved.
-// Use of this source code is governed by a BSD-style
-// license that can be found in the LICENSE file.
-
-// +build windows plan9
-
-package net
-
-func init() {
-	testingIssue5349 = true
-}
diff --git a/src/net/dial_test.go b/src/net/dial_test.go
index 42898d669f..15ee71b9a0 100644
--- a/src/net/dial_test.go
+++ b/src/net/dial_test.go
@@ -31,89 +31,6 @@ func newLocalListener(t *testing.T) Listener {
 	return ln
 }
 
-func TestDialTimeout(t *testing.T) {
-	origBacklog := listenerBacklog
-	defer func() {
-		listenerBacklog = origBacklog
-	}()
-	listenerBacklog = 1
-
-	ln := newLocalListener(t)
-	defer ln.Close()
-
-	errc := make(chan error)
-
-	numConns := listenerBacklog + 100
-
-	// TODO(bradfitz): It's hard to test this in a portable
-	// way. This is unfortunate, but works for now.
-	switch runtime.GOOS {
-	case "linux":
-		// The kernel will start accepting TCP connections before userspace
-		// gets a chance to not accept them, so fire off a bunch to fill up
-		// the kernel's backlog.  Then we test we get a failure after that.
-		for i := 0; i < numConns; i++ {
-			go func() {
-				_, err := DialTimeout("tcp", ln.Addr().String(), 200*time.Millisecond)
-				errc <- err
-			}()
-		}
-	case "darwin", "plan9", "windows":
-		// At least OS X 10.7 seems to accept any number of
-		// connections, ignoring listen's backlog, so resort
-		// to connecting to a hopefully-dead 127/8 address.
-		// Same for windows.
-		//
-		// Use an IANA reserved port (49151) instead of 80, because
-		// on our 386 builder, this Dial succeeds, connecting
-		// to an IIS web server somewhere.  The data center
-		// or VM or firewall must be stealing the TCP connection.
-		//
-		// IANA Service Name and Transport Protocol Port Number Registry
-		// <http://www.iana.org/assignments/service-names-port-numbers/service-names-port-numbers.xml>
-		go func() {
-			c, err := DialTimeout("tcp", "127.0.71.111:49151", 200*time.Millisecond)
-			if err == nil {
-				err = fmt.Errorf("unexpected: connected to %s!", c.RemoteAddr())
-				c.Close()
-			}
-			errc <- err
-		}()
-	default:
-		// TODO(bradfitz):
-		// OpenBSD may have a reject route to 127/8 except 127.0.0.1/32
-		// by default. FreeBSD likely works, but is untested.
-		// TODO(rsc):
-		// The timeout never happens on Windows.  Why?  Issue 3016.
-		t.Skipf("skipping test on %q; untested.", runtime.GOOS)
-	}
-
-	connected := 0
-	for {
-		select {
-		case <-time.After(15 * time.Second):
-			t.Fatal("too slow")
-		case err := <-errc:
-			if err == nil {
-				connected++
-				if connected == numConns {
-					t.Fatal("all connections connected; expected some to time out")
-				}
-			} else {
-				terr, ok := err.(timeout)
-				if !ok {
-					t.Fatalf("got error %q; want error with timeout interface", err)
-				}
-				if !terr.Timeout() {
-					t.Fatalf("got error %q; not a timeout", err)
-				}
-				// Pass. We saw a timeout error.
-				return
-			}
-		}
-	}
-}
-
 func TestSelfConnect(t *testing.T) {
 	if runtime.GOOS == "windows" {
 		// TODO(brainman): do not know why it hangs.
diff --git a/src/net/hook_plan9.go b/src/net/hook_plan9.go
new file mode 100644
index 0000000000..e053348505
--- /dev/null
+++ b/src/net/hook_plan9.go
@@ -0,0 +1,9 @@
+// Copyright 2015 The Go 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 net
+
+import "time"
+
+var testHookDialChannel = func() { time.Sleep(time.Millisecond) } // see golang.org/issue/5349
diff --git a/src/net/hook_unix.go b/src/net/hook_unix.go
index 626d07fbb0..bb198e0228 100644
--- a/src/net/hook_unix.go
+++ b/src/net/hook_unix.go
@@ -9,6 +9,8 @@ package net
 import "syscall"
 
 var (
+	testHookDialChannel = func() {} // see golang.org/issue/5349
+
 	// Placeholders for socket system calls.
 	socketFunc        func(int, int, int) (int, error)         = syscall.Socket
 	closeFunc         func(int) error                          = syscall.Close
diff --git a/src/net/hook_windows.go b/src/net/hook_windows.go
index 2a6e5bf267..59c32d2c98 100644
--- a/src/net/hook_windows.go
+++ b/src/net/hook_windows.go
@@ -4,9 +4,14 @@
 
 package net
 
-import "syscall"
+import (
+	"syscall"
+	"time"
+)
 
 var (
+	testHookDialChannel = func() { time.Sleep(time.Millisecond) } // see golang.org/issue/5349
+
 	// Placeholders for socket system calls.
 	socketFunc    func(int, int, int) (syscall.Handle, error)                                               = syscall.Socket
 	closeFunc     func(syscall.Handle) error                                                                = syscall.Closesocket
diff --git a/src/net/timeout_test.go b/src/net/timeout_test.go
index 9ef0c4d15c..b46321b13b 100644
--- a/src/net/timeout_test.go
+++ b/src/net/timeout_test.go
@@ -8,11 +8,59 @@ import (
 	"fmt"
 	"io"
 	"io/ioutil"
+	"net/internal/socktest"
 	"runtime"
 	"testing"
 	"time"
 )
 
+func TestDialTimeout(t *testing.T) {
+	const T = 100 * time.Millisecond
+
+	switch runtime.GOOS {
+	case "plan9", "windows":
+		origTestHookDialChannel := testHookDialChannel
+		testHookDialChannel = func() { time.Sleep(2 * T) }
+		defer func() { testHookDialChannel = origTestHookDialChannel }()
+		if runtime.GOOS == "plan9" {
+			break
+		}
+		fallthrough
+	default:
+		sw.Set(socktest.FilterConnect, func(so *socktest.Status) (socktest.AfterFilter, error) {
+			time.Sleep(2 * T)
+			return nil, errTimeout
+		})
+		defer sw.Set(socktest.FilterConnect, nil)
+	}
+
+	ch := make(chan error)
+	go func() {
+		// This dial never starts to send any SYN segment
+		// because of above socket filter and test hook.
+		c, err := DialTimeout("tcp", "127.0.0.1:0", T)
+		if err == nil {
+			err = fmt.Errorf("unexpectedly established: tcp:%s->%s", c.LocalAddr(), c.RemoteAddr())
+			c.Close()
+		}
+		ch <- err
+	}()
+	tmo := time.NewTimer(3 * T)
+	defer tmo.Stop()
+	select {
+	case <-tmo.C:
+		t.Fatal("dial has not returned")
+	case err := <-ch:
+		nerr, ok := err.(Error)
+		if !ok {
+			t.Fatalf("got %v; want error implements Error interface", err)
+		}
+		if !nerr.Timeout() {
+			t.Fatalf("got %v; want timeout error", err)
+		}
+	}
+}
+
 func isTimeout(err error) bool {
 	e, ok := err.(Error)
 	return ok && e.Timeout()