Workaround race condition when child process exits

There's race condition happening (so far observed only on CentOS) at the
moment when child process exits.  Sometimes reading from master fd
blocks forever even when the process is already reaped and there's no
more data to be read. Not sure what's the reason, it may be some CentOS
specific bug.

Using timeout does the job and seems to be relatively harmless.

terminal/terminal.go

@@ -6,6 +6,7 @@ import (
 	"os/exec"
 	"os/signal"
 	"syscall"
+	"time"
 
 	"code.google.com/p/go.crypto/ssh/terminal"
 
@@ -53,7 +54,7 @@ func (p *Pty) Record(command string, stdoutCopy io.Writer) error {
 	}()
 	defer close(signals)
 
-	// put stdin into raw mode (if it's a tty)
+	// put stdin in raw mode (if it's a tty)
 	fd := int(p.Stdin.Fd())
 	if terminal.IsTerminal(fd) {
 		oldState, err := terminal.MakeRaw(fd)
@@ -71,7 +72,23 @@ func (p *Pty) Record(command string, stdoutCopy io.Writer) error {
 
 	// copy pty master -> p.stdout & stdoutCopy
 	stdout := io.MultiWriter(p.Stdout, stdoutCopy)
+	stdoutWaitChan := make(chan struct{})
+	go func() {
 		io.Copy(stdout, master)
+		stdoutWaitChan <- struct{}{}
+	}()
+
+	// wait for the process to exit and reap it
+	cmd.Wait()
+
+	// wait for master -> stdout copying to finish
+	//
+	// sometimes after process exits reading from master blocks forever (race condition?)
+	// we're using timeout here to overcome this problem
+	select {
+	case <-stdoutWaitChan:
+	case <-time.After(200 * time.Millisecond):
+	}
 
 	// stop stdin -> master copying
 	stop()