rpc: remove shutdown deadlock

shutdown was originally written for Conn.Close and included a wait for
goroutines to report done, as is appropriate.  However, when shutdown
was called from one of the manager's goroutines (in case of an abort,
etc.), the goroutines would deadlock.  Now manager.wait is separated out,
and I made Conn.Wait use the manager.wait function.  This has the
side-effect of making Conn.Wait wait for the goroutines to finish, which
is more correct than the old behavior.

Fixes #45

rpc/manager.go

@@ -52,9 +52,9 @@ func (m *manager) do(f func()) {
 	}
 }
 
-// shutdown closes the finish channel and sets the error.
-// The first call to shutdown returns true; subsequent calls are no-ops
-// and return false.
+// shutdown closes the finish channel and sets the error.  The first
+// call to shutdown returns true; subsequent calls are no-ops and return
+// false.  This will not wait for the manager's goroutines to finish.
 func (m *manager) shutdown(e error) bool {
 	m.mu.Lock()
 	ok := !m.done
@@ -64,12 +64,16 @@ func (m *manager) shutdown(e error) bool {
 		m.e = e
 	}
 	m.mu.Unlock()
-	if ok {
-		m.wg.Wait()
-	}
 	return ok
 }
 
+// wait blocks until the manager is shut down and all of its goroutines
+// are finished.
+func (m *manager) wait() {
+	<-m.finish
+	m.wg.Wait()
+}
+
 // err returns the error passed to shutdown.
 func (m *manager) err() error {
 	m.mu.RLock()

rpc/rpc.go

@@ -129,7 +129,7 @@ func NewConn(t Transport, options ...ConnOption) *Conn {
 // Wait waits until the connection is closed or aborted by the remote vat.
 // Wait will always return an error, usually ErrConnClosed or of type Abort.
 func (c *Conn) Wait() error {
-	<-c.manager.finish
+	c.manager.wait()
 	return c.manager.err()
 }
 
@@ -139,6 +139,7 @@ func (c *Conn) Close() error {
 	if !c.manager.shutdown(ErrConnClosed) {
 		return ErrConnClosed
 	}
+	c.manager.wait()
 	// Hang up.
 	// TODO(light): add timeout to write.
 	ctx := context.Background()