testing: capture panics, present them, and mark the test as a failure.

R=r
CC=golang-dev
https://golang.org/cl/5633044

src/pkg/testing/testing.go

@@ -225,6 +225,19 @@ func (c *common) Fatalf(format string, args ...interface{}) {
 	c.FailNow()
 }
 
+// TODO(dsymonds): Consider hooking into runtime·traceback instead.
+func (c *common) stack() {
+	for i := 2; ; i++ { // Caller we care about is the user, 2 frames up
+		pc, file, line, ok := runtime.Caller(i)
+		f := runtime.FuncForPC(pc)
+		if !ok || f == nil {
+			break
+		}
+		c.Logf("%s:%d (0x%x)", file, line, pc)
+		c.Logf("\t%s", f.Name())
+	}
+}
+
 // Parallel signals that this test is to be run in parallel with (and only with) 
 // other parallel tests in this CPU group.
 func (t *T) Parallel() {
@@ -247,6 +260,13 @@ func tRunner(t *T, test *InternalTest) {
 	// a call to runtime.Goexit, record the duration and send
 	// a signal saying that the test is done.
 	defer func() {
+		// Consider any uncaught panic a failure.
+		if err := recover(); err != nil {
+			t.failed = true
+			t.Log(err)
+			t.stack()
+		}
+
 		t.duration = time.Now().Sub(t.start)
 		t.signal <- t
 	}()