cmd/testwrapper/flakytest: don't spam stderr in Mark when not under wrapper

If the user's running "go test" by hand, no need to spam stderr with
the sentinel marker. It already calls t.Logf (which only gets output
on actual failure, or verbose mode) which is enough to tell users it's
known flaky. Stderr OTOH always prints out and is distracting to
manual "go test" users.

Updates #cleanup

Change-Id: Ie5e6881bae291787c30f75924fa132f4a28abbb2
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>

cmd/testwrapper/flakytest/flakytest.go

@@ -19,7 +19,8 @@ import (
 const FlakyTestLogMessage = "flakytest: this is a known flaky test"
 
 // FlakeAttemptEnv is an environment variable that is set by cmd/testwrapper
-// when a flaky test is retried. It contains the attempt number, starting at 1.
+// when a flaky test is being (re)tried. It contains the attempt number,
+// starting at 1.
 const FlakeAttemptEnv = "TS_TESTWRAPPER_ATTEMPT"
 
 var issueRegexp = regexp.MustCompile(`\Ahttps://github\.com/tailscale/[a-zA-Z0-9_.-]+/issues/\d+\z`)
@@ -33,7 +34,11 @@ func Mark(t testing.TB, issue string) {
 	if !issueRegexp.MatchString(issue) {
 		t.Fatalf("bad issue format: %q", issue)
 	}
-
-	fmt.Fprintln(os.Stderr, FlakyTestLogMessage) // sentinel value for testwrapper
+	if _, ok := os.LookupEnv(FlakeAttemptEnv); ok {
+		// We're being run under cmd/testwrapper so send our sentinel message
+		// to stderr. (We avoid doing this when the env is absent to avoid
+		// spamming people running tests without the wrapper)
+		fmt.Fprintln(os.Stderr, FlakyTestLogMessage)
+	}
 	t.Logf("flakytest: issue tracking this flaky test: %s", issue)
 }