test.py: Report timings by adding up the time of all commands

tools/testing/dart/test_progress.dart

Index: tools/testing/dart/test_progress.dart
diff --git a/tools/testing/dart/test_progress.dart b/tools/testing/dart/test_progress.dart
index b288db5b064349b6eba742a5b032a1e7bf37b050..497bdca591293a26438be466f721afb9f17de585 100644
--- a/tools/testing/dart/test_progress.dart
+++ b/tools/testing/dart/test_progress.dart
@@ -213,17 +213,23 @@ class TimingPrinter extends EventListener {
   }
 
   void allDone() {
-    // TODO: We should take all the commands into account
+    Duration durationOfTest(testCase) {
+      var duration = const Duration();
+      for (var output in testCase.commandOutputs.values) {
+        duration += output.time;
+      }
+      return duration;
+    }

[kustermann, 2013/07/16 14:48:47]

This procedure will be called O(n*log n) which means that we call it quite a few
times per test. We could cache it, but looking through my performance glasses
this whole things should be changed:

We should use a MaxHeap data structure of 20 elements which would basically
reduce this to linear time. But that's for another CL I guess.

+
     Duration d = (new DateTime.now()).difference(_startTime);
     print('\n--- Total time: ${_timeString(d)} ---');
     _tests.sort((a, b) {
-      Duration aDuration = a.lastCommandOutput.time;
-      Duration bDuration = b.lastCommandOutput.time;
-      return bDuration.inMilliseconds - aDuration.inMilliseconds;
+      return
+          durationOfTest(b).inMilliseconds - durationOfTest(a).inMilliseconds;
     });
     for (int i = 0; i < 20 && i < _tests.length; i++) {
       var name = _tests[i].displayName;
-      var duration = _tests[i].lastCommandOutput.time;
+      var duration = durationOfTest(_tests[i]);
       var configuration = _tests[i].configurationString;
       print('${duration} - $configuration $name');
     }