Multitest annotations now accept multiple annotations on one line.

tools/testing/dart/test_runner.dart

Index: tools/testing/dart/test_runner.dart
diff --git a/tools/testing/dart/test_runner.dart b/tools/testing/dart/test_runner.dart
index d3c97e120ff8a83775c6a784639b05e0695b601d..2ac1bdaf6d714a49b8c1ffd46effd554e0212195 100644
--- a/tools/testing/dart/test_runner.dart
+++ b/tools/testing/dart/test_runner.dart
@@ -337,10 +337,12 @@ class AnalysisTestOutputImpl extends TestOutputImpl {
   }
   
   bool _didMultitestFail(List errors, List staticWarnings) {
-    String outcome = testCase.info.multitestOutcome;
-    if ((outcome == '' || outcome == 'compile-time error') && errors.length > 0) {
+    Set<String> outcome = testCase.info.multitestOutcome;

[Bill Hesse, 2012/03/08 10:21:49]

It should be checked that we never reach here with a null multitestOutcome -
this is only enforced by control flow.

[zundel, 2012/03/08 13:18:00]

Added assertion

+    if ((outcome.contains('') || outcome.contains('compile-time error')) 

[Bill Hesse, 2012/03/08 10:21:49]

Isn't this check about not containing 'run-time error', not about not containing
the other two?  '' shouldn't really be an element of the set anyway - you may
want to use the empty set instead.

[zundel, 2012/03/08 13:18:00]

Besides runtime-error, we also don't want 'errors' populated for dynamic-type
errors and static-type errors. I think the root of the problem is that a test
annotated with 'compile-time' error is supposed to come back marked as 'failed',
even though this is the expected outcome.  If we get rid of the 'negative test'
concept, I think this might be simpler.

+        && errors.length > 0) {
       return true;
-    } else if (outcome == 'static type error' && staticWarnings.length > 0) {
+    } else if (outcome.contains('static type error') 
+        && staticWarnings.length > 0) {
       return true;
     }
     return false;