Refactor stages 1, 2 and test_api overhaul.

scripts/slave/recipe_modules/auto_bisect/resources/significantly_different_inner.py

Index: scripts/slave/recipe_modules/auto_bisect/resources/significantly_different_inner.py
diff --git a/scripts/slave/recipe_modules/auto_bisect/resources/significantly_different_inner.py b/scripts/slave/recipe_modules/auto_bisect/resources/significantly_different_inner.py
index a2285bc17099b6c4c4e9764983fd97e2c2c1cb93..f6b9d74584f74cd8932df646d352bf49b1881a8d 100644
--- a/scripts/slave/recipe_modules/auto_bisect/resources/significantly_different_inner.py
+++ b/scripts/slave/recipe_modules/auto_bisect/resources/significantly_different_inner.py
@@ -13,12 +13,11 @@ except ImportError:
 else:
 
   def main():
-    if len(sys.argv) < 4:
+    if len(sys.argv) < 3:
       return 1
-    _, list_a, list_b, significance = sys.argv[:4]
+    _, list_a, list_b = sys.argv[:3]
     list_a = json.loads(list_a)
     list_b = json.loads(list_b)
-    significance = float(significance)
 
     shapiro_p_value = stats.shapiro(list_a)[1], stats.shapiro(list_b)[1]
     mann_whitney_p_value = stats.mannwhitneyu(list_a, list_b).pvalue
@@ -34,16 +33,6 @@ else:
         'welch_p_value': welch_p_value,
     }
 
-    # TODO(robertocn): It seems we haven't used the results of shapiro test for
-    # normality. We should remove this along with anderson darling and welch's.
-    if (results['shapiro_p_value'][0] < significance and
-        results['shapiro_p_value'][1] < significance):
-      results['normal-y'] = True
-    else:
-      results['normal-y'] = False
-    results['significantly_different'] = bool(
-        float(results['mann_p_value']) < float(significance))
-
     print json.dumps(results)
     return 0