Don't ever report a failure when the sum of the reference count differences

are zero.  This should help reduce the false positives.

The message about references leaking is maintained to provide as much
info as possible rather than simply suppressing the message at the source.

Lib/test/regrtest.py

@@ -682,9 +682,10 @@ def dash_R(the_module, test, indirect_test, huntrleaks):
             deltas.append(sys.gettotalrefcount() - rc - 2)
     print >> sys.stderr
     if any(deltas):
-        print >> sys.stderr, test, 'leaked', deltas, 'references'
+        msg = '%s leaked %s references, sum=%s' % (test, deltas, sum(deltas))
+        print >> sys.stderr, msg
         refrep = open(fname, "a")
-        print >> refrep, test, 'leaked', deltas, 'references'
+        print >> refrep, msg
         refrep.close()
 
 def dash_R_cleanup(fs, ps, pic):

Misc/build.sh

@@ -192,7 +192,7 @@ if [ $err = 0 -a "$BUILD_DISABLED" != "yes" ]; then
             ## ensure that the reflog exists so the grep doesn't fail
             touch $REFLOG
             $PYTHON $REGRTEST_ARGS -R 4:3:$REFLOG -u network $LEAKY_SKIPS >& build/$F
-            NUM_FAILURES=`egrep -vc "$LEAKY_TESTS" $REFLOG`
+            NUM_FAILURES=`egrep -vc "($LEAKY_TESTS|sum=0)" $REFLOG`
             update_status "Testing refleaks ($NUM_FAILURES failures)" "$F" $start
             mail_on_failure "refleak" $REFLOG