Fix bug 3789:

Turn off assert in automated tests for timer latency.
This test isn't helping a lot, so removing it has no 
negative effect.  But the test fails on some VMs which
run horribly slow.

Leaving the test in place because it still prints out
timer perf information, which can be useful for testing
on laptops, low-end machines, etc.

Review URL: http://codereview.chromium.org/14853

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@7245 0039d316-1c4b-4281-b951-d872f2087c98

1 files changed, 7 insertions, 1 deletions

diff --git a/base/time_unittest_win.cc b/base/time_unittest_win.cc
index d0f606d4..4b63f6b 100644
--- a/base/time_unittest_win.cc
+++ b/base/time_unittest_win.cc
@@ -190,7 +190,13 @@ TEST(TimeTicks, TimerPerformance) {
     for (int index = 0; index < kLoops; index++)
       cases[test_case].func();
     TimeTicks stop = TimeTicks::HighResNow();
-    EXPECT_LT((stop - start).InMilliseconds(), kMaxTime);
+    // Turning off the check for acceptible delays.  Without this check,
+    // the test really doesn't do much other than measure.  But the
+    // measurements are still useful for testing timers on various platforms.
+    // The reason to remove the check is because the tests run on many 
+    // buildbots, some of which are VMs.  These machines can run horribly
+    // slow, and there is really no value for checking against a max timer.
+    //EXPECT_LT((stop - start).InMilliseconds(), kMaxTime);
     printf("%s: %1.2fus per call\n", cases[test_case].description, 
       (stop - start).InMillisecondsF() * 1000 / kLoops);
     test_case++;