The real fix here was that the test was using > (greater than 1ms), but if the test ran exeptionally well then the difference was less than 1ms. So changing to >= actually fixes currentTimeMS, and nano just makes it unlikely ever to be =.

Andy.

On 16/07/2014 10:55, Jean-Louis Monteiro wrote:
Ah ok, did not answer yesterday cause I was wondering if my eyes were
blinking or if I was totally out.

What about the test?
Do we remove it?

--
Jean-Louis Monteiro
http://twitter.com/jlouismonteiro
http://www.tomitribe.com


On Wed, Jul 16, 2014 at 10:50 AM, Romain Manni-Bucau <[email protected]>
wrote:

Yep

the link was not linked (hehe) to this particular test but was more
generic.


Romain Manni-Bucau
Twitter: @rmannibucau
Blog: http://rmannibucau.wordpress.com/
LinkedIn: http://fr.linkedin.com/in/rmannibucau
Github: https://github.com/rmannibucau


2014-07-16 10:40 GMT+02:00 Jean-Louis Monteiro <[email protected]>:
right, wrong .... blue, green ... not enough
If we don't need high precision, still not see any link to the blog entry
you pointed.
The test explicitly test we don't fire any asynch method before the bean
become fully initialized (end of @PostConstruct).

That is the only test method of this class.
So, if wrong, fix it!
If not relevant, fix it or remove it!
If fine, then as we sleep 100ms, ms or ns is not that important, I agree,
but we are plenty of tests relying on time, and we know that on windows
it
sucks when using timemillis. Using nano, at least ensure that end -
start >
0


--
Jean-Louis Monteiro
http://twitter.com/jlouismonteiro
http://www.tomitribe.com


On Tue, Jul 15, 2014 at 7:08 PM, Romain Manni-Bucau <
[email protected]>
wrote:

we don't care ns or ms in our tests (which is fine since that's not
exactly what we test) but both are wrong, that's all :).


Romain Manni-Bucau
Twitter: @rmannibucau
Blog: http://rmannibucau.wordpress.com/
LinkedIn: http://fr.linkedin.com/in/rmannibucau
Github: https://github.com/rmannibucau


2014-07-15 18:19 GMT+02:00 Jean-Louis Monteiro <
[email protected]>:
Not sure to understand your point/msg.
I don't really care about the 10 pages benchmarks and I have no time
to
read it.
I only care about the java contract and JVM implementations
accordingly.
It looks out of context here IMHO as we don't care about the absolute
precision, we only care about the difference.
So regarding the test case above, I still agree going with nanoTime is
better than timeMillis.

The test is wrong btw, it should not compare absolute values but only
the
difference.
If there is an interesting outcome, feel free to share, it will be
appreciated I'm sure.

JLouis



--
Jean-Louis Monteiro
http://twitter.com/jlouismonteiro
http://www.tomitribe.com


On Tue, Jul 15, 2014 at 6:04 PM, Romain Manni-Bucau <
[email protected]>
wrote:

can't resist: http://shipilev.net/blog/2014/nanotrusting-nanotime/


Romain Manni-Bucau
Twitter: @rmannibucau
Blog: http://rmannibucau.wordpress.com/
LinkedIn: http://fr.linkedin.com/in/rmannibucau
Github: https://github.com/rmannibucau


2014-07-15 17:58 GMT+02:00 Jean-Louis Monteiro <
[email protected]>:
Definitely for measuring delay nanoTime is better and much more
accurate
than timeMillis especially on non unix machins.



--
Jean-Louis Monteiro
http://twitter.com/jlouismonteiro
http://www.tomitribe.com


On Tue, Jul 15, 2014 at 5:51 PM, <[email protected]> wrote:

Author: andygumbrecht
Date: Tue Jul 15 15:51:08 2014
New Revision: 1610729

URL: http://svn.apache.org/r1610729
Log:
1ms is quite a long time

Modified:


tomee/tomee/branches/tomee-1.7.x/container/openejb-core/src/test/java/org/apache/openejb/core/singleton/AsyncPostContructTest.java
Modified:

tomee/tomee/branches/tomee-1.7.x/container/openejb-core/src/test/java/org/apache/openejb/core/singleton/AsyncPostContructTest.java
URL:

http://svn.apache.org/viewvc/tomee/tomee/branches/tomee-1.7.x/container/openejb-core/src/test/java/org/apache/openejb/core/singleton/AsyncPostContructTest.java?rev=1610729&r1=1610728&r2=1610729&view=diff

==============================================================================
---

tomee/tomee/branches/tomee-1.7.x/container/openejb-core/src/test/java/org/apache/openejb/core/singleton/AsyncPostContructTest.java
(original)
+++

tomee/tomee/branches/tomee-1.7.x/container/openejb-core/src/test/java/org/apache/openejb/core/singleton/AsyncPostContructTest.java
Tue Jul 15 15:51:08 2014
@@ -16,10 +16,12 @@
   */
  package org.apache.openejb.core.singleton;

+import org.apache.openejb.OpenEJB;
  import org.apache.openejb.jee.EnterpriseBean;
  import org.apache.openejb.jee.SingletonBean;
  import org.apache.openejb.junit.ApplicationComposer;
  import org.apache.openejb.testing.Module;
+import org.junit.AfterClass;
  import org.junit.Test;
  import org.junit.runner.RunWith;

@@ -38,6 +40,12 @@ import static org.testng.Assert.assertTr

  @RunWith(ApplicationComposer.class)
  public class AsyncPostContructTest {
+
+    @AfterClass
+    public static void afterClass() throws Exception {
+        OpenEJB.destroy();
+    }
+
      @EJB
      private BuildMeAsync buildMeAsync;

@@ -50,9 +58,9 @@ public class AsyncPostContructTest {

      @Test
      public void postConstructShouldEndsBeforeAsyncCall() {
-        final long start = buildMeAsync.getStartEnd();
+        final long constructed = buildMeAsync.getStartEnd();
          final long async = buildMeAsync.getAsyncStart();
-        assertTrue(async > start);
+        assertTrue(async >= constructed);
          assertSame(buildMeAsync.getAsyncInstance(),
buildMeAsync.getAsyncInstance());
      }

@@ -77,12 +85,12 @@ public class AsyncPostContructTest {
              } catch (final InterruptedException e) {
                  // no-op
              }
-            startEnd = System.currentTimeMillis();
+            startEnd = System.nanoTime();
          }

          @Asynchronous
          public Future<Boolean> async() {
-            asyncStart = System.currentTimeMillis();
+            asyncStart = System.nanoTime();
              asyncInstance = this;
              return new AsyncResult<Boolean>(true);
          }




--
  Andy Gumbrecht

  http://www.tomitribe.com
  [email protected]
  https://twitter.com/AndyGeeDe

  TomEE treibt Tomitribe! | http://tomee.apache.org

Reply via email to