-
75. Re: Regression on AS6 org.jboss.test.jca.test.TransactionActiveUnitTestCase
timfox Jun 14, 2010 3:05 PM (in response to clebert.suconic)a) We can get the test to pass by reverting, so clearly your statement that it's impossible to fix without interacting with the tx mgr is false. As I've said a few times now, this test used to pass! Without any tx mgr interactions.
b) Once the AS test suite is passing, we look at fixing the TCK issue. Now, we all agree that it's not valid to fix a TCK issue by adding non portable requirements, e.g. tx mgr interactions.
Reverting would probably take about 1 hour. Like I said earlier, you could have done it several times over in the time taken to argue this
And finally, does anyone have a link to the thread where the TCK failure was discussed? I want to look at how this can be fixed.
-
76. Re: Regression on AS6 org.jboss.test.jca.test.TransactionActiveUnitTestCase
clebert.suconic Jun 14, 2010 3:11 PM (in response to timfox)a) Sure.. then we will break the TCK.
The "fixes" we had to make the test pass was by not allowing dual behaviour on XASessions.
What makes you think reverting will fix both the TCK and the Test?
I have spent a great deal of time hacking the TransactionManager. When a timeout happens, all we get is a rollback. And there is no way to known the TX is timed out.
b) Then we will need to revert a to make the TCK pass... and then we fix a. and then we fix b...
Interchanging between a and b is all we have done this past three months. We need a definitive solution that works on both.
The only way is by changing the TM or asking the TM if there is a pending TX. (Have I said that already?)
-
77. Re: Regression on AS6 org.jboss.test.jca.test.TransactionActiveUnitTestCase
clebert.suconic Jun 14, 2010 3:13 PM (in response to timfox)"And finally, does anyone have a link to the thread where the TCK failure was discussed? I want to look at how this can be fixed."
These are all the links I have regarding this thread. (There's a link on the first page of the thread):
https://jira.jboss.org/browse/HORNETQ-328
-
78. Re: Regression on AS6 org.jboss.test.jca.test.TransactionActiveUnitTestCase
timfox Jun 14, 2010 3:16 PM (in response to clebert.suconic)Clebert Suconic wrote:
What makes you think reverting will fix both the TCK and the Test?
Are you actually bothering to read my posts before your reply?
I never said reverting would fix both test suite and TCK issues. I clearly said it would only fix the AS test suite issue, and then we look at fixing the TCK.
And we all agree that it's not valid to fix a TCK issue by adding non portable requirements.
You're making statements that's it's impossible to fix the AS test suite without adding tx mgr interactions, for the third time, I am saying this is clearly false, this test used to pass without any such code.
I'm calling time on this thread.
Andy, please revert back the code tomorrow, then we'll look at fixing the TCK issue PROPERLY.
-
79. Re: Regression on AS6 org.jboss.test.jca.test.TransactionActiveUnitTestCase
timfox Jun 14, 2010 3:23 PM (in response to clebert.suconic)Clebert Suconic wrote:
"And finally, does anyone have a link to the thread where the TCK failure was discussed? I want to look at how this can be fixed."
These are all the links I have regarding this thread. (There's a link on the first page of the thread):
https://jira.jboss.org/browse/HORNETQ-328
Those threads don't discuss the TCK failure. I'm looking for the original thread, probably from Jeff which discusses the TCK failure, NOT the AS test suite failure
-
80. Re: Regression on AS6 org.jboss.test.jca.test.TransactionActiveUnitTestCase
clebert.suconic Jun 14, 2010 3:28 PM (in response to timfox)"And we all agree that it's not valid to fix a TCK issue by adding non portable requirements."
Actually no.. I've talked to both Andy and Jeff this morning on IRC, and you're the only one who thinks the fix is invalid.
We will revert the code, and the issue of having to interact or change the TM will raise again.
"You're making statements that's it's impossible to fix the AS test suite without adding tx mgr interactions, for the third time, I am saying this is clearly false, this test used to pass without any such code."
Then the issue here is beyond my expertise. I already provided you facts that we never had the TCK and the Test passing at the same time.
The TM doesn't give us the distinction of timeouts.
-
81. Re: Regression on AS6 org.jboss.test.jca.test.TransactionActiveUnitTestCase
timfox Jun 14, 2010 3:33 PM (in response to clebert.suconic)Clebert Suconic wrote:
Then the issue here is beyond my expertise. I already provided you facts that we never had the TCK and the Test passing at the same time.
Again, you are putting words in my mouth. I never said it would fix both test suite and TCK!!
But it will fix the test suite.
You said it was impossible to fix the test suite without adding tx mgr interaction. That is clearly FALSE.
-
82. Re: Regression on AS6 org.jboss.test.jca.test.TransactionActiveUnitTestCase
clebert.suconic Jun 14, 2010 3:37 PM (in response to timfox)"You said it was impossible to fix the test suite without adding tx mgr interaction. That is clearly FALSE."
No, I said it's impossible to fix both the test and the TCK without adding the tx mgr integration or fixing the TX MGr to send us a failure when the timeout happened. (post edited this line here.. had a typo on test and the testsuite.. Clebert)
Message was edited by: Clebert Suconic
-
83. Re: Regression on AS6 org.jboss.test.jca.test.TransactionActiveUnitTestCase
timfox Jun 14, 2010 3:36 PM (in response to timfox)Andy, Clebert -
Once you have reverted the code so the test suite passes, please post details of the failing TCK test and we will take it from there.
-
84. Re: Regression on AS6 org.jboss.test.jca.test.TransactionActiveUnitTestCase
timfox Jun 14, 2010 3:41 PM (in response to clebert.suconic)Clebert Suconic wrote:
No, I said it's impossible to fix both the test and the TCK without adding the tx mgr integration or fixing the TX MGr to send us a failure when the timeout happened. (post edited this line here.. had a typo on test and the testsuite.. Clebert)
Message was edited by: Clebert Suconic
No, you didn't say that,
Try reading the thread again.
This is exactly what you said:
"*There is no way to fix TransactionActiveUnitTestCase without involving the TM somehow. (By fixint it or interacting with it).
I think I made my point very clear. If you want to ignore the warning and spend the whole week fixing an issue that didn't exist.. that's fine."
This statement doesn't mention the TCK at all. It says it's impossible to fix the test suite test without involving TM. That is FALSE
-
85. Re: Regression on AS6 org.jboss.test.jca.test.TransactionActiveUnitTestCase
clebert.suconic Jun 14, 2010 3:46 PM (in response to timfox)The context here is fixing both the TCK and TransactionActiveUnitTestCase.
I can certainly fix TransactionActiveUnitTestCase by not letting XASessions to have the dual behaviour expected by the Testsuite. We have bounced between adding and removing this somehow over the past three months. I have already said this on the thread.
And I don't want to do it again.
I certainly don't like the Sessions cached on the ManagedConnection, but that won't change things. In the end you will have to either raise a bug on the TM and ask them and convince them to change their code, or change our code to interact with the TM.
-
86. Re: Regression on AS6 org.jboss.test.jca.test.TransactionActiveUnitTestCase
clebert.suconic Jun 14, 2010 3:47 PM (in response to clebert.suconic)"And I don't want to do it again"
^^ By this phrase I meant... I don't want to bounce over removing one fix and breaking the other ^^
-
87. Re: Regression on AS6 org.jboss.test.jca.test.TransactionActiveUnitTestCase
clebert.suconic Jun 14, 2010 3:58 PM (in response to clebert.suconic)If we revert the cached Sessions on the adapter, first thing we need to do is let the XASession to behave as non-transacted when not-enlisted.
-
88. Re: Regression on AS6 org.jboss.test.jca.test.TransactionActiveUnitTestCase
ataylor Jun 14, 2010 4:04 PM (in response to clebert.suconic)I have no idea what needs to be changes need to be reversed as I was not involved, Clebert ig you can do this i will look at the original TCK issue tomorrow.
-
89. Re: Regression on AS6 org.jboss.test.jca.test.TransactionActiveUnitTestCase
clebert.suconic Jun 14, 2010 4:07 PM (in response to ataylor)I don't know what to do:
I can either avoid the Session from having the dual behaviour. What will break TCK tests:
Here is the list of tests that will fail (for Jboss employees only):
http://hudson.qa.jboss.com/hudson/view/TCK5-EAP51-JDK6/job/tck5-eap51-jdk6-jms-hornetq/8/
Or we can let the session having dual behaviour, what will let timed out Transactions to be accepted as non transactional. To fix the later now you need the TM in place to test and validate an error.