Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [jakartabatch-dev] Rundown of remaining Batch 2.1 items including finalizing module-info , TCK

I've tested the new JBatch version with GlassFish 7 and all tests now pass, thanks Scott!

Ondro

po 14. 2. 2022 o 18:53 Ondro Mihályi <ondrej.mihalyi@xxxxxxxxx> napísal(a):
Hi Scott,

In the TCK, the symptoms are different than in the issue https://github.com/payara/Payara/issues/5133 but they are fixed with the same fix - when caontainer-managed transactions in the EJB are disabled, all works.

In the TCK, the symptom is that the job is never started, it remains in the "starting" state and the test fails on a timeout (which can be configured with  'tck.execution.waiter.timeout' system property in GlassFish).

I'll test your fix with GlassFish and will let you know soon.

Ondro

po 14. 2. 2022 o 17:13 Scott Kurz <skurz@xxxxxxxxxx> napísal(a):

Ondro had mentioned this issue describes the failure: https://github.com/payara/Payara/issues/5133

 

Let me mention that I released version 2.1.0-M2 of jbatch:

https://search.maven.org/search?q=g:com.ibm.jbatch

 

My local testing seemed to both recreate (before the new version) and resolve (with 2.1.0-M2) a similar-seeming issue.

 

 

From: jakartabatch-dev <jakartabatch-dev-bounces@xxxxxxxxxxx> On Behalf Of Scott Marlow
Sent: Monday, February 14, 2022 11:02 AM
To: jakartabatch developer discussions <jakartabatch-dev@xxxxxxxxxxx>; Ondro Mihályi <ondrej.mihalyi@xxxxxxxxx>
Subject: [EXTERNAL] Re: [jakartabatch-dev] Rundown of remaining Batch 2.1 items including finalizing module-info , TCK

 

 

On 2/12/22 4:25 AM, Ondro Mihályi wrote:

Hi, as we're progressing towards the final release, we communicate mostly via github PRs and Slack.

 

I'll post a brief update here for those that only follow the mailing list.

 

We're now very close to the 2.1 release. From the points that Scott mentioned, we've finished:

 

1. EJB execution in TCK

2. module-info

4. Spec change history

 

We also cleaned the copyright headers (point 5)

 

The only remaining thing is to update the TCK guide (point 3) and the rest of the misc. tasks in 5. While doing it, we plan to release the final version of the Batch 2.1 artifacts to a staging repository, and prepare for submitting a release review.

 

One more thing is that GlassFish now doesn't pass the transaction tests in the EJB execution. This can be fixed later, because GlassFish is needed only to pass the full Jakarta EE TCK so that Jakarta EE 10 can be released, it doesn't block releasing Batch 2.1. We discussed this with Scott and we think that this can be fixed in the JBatch implementation, following the approach of OpenLiberty. OpenLiberty also uses JBatch but provides its own transaction-aware JobOperator, unlike GlassFish, which just uses the default JobOperator from JBatch. Scott will enhance the JBatch job operator to be transaction-aware, so that the GlassFish team doesn't need to do anything to fix this.

Do you have a link for the GlassFish EJB failure just so that we can see more context of what is currently failing and possibly how that might impact other Batch implementations (e.g. whether other implementations may also need fixing to pass EJB tests even after the JBatch job operator is transaction-aware).

Thanks,
Scott

 

 

All the best,

Ondro

 

 

On Wed, Jan 19, 2022, 17:39 Scott Kurz <skurz@xxxxxxxxxx> wrote:

Let me just run down quickly the remaining items for Batch 2.1 EE 10.  

I'll just say upfront we basically have resource lined up (in the form of Ondro and me) to do this...

What I'd most ask for help on is:

* review/feedback for the TCK (e.g. attempt to execute the tests in the milestone release, or even the latest in 'master') .  
* Any last input on module-info (see 2. below).

So the list:

1.  TCK - add EJB execution to Arquillian - Ondro is working on this (thank you)

2.  Review our module-info.   We had a good amount of discussion and ended up deciding to declare cdi and inject each as optional ('static', compile time) deps, and reworking to remove the hard java.logging dependency.

module jakarta.batch {

    requires static jakarta.cdi;
    requires transitive static jakarta.inject;

    uses jakarta.batch.operations.JobOperator;
    // plus export all packages

I dropped the PR approval after pushing the latest so would appreciate someone adding a review:  
https://github.com/eclipse-ee4j/batch-api/pull/192

3. TCK Guide - rework (as we moved from Ant->Maven) - I am planning on doing this.

4. Spec doc - Change History - I will do this if needed, but someone else could take a shot at it if desired.

5 . Everything else  / misc.  (I will plan on doing all this bookkeeping.. might be hard to divide up to get assistance at this point):
  - Legal - copyright review
 - remove batch from platform TCK
 - record CI results,
 - specifications PR

Thanks, getting close..
------------------------------------------------------
Scott Kurz
WebSphere / Open Liberty Batch and Developer Experience
skurz@xxxxxxxxxx
--------------------------------------------------------

_______________________________________________
jakartabatch-dev mailing list
jakartabatch-dev@xxxxxxxxxxx
To unsubscribe from this list, visit https://www.eclipse.org/mailman/listinfo/jakartabatch-dev



_______________________________________________
jakartabatch-dev mailing list
jakartabatch-dev@xxxxxxxxxxx
To unsubscribe from this list, visit https://www.eclipse.org/mailman/listinfo/jakartabatch-dev

Back to the top