Fix WebServiceContext injection into EJB3 beans

<Jaikiran> ropalka: hi
<ropalka> Jaikiran, I found one regression
<ropalka> Jaikiran,
<ropalka> —
<ropalka> tests-ws:
<ropalka> [junit] Running org.jboss.test.ws.jaxws.ejb3Integration.WebServiceTestCase
<ropalka> [junit] Tests run: 5, Failures: 0, Errors: 1, Time elapsed: 12.337 sec
<ropalka> [junit] Test org.jboss.test.ws.jaxws.ejb3Integration.WebServiceTestCase FAILED
<Jaikiran> what's that?
<ropalka> —
<Jaikiran> what's the error?
<ropalka> Jaikiran, NPE
<ropalka> Jaikiran, This test http://fpaste.org/iK7w/

• istudens_wfh has quit (Read error: Connection reset by peer)
  <ropalka> Jaikiran, don't get injected @Resource
• jpechane (~jpechane@dhcp-1-171.brq.redhat.com) has joined #jbossas
  <Jaikiran> let me check
  <ropalka> Jaikiran, This is the NPE - http://fpaste.org/3fZc/
  <ropalka> Jaikiran, U need to update webservices AS module and rebuild it (or whole AS first)
  <Jaikiran> ropalka: is this committed in AS trunk? I just tried running this test but see a different error
  <Jaikiran> ok, one sec
  <ropalka> Jaikiran, U can rebuild just webservices module
  <ropalka> Jaikiran, and created artifact rename to jbossws-jboss60.jar and copy to AS/common/lib
  <ropalka> Jaikiran, this is the fastest way to go
  <carlo> do we still have a working internal pastebin?
  <Jaikiran> ropalka: can you enable trace level logs of org.jboss.switchboard and send me the logs?
  <ropalka> carlo, I didn't try, why?
  <ropalka> Jaikiran, give me a sec
  <carlo> for a completely other problem
  <ropalka> carlo, aha, ok
  <ropalka> Jaikiran, How can I run just this one particular test?
  <ropalka> Jaikiran, I'm using
  <ropalka> —
  <Jaikiran> ropalka: ./build.sh one-test -Dtest=org.jboss.test.ws.jaxws.ejb3Integration.WebServiceTestCase
  <ropalka> [nucleus][/home/opalka/svn/jbossas/trunk/testsuite]>./build.sh -Dnode0=localhost tests-ws
  <ropalka> —
  <carlo> FYI: pastebin.test.redhat.com
  <Jaikiran> ah, looks like the url changed
  <ropalka> Jaikiran, thanks
  <ropalka> Jaikiran, Where should I enable logging?
  <ropalka> Jaikiran, in jboss-6.0.0-SNAPSHOT/server/all/deploy/jboss-logging.xml?
  <Jaikiran> ropalka: yes
  <ropalka> Jaikiran, what level, DEBUG?
  <ropalka> Jaikiran, or TRACE?
• pbravo has quit (Ping timeout: 240 seconds)
  <Jaikiran> ropalka: trace should be better. enable for both org.jboss.switchboard and org.jboss.injection
  <ropalka> Jaikiran, OK
• istudens (~istudens@vpn2-11-241.ams2.redhat.com) has joined #jbossas
• istudens is now known as istudens_wfh
• mschoene (~mschoene@vpn1-7-138.ams2.redhat.com) has joined #jbossas
  <ropalka> Jaikiran, I'm sending U trace log via e-mail
  <Jaikiran> ok
  <ropalka> Jaikiran, sent
  <Jaikiran> ropalka: looking at the logs, the injection seems to be happening.
  <Jaikiran> ropalka: are you sure the getMessageContext() returned by the context here http://fpaste.org/iK7w/ is not null?
  <ropalka> Jaikiran, will debug
• shaun (~sappleto@dhcp-1-10.fab.redhat.com) has joined #jbossas
• mlinhard (~mlinhard@vpn-11-2.rdu.redhat.com) has joined #jbossas
• tkimura has quit (Ping timeout: 612 seconds)
• tkonishi has quit (Quit: Leaving)
• tkimura (~tkimura@dhcp-193-208.nrt.redhat.com) has joined #jbossas
• rhusar (~rhusar@vpn1-7-197.ams2.redhat.com) has joined #jbossas
  <ropalka> Jaikiran, I confirm the problem is returned wsctx is null
• bgeorges has quit (Remote host closed the connection)
  <ropalka> Jaikiran, I'm having a look ...
  <Jaikiran> ropalka: wsCtx is null or messagecontext?
  <Jaikiran> looking at the logs i did notice that it was picking up the value for injection from env/org.jboss.test.ws.jaxws.ejb3Integration.HandlerContextEndpoint/wsCtx for injecting into targetclass class org.jboss.test.ws.jaxws.ejb3Integration.HandlerContextEndpoint targetname wsCtx
  <ropalka> Jaikiran, messagecontext, sorry for confusion
  <Jaikiran> oh ok
  <ropalka> Jaikiran,
  <ropalka> Jaikiran, I found the reason but I'd like to discuss the fix with you
  <Jaikiran> ropalka: sure
  <ropalka> Jaikiran, I'd like to know EJB3 impl details
  <ropalka> Jaikiran, The problem is I don't set message context for EJB3 beans (just JAXWS POJO beans)
• emuckenhuber (~emuckenh@dhcp-144-239.gva.redhat.com) has joined #jbossas
  <ropalka> Jaikiran, I can set it up for EJB3 beans
  <ropalka> Jaikiran, But I need to be sure
  <Jaikiran> ropalka: so it was EJB3 which was setting this up so far?
  <ropalka> Jaikiran, EJB3 beans are not executed using ThreadPools or staff like that
  <ropalka> Jaikiran, Before yes
  <ropalka> Jaikiran, Since SB changes no.
  <ropalka> Jaikiran, But this is fixable in our AS integration layer
  <Jaikiran> hmm i think this is related to that pending jira we have
  <ropalka> Jaikiran, Which one?
  <Jaikiran> ropalka: EJBTHREE-2161 i guess
• mnovotny (~mnovotny@dhcp-27-205.brq.redhat.com) has joined #jbossas
  <JiraBot> EJBTHREE-2161 Implement @WebService support for @Singleton EJB3 beans [Coding In Progress, Critical, jaikiran pai] https://jira.jboss.org/jira/browse/EJBTHREE-2161
  <ropalka> Jaikiran, This is @Stateless bean
  <ropalka> Jaikiran, Maybe JIRA description needs review
  <ropalka> Jaikiran, No
  <ropalka> Jaikiran, This is JIRA I created
  <ropalka> Jaikiran, This is correct description
  <ropalka> Jaikiran, As I said earlier
  <Jaikiran> ok, so it's a different thing then
  <ropalka> Jaikiran, This is fixable in our integration layer
  <ropalka> Jaikiran, But I need to be sure that if I set thread local
  <Jaikiran> sure. some background on the issue would be helpful, please
  <ropalka> Jaikiran, EJB3 isn't using custom thread pool (so it will be visible to EJB3 layer too)
  <ropalka> Jaikiran, OK
  <ropalka> Jaikiran, Do U have time now?
  <Jaikiran> ropalka: yep
  <ropalka> Jaikiran, very good
  <ropalka> Jaikiran, The problem is
  <ropalka> Jaikiran, This was before handled in EJB3 component
  <ropalka> Jaikiran, and seems the proper place to fix it now is in WS integration layer (at least IMHO)
  <ropalka> Jaikiran, How it works for JAXWS POJO beans
  <ropalka> Jaikiran, There's InvocationHandler abstraction in JBossWS SPI
  <ropalka> Jaikiran, JAXWS POJO invocation handler (org.jboss.wsf.common.invocation.InvocationHandlerJAXWS)
  <ropalka> Jaikiran, implements org.jboss.wsf.common.invocation.AbstractInvocationHandlerJSE
  <ropalka> Jaikiran, which provides the following template methods
  <ropalka> —
  <ropalka> protected void onBeforeInvocation(final Invocation invocation) throws Exception
  <ropalka> protected void onAfterInvocation(final Invocation invocation) throws Exception
  <ropalka> —
• darran (~darran@vpn1-7-205.ams2.redhat.com) has joined #jbossas
  <ropalka> Jaikiran, these template methods are implemented in JAXWS POJO invocation handler (org.jboss.wsf.common.invocation.InvocationHandlerJAXWS)
  <ropalka> —
  <ropalka> @Override
  <ropalka> protected void onBeforeInvocation(final Invocation invocation)
  <ropalka> { <ropalka> final WebServiceContext wsContext = this.getWebServiceContext(invocation); <ropalka> ThreadLocalAwareWebServiceContext.getInstance().setMessageContext(wsContext); <ropalka> }

  <carlo> ropalka: this goes back to EAP 4 where I did a proper fix for WebServiceContext
  <ropalka> @Override
  <ropalka> protected void onAfterInvocation(final Invocation invocation)
  <ropalka>

  { <ropalka> ThreadLocalAwareWebServiceContext.getInstance().setMessageContext(null); <ropalka> }

  <ropalka> —
  <ropalka> carlo, ???
  <carlo> This whole before invocation and after invocation is horseshit
  <ropalka> carlo, listening ...
  <carlo> WebServiceContext needs to be injected before @PostConstruct
  <carlo> And then be done with it.
  <ropalka> carlo, stop
  <ropalka> carlo, This is not our problem
  <ropalka> carlo, I'm installing ThreadLocalAwareWebServiceContext before @PostConstruct is called
  <ropalka> carlo, So U have it available
  <ropalka> carlo, Then onBeforeInvocation() and onAfterInvocation() just use thread locals to set proper context before/after method invocations
  <ropalka> carlo, In order to have it working in EJB3 as well (this is working for POJOs)
  <ropalka> carlo, I need to be sure EJB3's not using custom thread pools (iow my thread local changes will be visible there)
  <ropalka> carlo, Is it clear now how it works for you?

• maeste has quit (Ping timeout: 622 seconds)
  <ropalka> carlo, If not, as questions
  <ropalka> s/as/ask/
• eko (~eko@dhcp-2-223.brq.redhat.com) has joined #jbossas
  <ropalka> Jaikiran, carlo
  <ropalka> Jaikiran, carlo I propose the same fix for EJB3
  <ropalka> Jaikiran, carlo I mean the one we're using for JAX-WS POJOs
  <carlo> ropalka: where is onAfterInvocation ?
  <carlo> What I want to get rid of is BeanContextLifecycleCallback
  <ropalka> carlo, This would be called after EJB3 method invocation
• tkobayas has quit (Quit: Leaving)
  <ropalka> carlo, In webservices AS integration module
  <ropalka> carlo, concretely in finally block of method invoke in org.jboss.webservices.integration.invocation.InvocationHandlerEJB3
  <carlo> ropalka: where is that code?
  <ropalka> carlo, jbossas/trunk/webservices/src/main/java/org/jboss/webservices/integration/invocation/InvocationHandlerEJB3.java
  <ropalka> carlo, This code isn't there yet (WS context injection population)
  <ropalka> carlo, I'm now proposing fix and need to verify with U (EJB3) this is proper one
• stliu has quit (Quit: Leaving)
  <carlo> ropalka: the problem is that you are not allowed to inject anything after @PostConstruct, is that fixed?
  <ropalka> carlo, Yes
  <ropalka> carlo, This is impl. detail
  <ropalka> carlo, As I said earlier
  <ropalka> carlo, Our jbossas/trunk/webservices/src/main/java/org/jboss/webservices/integration/injection/WebServiceContextResourceProvider.java
  <ropalka> carlo, installs factory object
  <ropalka> —
  <ropalka> @Override
  <ropalka> public Resource provide(final DeploymentUnit unit, final JBossResourceEnvRefType resEnvRef)
  <ropalka> { <ropalka> return new WebServiceContextResource(ThreadLocalAwareWebServiceContext.getInstance()); <ropalka> }

  <ropalka> —
  <ropalka> carlo, The staff we're discussing now is about proper setup of message context (which is always unique for each invocation)
  <ropalka> carlo, this default WS context factory uses thread locals (impl. detail)
  <carlo> ropalka: just make it a theadlocal
  <ropalka> carlo, what?
  <carlo> the current MessageContext
  <ropalka> carlo, yep
  <ropalka> carlo, The WS context factory is singleton
  <ropalka> carlo, using thread locals for message context propagation
  <ropalka> carlo, is it clear now?
  <carlo> Yes, so we don't need any after or before invocation
  <ropalka> carlo, EJB3 team not
  <ropalka> carlo, But I need to fix it in webservices AS trunk module
  <ropalka> carlo, because this is not propagate now
  <ropalka> carlo, as i said message context is unique per invocation
  <carlo> There is one tricky issue though: JBPAPP-4429
  <JiraBot> JBPAPP-4429 Calling getContextData() on InvocationContext passed to EJB3 Interceptor should return the JAX-WS MessageContext for JAX-WS Calls. [Open, Major, Carlo de Wolf] https://jira.jboss.org/jira/browse/JBPAPP-4429

• ropalka 's having a look
• lami (~lakagwu@dhcp-1-142.fab.redhat.com) has joined #jbossas
• jhalliday (~jhalli@ovpn-113-66.phx2.redhat.com) has joined #jbossas
• bbelovic is now known as bbelovic-lunch
  <ropalka> carlo, OK, what the question U wanna ask?
  <ropalka> s/what/what's/
  <carlo> ropalka: nothing, just get rid of these weird callbacks
  <ropalka> carlo, so U like my proposal fix?
• pskopek is now known as pskopek|otl
  <carlo> ropalka: yes
  <ropalka> carlo, OK, now question of the day I need to know answer for.
  <ropalka> carlo, If I set thread local before invoking EJB3 method
  <carlo> ropalka: but in future we may switch threads
  <carlo> but I'm not sure whether that's worth pondering upon now
  <ropalka> carlo, "Is it true EJB3 AS6 impl. don't use custom thread pools, so my change will be visible to EJB3?"
  <carlo> Yes, use ThreadLocal
  <ropalka> carlo, Jaikiran thanks for confirmation
  <carlo> You may need ThreadLocalStack though
  <ropalka> carlo, ???
  <ropalka> carlo, This is some EJB3 impl?
  <ropalka> carlo, I googled http://docs.jboss.org/jbossaop/docs/1.5.0.GA/docs/aspect-framework/api/org/jboss/aop/util/ThreadLocalStack.html
  <carlo> picture the following scenario: 1. call A. 2. call B. 3. back in A
  <carlo> We got a 1000+ impls
• ropalka 's listening ...
  <carlo> the one in ejb3-context is the fastest
  <ropalka> carlo, Could U give me the pointer?
  <carlo> Anyway in A I have MessageContext A in the ThreadLocal, now I call B
  <ropalka> U mean U call B from A?
  <carlo> In B I have MessageContext B, now I return to A
  <carlo> Yes
  <ropalka> carlo, ok, go on ...
  <carlo> If I use a regular ThreadLocal I would be staring at MessageContext B, instead of A
  <ropalka> carlo, Let me clarify scenario first before continuing in discussion
  <ropalka> —
  <ropalka> WS servlet -> A EJB3 (msgCTX A) -> B EJB3 (msgCTX A)
  <ropalka> —
  <ropalka> or U would like to have
  <ropalka> —
  <ropalka> WS servlet -> A EJB3 (msgCTX A) -> B EJB3 (msgCTX B)
  <ropalka> —
  <ropalka> carlo, What's the proper/expected behavior?
  <carlo> I mean: WS Servet A -> A EJB 3 (msgCtx A) -> WS Servlet B -> B EJB (msgCtx B)
  <ropalka> carlo, WS can set just msgCTX A, because this is related to current invocation
  <ropalka> carlo, OK, this should be correct
• jcosta (~jcosta@dhcp-27-223.brq.redhat.com) has joined #jbossas
  <ropalka> carlo, What's the problem U're trying to discuss here?
  <carlo> Right, then I come back in A (after B call)
  <carlo> I would see msgCtx B instead of A when using ThreadLocal
  <ropalka> carlo, ok, go on
  <ropalka> carlo, I don't think so?
  <ropalka> carlo, Tomcat container uses thread pols
  <carlo> why not?
  <ropalka> s/pols/pools/
  <ropalka> carlo, Each servlet invocation should use unique thread AFAIK
  <carlo> impl detail
• Devang|Training is now known as Devang
  <ropalka> carlo, U mean this is impl. detail I shouldn't rely on?
  <carlo> Yes
  <ropalka> carlo, I don't think so. AFAIK this is covered with Servlet spec.
  <ropalka> carlo, Or it changed in Servlet 3.0 spec?
  <carlo> I don't know, nor want to know
• ropalka didn't read Servlet 3.0 spec yet
  <ropalka> carlo, Well, IMHO this is something in servlet spec since 2.x version
  <carlo> If you say you can get away without ThreadLocalStack, then go for it
  <ropalka> carlo, I think so
  <ropalka> carlo, Just for clarification
  <ropalka> carlo, What U mean with "without ThreadLocalStack"?
  <ropalka> carlo, I need to understand what U mean
  <carlo> ThreadLocalStack is meant to solve scenarios which I outlined
  <carlo> We use it over local (/ same thread) invocations.
  <ropalka> carlo, So it isn't InheritableThreadLocal related thing
  <carlo> No, InheritableThreadLocal is meant for child thrads
  <carlo> threads
  <ropalka> carlo, Could I have a link for your ThreadLocalStack impl. to see how it looks like?
  <ropalka> carlo, right
  <carlo> https://github.com/jbossejb3/jboss-ejb3-context/blob/master/base/src/main/java/org/jboss/ejb3/context/util/ThreadLocalStack.java
• liweinan has quit (Read error: Connection reset by peer)
• ropalka 's having a look ...
  <ropalka> carlo, aha
  <ropalka> carlo, I see what U meant now
  <ropalka> carlo, U mean "STACK" behavior based on top of thread locals
  <carlo> Yes
  <carlo> A piece of code says a thousand words
  <ropalka> carlo, exactly. For me this way works the best
  <ropalka> carlo, Thanks for clarification!
  <ropalka> carlo, I don't think I'll need it for WS msg ctx injection. It should work IMHO in the way I proposed it
  <carlo> Yes, looks good
  <ropalka> carlo, Jaikiran Thanks for chat folks. I'm going to fix it