10 a.m.
Because of this change
http://fisheye.jboss.org/browse/JBossAS/trunk/server/src/main/java/org/jb...
annotation scanning doesn't work anymore and thus EJBs (in wars) won't
get deployed.
Suggestions?
Carlo
10:26 a.m.
On Thu, Nov 26, 2009 at 5:00 PM, Carlo de Wolf <cdewolf(a)redhat.com> wrote:
Because of this change
http://fisheye.jboss.org/browse/JBossAS/trunk/server/src/main/java/org/jb...
annotation scanning doesn't work anymore and thus EJBs (in wars) won't
get deployed.
Suggestions?
Oooops. At the moment annotations that are processed by
Web30MetaDataCreator need to get special treatment in a different
deployer, so this was a (bad) attempt to avoid doing things twice.
Fixed.
I have the impression that the other annotations will need to be
processed per JAR eventually, but this is not urgent ...
Rémy
10:45 a.m.
On 11/26/2009 05:26 PM, Rémy Maucherat wrote:
On Thu, Nov 26, 2009 at 5:00 PM, Carlo de
Wolf<cdewolf(a)redhat.com> wrote:
> Because of this change
>
http://fisheye.jboss.org/browse/JBossAS/trunk/server/src/main/java/org/jb...
> annotation scanning doesn't work anymore and thus EJBs (in wars) won't
> get deployed.
>
> Suggestions?
Oooops. At the moment annotations that are processed by
Web30MetaDataCreator need to get special treatment in a different
deployer, so this was a (bad) attempt to avoid doing things twice.
Fixed.
I have the impression that the other annotations will need to be
processed per JAR eventually, but this is not urgent ...
Hmm we also need to make sure that you can use the AnnotationEnvironment
(or the outstanding Papaki annotation scanning thingy) instead of the
AnnotatedClassFilter in future. AFAIK the later one was much slower. It
also means that the classpath is scanned twice for annotations (first
from AE and then from the AnnotatedClassFilter) which should be avoided.
I guess you cannot use the AnnotationEnvironment for now since you need
scanning the reference to the web-fragment/jar?
10:54 a.m.
On Thu, Nov 26, 2009 at 5:45 PM, Emanuel Muckenhuber
<emuckenh(a)redhat.com> wrote:
Hmm we also need to make sure that you can use the
AnnotationEnvironment
(or the outstanding Papaki annotation scanning thingy) instead of the
AnnotatedClassFilter in future. AFAIK the later one was much slower. It
also means that the classpath is scanned twice for annotations (first
from AE and then from the AnnotatedClassFilter) which should be avoided.
I guess you cannot use the AnnotationEnvironment for now since you need
scanning the reference to the web-fragment/jar?
Yes, scanning per JAR is needed, I did put it in the requirements wiki
for future annotation work.
Right now, it's either:
1) use the slow AnnotationMetaDataDeployer everywhere
2) use WarAnnotationMetaDataDeployer for web stuff and
OptAnnotationMetaDataDeployer elsewhere
Solution 2) is used.
For Carlo: You can update OptAnnotationMetaDataDeployer as you need,
web annotations processing is not done there. Your change looks ok to
me.
Rémy
10:31 a.m.
The AnnotationMetaDataDeployer should be split up anyway. Seems Remy did
the first step.
Was this really working before? As far as i can remember we always
skipped ejb3 processing when there was a WebMetaData present?
On 11/26/2009 05:00 PM, Carlo de Wolf wrote:
Because of this change
http://fisheye.jboss.org/browse/JBossAS/trunk/server/src/main/java/org/jb...
annotation scanning doesn't work anymore and thus EJBs (in wars) won't
get deployed.
Suggestions?
Carlo
_______________________________________________
jboss-development mailing list
jboss-development(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/jboss-development
10:41 a.m.
On Thu, Nov 26, 2009 at 5:31 PM, Emanuel Muckenhuber
<emuckenh(a)redhat.com> wrote:
The AnnotationMetaDataDeployer should be split up anyway. Seems Remy
did
the first step.
Was this really working before? As far as i can remember we always
skipped ejb3 processing when there was a WebMetaData present?
I did remove the "skip if it's a war" test, but indeed, there is a
test like that:
if(webMetaData != null)
{
processJBossWebMetaData(unit, finder);
}
else if (clientMetaData != null)
{
String mainClassName = getMainClassName(unit);
if(mainClassName != null)
processJBossClientMetaData(unit, finder, mainClassName);
}
else
{
String mainClassName = getMainClassName(unit);
if (mainClassName != null)
processJBossClientMetaData(unit, finder, mainClassName);
else
processJBossMetaData(unit, finder);
}
Rémy
10:50 a.m.
At some point in AS 5.0 we did do EJBs in wars (
https://jira.jboss.org/jira/browse/JBAS-5281 ), but that lead to
problems were people would actually package EJBs in wars. :-)
In AS 5.1 it doesn't work ( https://jira.jboss.org/jira/browse/JBAS-7312
), but that's because we needed to forcibly disable it and thus it
wasn't tested anymore.
Now for AS 6 we need it again for EJB 3.1 spec. :-)
I've locally modified that to:
protected void processMetaData(VFSDeploymentUnit unit, WebMetaData webMetaData,
ApplicationClientMetaData clientMetaData, List<VirtualFile> classpath) throws
Exception
{
AnnotationFinder<AnnotatedElement> finder = new
DefaultAnnotationFinder<AnnotatedElement>();
// don't do webMetaData anymore (rev 96033)
String mainClassName = getMainClassName(unit);
if(mainClassName != null)
processJBossClientMetaData(unit, finder, mainClassName);
else
processJBossMetaData(unit, finder);
}
But yes, these concerns need to be separated out.
Carlo
On 11/26/2009 05:41 PM, Rémy Maucherat wrote:
On Thu, Nov 26, 2009 at 5:31 PM, Emanuel Muckenhuber
<emuckenh(a)redhat.com> wrote:
> The AnnotationMetaDataDeployer should be split up anyway. Seems Remy did
> the first step.
>
> Was this really working before? As far as i can remember we always
> skipped ejb3 processing when there was a WebMetaData present?
>
I did remove the "skip if it's a war" test, but indeed, there is a
test like that:
if(webMetaData != null)
{
processJBossWebMetaData(unit, finder);
}
else if (clientMetaData != null)
{
String mainClassName = getMainClassName(unit);
if(mainClassName != null)
processJBossClientMetaData(unit, finder, mainClassName);
}
else
{
String mainClassName = getMainClassName(unit);
if (mainClassName != null)
processJBossClientMetaData(unit, finder, mainClassName);
else
processJBossMetaData(unit, finder);
}
Rémy
_______________________________________________
jboss-development mailing list
jboss-development(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/jboss-development
4938
days inactive
4938
days old
jboss-development@lists.jboss.org
6 comments
3 participants
participants (3)
-
Carlo de Wolf -
Emanuel Muckenhuber -
Rémy Maucherat