Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [ajdt-dev] strange exception problem

Hi Andy,

ps. if you get a chance, I'd be interested in getting you to try out
the weaver in https://bugs.eclipse.org/bugs/show_bug.cgi?id=210470 -
since it is meant to be much better behaved in terms of allowing
itself to be gc'd when the related classloader is orphaned.

I tried this weaver version and it works fine in my environment. So I have tested only the general weaving functionality at the moment, not the gc of orphaned classloaders and my testing was not very extensive, but my equinox aspects examples work well with this weaver.

Great to see improvements in this area. I believe that this gc of classloaders becomes more important for me. During the past days I got equinox aspects to fully support bundles dynamics. Means you can install, update and uninstall aspect bundles at runtime and get the rest of the system updated automatically including re-weaving of affected bundles and so on. Really cool... :-) And for uninstalled bundles gc'ing the weaver sounds like a good thing to do... ;-)

BTW: While analyzing the cflow problem I quickly read through the CflowPointcut class and it contains some static stuff in it (some static collections that are cleared sometimes). I am not sure whether this is related to my current problem or not but couldn't that cause trouble when you have multiple separate weavers for different classloaders around?

Thanks again for your help. Will let you know when I find out more about the cflow field problem...

-Martin





2008/5/2 Andy Clement <andrew.clement@xxxxxxxxx>:
I can't think of a change, but there might be one.  You could check
 bugzilla for fixes committed between the dates of old and new (AspectJ
 build dates, not AJDT ones).

 Does it work if you let the aspects get woven?



 Andy.

 2008/5/2 Martin Lippert <lippert@xxxxxxx>:
 > Hi Andy,
 >
 >
 >
 > > I think it needs to weave the aspect where the cflow pointcut occurs -
 > > but I'd need to double check.
 > >
 > > I don't believe the behaviour has changed, the only thing I added
 > > recently was an error message saying "you need to weave the aspect"
 > > which comes out with the cryptic cflowcounter message.  if you aren't
 > > seeing that message then maybe your problem really is different.
 > >
 >
 >  Hm, this seems to be very interesting. I have a hierarchy of three aspects
 > where the cflow pointcut is defined in the base class of this hierarchy and
 > I decompiled all three classfiles after weaving. Noone contains a field
 > definition for the cflow. Seems to me that my case might be indeed a
 > different one. Could any other change to the weaver code possibly created
 > this problem? I am so puzzled because an older build worked fine...
 >
 >
 >
 >  -Martin
 >
 >
 >
 > >
 > > Andy
 > >
 > > 2008/5/2 Martin Lippert <lippert@xxxxxxx>:
 > >
 > > > Hi Andy,
 > > >
 > > >
 > > >
 > > >
 > > > > I've seen that if you exclude the aspect from being woven in the XML
 > > > > configuration for ltw - we must be able to weave the aspect if you are
 > > > > using cflow since it will have that field added to it.   See bugs like
 > > > > https://bugs.eclipse.org/bugs/show_bug.cgi?id=166647
 > > > >
 > > > > Are you including the aspect in what can be woven?
 > > > >
 > > > >
 > > >  Thanks for the quick pointer to the bug. My issue seems to be closely
 > > > related but also slightly different... ;-)
 > > >
 > > >  In my case I have an abstract aspect containing the cflow pointcut and
 > a
 > > > concrete aspect inheriting from that aspect (so no aop.xml-based
 > concretized
 > > > aspect). And in my setting the abstract aspect seems to be excluded from
 > > > weaving. Could that cause the problem? And has this behavior changed
 > really
 > > > between the builds I mentioned? I just wanna be sure where the problem
 > is to
 > > > find a good solution for my case... :-)
 > > >
 > > >  Thanks a lot for your help!
 > > >  -Martin
 > > >
 > > >
 > > >
 > > >
 > > >
 > > >
 > > >
 > > >
 > > >
 > > >
 > > > > Andy.
 > > > >
 > > > > 2008/5/2 Martin Lippert <lippert@xxxxxxx>:
 > > > >
 > > > >
 > > > > > Hi!
 > > > > >
 > > > > >  When using load-time weaving with AJDT 1.5.2 I get this exception:
 > > > > >
 > > > > >  java.lang.NoSuchFieldError: ajc$cflowCounter$0
 > > > > >  (invoking a woven class)
 > > > > >
 > > > > >  I tried to track this down. Seems like this problem was introduced
 > by
 > > > > >
 > > > >
 > > > build
 > > >
 > > > >
 > > > > > 1.5.2.200804031532 of AJDT. It worked well with the previous build
 > > > > > 1.5.2.200801291327.
 > > > > >
 > > > > >  Any idea what is going on?
 > > > > >
 > > > > >  Thanks in advance!
 > > > > >  -Martin
 > > > > >
 > > > > >  _______________________________________________
 > > > > >  ajdt-dev mailing list
 > > > > >  ajdt-dev@xxxxxxxxxxx
 > > > > >  https://dev.eclipse.org/mailman/listinfo/ajdt-dev
 > > > > >
 > > > > >
 > > > > >
 > > > > _______________________________________________
 > > > > ajdt-dev mailing list
 > > > > ajdt-dev@xxxxxxxxxxx
 > > > > https://dev.eclipse.org/mailman/listinfo/ajdt-dev
 > > > >
 > > > >
 > > > >
 > > > >
 > > > >
 > > >  _______________________________________________
 > > >  ajdt-dev mailing list
 > > >  ajdt-dev@xxxxxxxxxxx
 > > >  https://dev.eclipse.org/mailman/listinfo/ajdt-dev
 > > >
 > > >
 > > _______________________________________________
 > > ajdt-dev mailing list
 > > ajdt-dev@xxxxxxxxxxx
 > > https://dev.eclipse.org/mailman/listinfo/ajdt-dev
 > >
 > >
 > >
 >
 >  _______________________________________________
 >  ajdt-dev mailing list
 >  ajdt-dev@xxxxxxxxxxx
 >  https://dev.eclipse.org/mailman/listinfo/ajdt-dev
 >

_______________________________________________
ajdt-dev mailing list
ajdt-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/ajdt-dev






Back to the top