[
Date Prev][
Date Next][
Thread Prev][
Thread Next][
Date Index][
Thread Index]
[
List Home]
Re: [aspectj-users] java.lang.VerifyError: stack shape inconsistent
|
OK, seems like things are now working in my MAVEN BUILD ONLY after collapsing the abstract class into the concrete impls. Doing this in RAD7/Eclipse (AJDT v1.4.2.200705221209 - newest I can go to using RADv7 built on Eclipse 3.2) failed with the same error on a different class (as mentioned previously).
hmmm.... Could this actually be a bug in the AJDT version I'm using!? I hope not, since I can't really upgrade! I will say this is something else that has changed since the last time this worked. However, I can't easily roll this plugin version back without also rolling back my version of SpringIDE (which depends on this version of AJDT I think...)....
On 4/30/08, Bryan Jones <bjones@xxxxxxxxxxxxxxxxxxxx> wrote:
I commented out a bunch of code trying to refactor the code that was causing the issue before. I now was able to get the same error to happen on another class. This is more information at least... it does not seem to be isolated to one type. It does seem to be an issue related to inheritance... the errors come from the abstract class being extended. I got rid of the previous errors by temporarily collapsing the abstract class into the concrete impl, and the problem went away there. Now I'm getting the problem on another abstract class... ????
It does seem to me that the aspect I'm using, which has been working fine in production for almost 2 years, seems to not be creating valid joinpoints anymore. I wouldn't think so, but is it possible that the move to aspectj 1.5.4 changed the semantics of basic pointcuts and advice?? I wouldn't think so....
I'm still looking at things closer.
On 4/30/08, Bryan Jones <bjones@xxxxxxxxxxxxxxxxxxxx> wrote:
I was able to recreate the issue with the Sun JRE as well...
So, this seem like it could this be an AspectJ bug. I'm going to try to isolate the issue better.
Thanks,
Bryan
On 4/30/08, Bryan Jones <bjones@xxxxxxxxxxxxxxxxxxxx> wrote:
Yes, I'm now running with IBM J9 1.5 SR6b (previously running with SR2):
" java version "1.5.0"
Java(TM) 2 Runtime Environment, Standard Edition (build pwi32devifx-20071025 (SR6b))
IBM J9 VM (build 2.3, J2RE 1.5.0 IBM J9 2.3 Windows 2000 x86-32 j9vmwi3223-20071007 (JIT enabled)
J9VM - 20071004_14218_lHdSMR
JIT - 20070820_1846ifx1_r8
GC - 200708_10)
JCL - 20071025"
I will try to run against the Sun JRE to see if the issue still exists.
On 4/30/08, Andy Clement <andrew.clement@xxxxxxxxx> wrote:
So you are now on IBM J9 1.5 SR6?
Are you able to try a one off test with a Sun vm to see if the problem
still occurs? If so, it is obviously more likely to be an AJ bug than
a VM issue.
Can you temporarily exclude the problematic method?
Andy.
2008/4/30 Bryan Jones <bjones@xxxxxxxxxxxxxxxxxxxx>:
> No luck! The upgrade of the IBM J9 JRE did not fix the issue! It may be a
> real pain for me to revert to the previous version of AspectJ, with no
> guarantee that the revert will even fix this issue...
>
> I guessing it could be an issue in my environment somehow, which is why I'll
> probably hold off on opening a bug at this point.
>
> I'm at a loss at this point.
>
> Any other thoughts or suggestions!?
>
> Thanks,
> Bryan
>
>
>
>
>
> On 4/29/08, Andy Clement <andrew.clement@xxxxxxxxx> wrote:
> > Hmmmm - right now I can't think why moving from a runtime of 1.5.3 to
> > a runtime of 1.5.4 should make a difference, very very little ever
> > changes in the runtime jar. I can possibly imagine moving from the
> > 1.5.3 weaver to a 1.5.4 weaver could cause some kind of issue like
> > this to surface, but I don't know which change in AspectJ would cause
> > the IBM VM to start having problems, I would need to look through all
> > the issues resolved in 1.5.4 to see if there are any likely
> > candidates. If it isn't straightforward for you to pick up the latest
> > level of IBM J9 then by all means revert to aspectjrt 1.5.3 if that
> > gets you going - I just can't explain why that should at the moment :)
> >
> > Andy.
> >
> > 2008/4/29 Bryan Jones <bjones@xxxxxxxxxxxxxxxxxxxx>:
> > > Thanks Andy. That's hopefully it. We are in fact using an IBM J9 VM.
> > >
> > > Things were all fine with this project in the past though... but I think
> it
> > > was running with AspectJ RT 1.5.3. I think it may have been the
> > > introduction of aspectjrt 1.5.4 that caused this problem to surface.
> Does
> > > this make sense to you that this could / would happen? If so, I would
> > > consider rolling back the aspectjrt upgrade, in order to resolve this
> issue
> > > in the shorter term, as there are other more pressing issues on my
> plate.
> > >
> > > Thanks,
> > > Bryan
> > >
> > >
> > >
> > >
> > > On Tue, Apr 29, 2008 at 1:32 PM, Andy Clement <andrew.clement@xxxxxxxxx>
> > > wrote:
> > > > Are you by any chance running on the IBM J9 VM? If so, are you on the
> > > > latest? For example, see
> > > >
> > >
> http://www-1.ibm.com/support/docview.wss?rs=2344&context=SS3PGL&dc=DB520&dc=DB560&uid=swg21298148&loc=en_US&cs=UTF-8&lang=en&rss=ct2344tivoli
> > > > which describes how after weaving a system with ITCAM (this happens to
> > > > use AspectJ inside) you then cause the VM to throw a stack shape
> > > > inconsistent problem and the fix is to upgrade the VM.
> > > >
> > > > I don't believe 'stack shape inconsistent' is due to a mismatch in
> > > > weaver/compiler versions used. We have no open bugs on that error.
> > > > The only similar verify issues I have seen are relating to incorrect
> > > > stack height - and they are usually because CGLIB generated code is
> > > > being woven.
> > > >
> > > > If it isn't a VM issue, then feel free to raise a bug and we can
> > > > investigate further.
> > > >
> > > > cheers,
> > > > Andy.
> > > >
> > > > 2008/4/29 Bryan Jones <bjones@xxxxxxxxxxxxxxxxxxxx>:
> > > >
> > > >
> > > >
> > > > > I'm receiving a 'VerifyError,' which I understand is typically a
> sign of
> > > 'a
> > > > > bug' but I wanted to float this out there first, before reporting
> > > anything.
> > > > >
> > > > > The error I'm receiving at runtime is:
> > > > > ..."Initialization of bean failed; nested exception is
> > > > > java.lang.VerifyError: stack shape inconsistent (class: [hidden]
> method:
> > > > >
> [hidden]_aroundBody0(L[hidden];LAbstract[hidden];Ljava/lang/String;)V)
> > > at
> > > > > pc: 2"
> > > > >
> > > > >
> > > > >
> > > > > I have seen this sort of error at run time when there are aspectjrt
> and
> > > / or
> > > > > weaver version related issues. I.e. compiled against a different
> > > version of
> > > > > aspectj then running against. I may be remembering incorrectly on
> this,
> > > but
> > > > > I believe those issues were cause by using a version of
> > > > > "aspectj-maven-plugin" compiling and weaving using one version of
> > > aspectj,
> > > > > while running the application using a different (seemingly
> incompatible
> > > > > version) of aspectj. The conflict in the past was really specific
> to
> > > server
> > > > > runtime and eclipse AJDT - i.e. deployed app and AJDT were using
> > > different
> > > > > and incompatible AspectJ versions. This can get quite painful...
> but
> > > those
> > > > > problems have been worked out in the past.
> > > > >
> > > > > Now we're getting similar issues, but even outside of Eclipse /
> AJDT...
> > > > > rather, now the issue is manifesting itself in a Maven build. I've
> > > taken
> > > > > care to ensure that a single version of AspectJ is ever used - so
> the
> > > same
> > > > > version is compiled with using "aspectj-maven-plugin" and ran
> against
> > > via
> > > > > the appropriate POM dependency. Prior to doing this, I was
> receiving
> > > > > warnings in the maven build as follows: "[WARNING] bad version
> number
> > > found
> > > > > in C:\dev\.m2\repository\aspectj\aspectjrt\1.5.3\aspectjrt-1.5.3.jar
> > > > > expected 1.5.4 found 1.5.3". So, after cleaning things up, this
> warning
> > > > > went away, and I was REALLY hoping that would have solved the issue.
> No
> > > > > such luck!
> > > > >
> > > > > Has anyone ever seen this error before - specifically the "stack
> shape
> > > > > inconsistent" portion? I think this is the key part of the error
> > > message,
> > > > > but I could be wrong. Another related detail here that _could_ be
> of
> > > > > importance is the error is actually happening on a class that is in
> > > another
> > > > > internal 'commons' jar - which is NOT compiled with AspectJ
> (standard
> > > > > compiler). Could this be causing the issue somehow?
> > > > >
> > > > > Thanks,
> > > > > Bryan
> > > > >
> > > > >
> > > > > --
> > > > > Bryan Jones
> > > > > http://www.chariotsolutions.com
> > > > > cell: 571.344.0698
> > > > > _______________________________________________
> > > > > aspectj-users mailing list
> > > > > aspectj-users@xxxxxxxxxxx
> > > > > https://dev.eclipse.org/mailman/listinfo/aspectj-users
> > > > >
> > > > >
> > > > _______________________________________________
> > > > aspectj-users mailing list
> > > > aspectj-users@xxxxxxxxxxx
> > > > https://dev.eclipse.org/mailman/listinfo/aspectj-users
> > > >
> > >
> > >
> > >
> > > --
> > >
> > > Bryan Jones
> > > http://www.chariotsolutions.com
> > > cell: 571.344.0698
> > > _______________________________________________
> > > aspectj-users mailing list
> > > aspectj-users@xxxxxxxxxxx
> > > https://dev.eclipse.org/mailman/listinfo/aspectj-users
> > >
> > >
> > _______________________________________________
> > aspectj-users mailing list
> > aspectj-users@xxxxxxxxxxx
> > https://dev.eclipse.org/mailman/listinfo/aspectj-users
> >
>
>
>
> --
> Bryan Jones
> http://www.chariotsolutions.com
> cell: 571.344.0698
> _______________________________________________
> aspectj-users mailing list
> aspectj-users@xxxxxxxxxxx
> https://dev.eclipse.org/mailman/listinfo/aspectj-users
>
>
_______________________________________________
aspectj-users mailing list
aspectj-users@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/aspectj-users
--
--
--
--
Bryan Jones
http://www.chariotsolutions.comcell: 571.344.0698