Re: Unsupported Operation Exception in feature resolution

[Ryan Moquin <fr...@gmail.com>]

Matteo,

Ironically, this same error had been plaguing me lately and so I turned on
debugging to see if I could get any details on what was happening.  This
resulted in me finding out what JB mentioned above about a bundle being
uninstalled.  The suppressed exception though was a collection not
modifiable exception (I forget the exact class).  My features result in a
refresh at runtime and that never caused a verification issue.

I have learned to temporarily skip verification when I run into any issue
with it (either that exception or if it fails after running out of
memory).  I then load my feature into karaf which will give me details of
what is wrong.

In this case, there were some missing requirements trying to resolve a
feature that was marked as a prerequisite.  After fixing that, I still got
the same error during validation.  I decided to try removing the
prerequisite attributes I was experimenting with and got everything to work
again.

It can be really hard to get prerequisite features working correctly unless
you are sure everything is perfectly correct.

When you hit issues, try your features in karaf and install small pieces at
a time which will let you narrow down the issue.

Hopefully this is helpful.

Ryan

On Fri, Jul 26, 2019, 10:33 AM Matteo Rulli <ma...@gmail.com> wrote:

> I see, thank you for the explanation and for adding a more permissive
> behaviour to the karaf-maven-plugin.
>
> Just to know a little bit better how to control the refreshes, in general
> my understanding is the following
>
>    - <bundle>/<feature> directives will install the bundle or the
>    feature, no matter what is already installed
>    - dependency=“true" will install the bundle/feature if and only if it
>    is required by the resolver and it is not already installed
>    - prerequisite=“true" will guarantee the prerequisite feature will be
>    installed before the other stuff in the feature declaration
>
> Is this correct? Is there any other attribute that can be used to control
> feature-based provisioning?
>
> Thank you very much,
> Matteo
>
> On 26 Jul 2019, at 13:12, Jean-Baptiste Onofré <jb...@nanthrax.net> wrote:
>
> Actually, your features is causing a refresh (the flairkit-alarm-rest
> for instance).
>
> So, the features resolver tries to uninstall a bundle to reinstall it.
>
> It happens here:
>
>
> https://github.com/apache/karaf/blob/karaf-4.2.x/features/core/src/main/java/org/apache/karaf/features/internal/service/StaticInstallSupport.java#L50
>
> So, your feature will could work at runtime, but with a refresh.
>
> I would recommend to fix your bundles to avoid the refresh.
>
> Anyway, instead of failing, I can log a WARN message with the bundles in
> cause.
>
> I created KARAF-6370 about that.
>
> Regards
> JB
>
> On 26/07/2019 11:40, Matteo Rulli wrote:
>
> Hello,
>
> Thanks. Yes we use the verify step of the karat-maven-plugin. Does this
> mean that the “verify” step can fail at build-time but the feature will
> work in a real deployment? If this is the case, what is the right approach
> to verify a feature before it is used or deployed?
>
> Thanks,
> Matteo
>
> On 26 Jul 2019, at 11:31, Jean-Baptiste Onofré <jb...@nanthrax.net> wrote:
>
> Hi,
>
> I guess you have this with karaf-maven-plugin, right ?
>
> The reason is that we use a "simplified" resolver in the plugin, not
> implementing all methods. That's why you have
> UnsupportedOperationException.
>
> Regards
> JB
>
> On 26/07/2019 11:05, Matteo Rulli wrote:
>
> Hello,
>
> Karaf feature resolution fails due to
> java.lang.UnsupportedOperationException. It is hard to understand the
> root cause.
>
> I managed to reproduce the issue in a relatively simple settings, you
> can find it here:
> https://github.com/mrulli/karaf-426-feature-unsupported-op
>
> Could you please help me to understand what is it happening here?
>
> Thanks,
> Matteo
>
>
> --
> Jean-Baptiste Onofré
> jbonofre@apache.org
> http://blog.nanthrax.net
> Talend - http://www.talend.com
>
>
>
> --
> Jean-Baptiste Onofré
> jbonofre@apache.org
> http://blog.nanthrax.net
> Talend - http://www.talend.com
>
>
>