ant elder wrote:
The actual error cause was:
Caused by: java.lang.ClassNotFoundException:
org.apache.tuscany.sca.assembly.builder.ComponentPreProcessor
at
org.eclipse.osgi.framework.internal.core.BundleLoader.findClassInternal(BundleLoader.java:481)
but I guess thats just caused by the spring modules not being finished,
I saw you committed some updates to the spring manifests last night so
i've tried again now it it works ok fine today.
What was unexpected (to me anyway at the time) was that the
node-launcher-equinox tests were failing because it was discovering a
module that i didn't even have included in the build just happened to be
within the modules/ directory, so thats the bit i was asking if that was
the expected behaviour. I'm not saying its a problem, now that i know
this is whats going on i guess it wont be unexpected if it happens again.
...ant
Ant,
Thanks for the post.
The error you were getting was the one I was working on, so that's good - and yes, I fixed it. It
was due to an incomplete adjustment of the code and the manifest files to permit clean running under
OSGi - the code was working fine outside of OSGi.
The problem here is perhaps one we should worry about - code that you have no intention of using
gets "touched" and eagerly loaded. So you have no Spring stuff in your Tuscany app, but simply
because the Spring implementation module is there, declaring that it supports
<implementation.spring/>, it gets initialized by the core runtime.
I would like us to consider moving to a lazy loading scheme, where the code is only loaded and
initialized when it is found to be required.
Thoughts??
Yours, Mike.
