Thanks for the info Raymond
snip...
> 1) The maven plug finds the dependencies and use them as the list of jars
> (bundles) to be installed to the Equinox runtime. We also package the main
> and test code into a bundle so that all the classes (main/test and
> tuscany/3rd-party) are loaded by OSGi. Typically the current project will
> have dependencies on the SCA API/annotations or Tuscany API, the maven
> surefire plugin will create a URL classloader and it makes the main/test
> classes get the SCA/Tuscany API classes from JSE instead of OSGi.
>
> 2) I assume most of our Junit test cases start/stop Node using the
> Tuscany Node APIs. The plugin approach makes it possible to run the test
> cases under both JSE and OSGi without changing the code.
>
What I was getting at in my initial question is that this sounds a lot like
the launcher code we have + the code that wraps up the module and the test
and + the code that fires up junit. So it sounds like we are describing a
set of launcher functions. The current base function being
Launcher (for command line)
Get dependencies and create OSGi runtime
Get contributions and start node
And extending this you are creating another launcher to handle a specific
environment
Launcher (for junit)
Bundle up module and tests
Fire up JUnit
It just so happens you have wrapped all this up in a maven plugin. The
Junit launcher function could be useful from the command line or from
Eclipse. So I think this is good input to our launcher discussion
I'm not particularly proposing to change anything immediately but lets see
what happens on the other discussion. I'd rather have a coherent launcher
story and a maven plugin that exploits it rather thant some specific
function buried in the maven plugin mojo. That's all I was getting at.
Regards
Simon
