I reported two bugs about my problems with cycles from PDE. (246615, 246617)

ekke

…and report some bugs, but its some work to make it reproducable in a small test-environment and at the moment I have to finish some prototyping-work so my project can go on…

ekke

Indeed I'm also suffering using different repositories using different OSGi packaging. It's a pain in the ass for average developer to identify the problem in such situation.

I agree on the need of "best practices" between repositories

>I also found some problems with PDE where PDE detects cycles in 'foreign' bundles – but they are running perfect. but thats another story to blog about

I ran also thru this "cycle" problem with sl4j bundles. It cames out that one bundle was a fragment hosted by the other one.

from my sight I’m using these rules:

* if using a ‘common’ bundle from ‘foreign’ projects always to use import package
* using required bundle only if working in the own universe where you have total control over the bundles

I always try to use import bundle where possible. its some more work, but worth…

BTW: I’ll publish my experiences with this next week. its not only the required-bundle vs import-package problematic – there’s also the problem using other bundles where you find ‘core’ bundles with 30+ jarfiles inside (Easybeans OSGI)
logging is another theme….

I think we need some rules what a good OSGI developer should do and what not

otherwise its a very hard job to combine different OSGI projects. I also found some problems with PDE where PDE detects cycles in ‘foreign’ bundles – but they are running perfect. but thats another story to blog about

ekke