| I didn’t much like my solution either. Fortunately we were able to modify the bundle that introduced the use of javax.annotation: the javax.annotation import was a compile-time dependency, not a runtime dependency. So all is now well.
Thanks for your feedback!
Brian.
Yes it’s a very common misconception that the Framework always chooses the highest version it can find — I have to address this in every training course I run.
It *may* prefer the highest version if everything else about the resulting outcomes is equal. But that is an unrealistic assumption, and certainly not something that should be relied upon for proper functioning of your application.
Neil
Note that a version bump does not guarantee that it will be the provider of the package. The framework can heuristics to do this but these are not specified to allow optimizations we could not foresee.
Another idea to use is to put the messed up packages on the framework’s classpath. This will export them from the framework side. Since these are resolved from the start and the uses constraints match the framework it is extremely unlikely they are not picked.
You’re in a nice mess …
Kind regards,
Peter Kriens
On 17-Jun-2016, at 1:08 PM, Neil Bartlett < njbartlett@xxxxxxxxx> wrote: If you’re unable to change the bundles then no, not really. The only way is to not install these bundles that have exports of javax.annoation.
I ended up following the process we used with hollowing out javax.annotation: making a bundle that:
- includes the other bundles as jars
- “Require-Bundle: system.bundle” to pull in the JVM’s javax.annotation classes
- Import-Package of the various javax.annotation packages in the bundled jars in case they’re already around
- Export-Package of the various javax.annotation packages in the bundled jars with a micro version bump to beat any of the existing jars.
That seems to work.
Brian. _______________________________________________ equinox-dev mailing list equinox-dev@xxxxxxxxxxxTo change your delivery options, retrieve your password, or unsubscribe from this list, visit https://dev.eclipse.org/mailman/listinfo/equinox-dev
_______________________________________________ equinox-dev mailing list equinox-dev@xxxxxxxxxxxTo change your delivery options, retrieve your password, or unsubscribe from this list, visit https://dev.eclipse.org/mailman/listinfo/equinox-dev
_______________________________________________ equinox-dev mailing list equinox-dev@xxxxxxxxxxxTo change your delivery options, retrieve your password, or unsubscribe from this list, visit https://dev.eclipse.org/mailman/listinfo/equinox-dev
|
