[
https://issues.apache.org/jira/browse/FELIX-29?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12601435#action_12601435
]
Richard S. Hall commented on FELIX-29:
--------------------------------------
I have committed a very preliminary step toward fragment support. The new patch
allows fragments to be used to extend the bundle class path of the host bundle
with content from the fragment. Nothing else is supported at this point (i.e.,
fragments cannot import, require, export, nor contain native code). Further,
fragments are not yet hooked into the update/refresh mechanisms of Package
Admin, so don't try to dynamically update and refresh them.
Currently, a fragment content is searched after the host bundle class path
according to the fragment's bundle class path. It is also possible for the host
bundle to reference fragment content on its own bundle class path to interpose
fragment content with its own.
My priorities are:
1. Get update/refresh working.
2. Get native code working.
3. Get import/require working.
4. Get export working.
At first blush, it appears that (4) will be the trickiest since it breaks some
assumptions in our resolving algorithm. I think the others should be reasonably
straightforward, but the devil is in the details. It might be the case that
between steps (3) and (4) some major refactoring will be done to clean things
up quite a bit, but if we can get 1-3 working well, then that should be a good
start.
> Implement bundle fragments
> --------------------------
>
> Key: FELIX-29
> URL: https://issues.apache.org/jira/browse/FELIX-29
> Project: Felix
> Issue Type: New Feature
> Components: Framework, Specification compliance
> Reporter: Richard S. Hall
> Assignee: Richard S. Hall
>
> This issue is described in section 3.14 of the OSGi R4 specification. The
> main jist is that a logical bundles can now be defined in more than one
> physical bundle. This means that fragment bundles logically merge with their
> host bundle at run time. This issue impacts the resolver algorithm and will
> likely have some impacts on the run-time class loading search order.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
