[
https://issues.apache.org/jira/browse/FELIX-3587?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13408848#comment-13408848
]
Sahoo commented on FELIX-3587:
------------------------------
Thanks for your comments, Richards. I think a resolver can't always detect the
need for a fragment. So, resolving framework extensions eagerly is a desired
thing, but it should not always be done. I will try to state my understanding
and you can comment on it. This is my understanding:
Felix does not allow fragments to be attached to regular RESOLVED bundles
without requiring the host to be updated for very good reasons. As a special
case, it allows system bundle fragments to be attached without requiring the
system bundle to be updated so that one can create a framework and install
framework extensions and those extensions will take immediate effect. This is a
good thing, but it can't always be allowed as it can lead to the kind of issues
mentioned in the bug description. So the framework has to be more selective
about resolving framework extensions. e.g.:
a) Allow system bundle fragments to be attached only if they are installed when
the framework is in STARTING state.
b) Allow system bundle fragments to be attached only if they are exporting
packages which are not already in use.
#b is superset of #a, but #b might require more effort to implement.
> Uses constraint violation due to framework extension bundle
> -----------------------------------------------------------
>
> Key: FELIX-3587
> URL: https://issues.apache.org/jira/browse/FELIX-3587
> Project: Felix
> Issue Type: Bug
> Components: Framework
> Reporter: Sahoo
>
> There seems to be an issue with framework extension bundle and I have not
> checked the spec yet. Felix allows extension bundle to be attached at
> runtime, but that can sometimes lead to uses constraint violation if a
> package exported by the framework extension is also exported by some other
> regular bundle and that package from regular bundle is in use. A
> refreshPackages solves the issue, but that seems to be putting the onus on
> user. Can the framework be smart enough to not attach an extension bundle to
> host when it sees that an extension is exporting a package which is already
> in use?
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
