[
https://issues.apache.org/jira/browse/FELIX-2560?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12919073#action_12919073
]
Richard S. Hall commented on FELIX-2560:
----------------------------------------
After looking into this a little bit more, I don't think it would necessarily
have to impact the bundle cache in any structural way. The main issue is to not
reset the revision-id on a refresh.
Normally, a revision directory is named something like version0.0 and after an
update a new revision directory is created called version0.1. The first number
is a refresh counter, while the second number is the revision-id. During a
refresh, the version0.0 directory is deleted and the version0.1 directory is
renamed to version1.0.
So, the simple solution is to keep the revision-id as just an increasing number
that doesn't reset. Instead of renaming the above current revision directory to
version1.0 on a refresh, we'd rename it to version1.1. Likewise if there were
two more updates, followed by a refresh, then the current revision directory
would be version2.3, since there would have been three updates and two
refreshes total.
> Bundle URLs do not survive refreshes
> ------------------------------------
>
> Key: FELIX-2560
> URL: https://issues.apache.org/jira/browse/FELIX-2560
> Project: Felix
> Issue Type: Improvement
> Components: Framework
> Affects Versions: framework-3.0.2
> Reporter: Richard S. Hall
> Priority: Minor
> Fix For: framework-3.2.0
>
>
> Currently, bundle URLs are encoded like this:
> bundle:// <bundle-id> . <revision-id> : <class-path-idx>/path/to/resource
> This is convenient to access our internal data structures and survives
> toString() round-tripping, but it does have the disadvantage that it doesn't
> survive a module being refreshed. The <revision-id> of a module gets reset
> after a refresh, which makes existing URLs become invalid. If <revision-id>
> were something like a <revision-name> that didn't get reset, then it would
> continue to work.
> This is clearly a corner case and not super important, but we could
> investigate making such a change. It may impact the bundle cache too. This is
> also not spec-mandated behavior, so bundles shouldn't really depend on this
> behavior either.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
