In this case I would say that the Knopflerfish behavior is the most correct behavior followed by Equinox. There was a clarification to the specification recently that mandated that when resolving a bundle the Framework must not allow a bundle to get wired to a package which is exported from an uninstalled bundle. It sounds like Felix still allows bundles to resolve to packages exported by uninstalled bundles.
Technically Equinox does not break the contract of the specification
because the refreshPackages method allows for a great range of freedom when
it comes to deciding which bundles belong to the graph of bundles that get
refreshed. Note that one acceptable implementation is to restart the
Framework which would essentially restart all bundles. In the Equinox case
a call to refreshPackages pulls in any uninstalled bundles that are removal
pending state, in this case Bundle A. This forces both Bundle B and Bundle
C to be refreshed and unresolved.
Tom
From: "Alin Dreghiciu" <[EMAIL PROTECTED]>
To: "OSGi Alliance - General" <[email protected]>
Date: 07/29/2008 03:35 AM
Subject: [osgi-dev] Refresh package behavior
Hi guys,
I stumbled yesterday on different behavior while refreshing packages
on the 3 OS framework implementations Felix/Equinox/Knopflerfish and
I'm wondering which one do you think is according to the specs.
So, I have:
1. bundle A exporting packages P1 and P2
2. Bundle B importing package P1
3. Bundle C importing package P2
4. Bundles A,B,C are started (so resolved)
Then I uninstall bundle A. As I expected bundle B and C remain started
so packages P1 and P2 are resolved and wired to bundle A.
Then I do a refresh on bundle B. After this the state of the bundles B
and C is different on the 3 implementations:
Felix: bundle B and C will still be resolved and started.
Equinox: both bundles B and C will be in installed state and cannot be
resolved (missing imports for P1 and P2)
Knopflerfish: bundle B will in installed state and cannot be resolved
(missing import for P1) and bundle B will be in started state (so it
is resolved)
So, which one is the correct behavior. Note that I used the shell
command (refresh) available in all of this implementations not the
package admin refresh packages so it can be that the refresh command
did different things not only call the package admin refreshPackages.
This fact I did not check.
Regards,
--
Alin Dreghiciu
http://www.ops4j.org - New Energy for OSS Communities - Open
Participation Software.
http://www.qi4j.org - New Energy for Java - Domain Driven Development.
http://malaysia.jayway.net - New Energy for Projects - Great People
working on Great Projects at Great Places
_______________________________________________
OSGi Developer Mail List
[email protected]
https://mail.osgi.org/mailman/listinfo/osgi-dev
<<inline: graycol.gif>>
<<inline: ecblank.gif>>
_______________________________________________ OSGi Developer Mail List [email protected] https://mail.osgi.org/mailman/listinfo/osgi-dev
