I should be more careful with my language, yes of course I'm talking about stopping a bundle.
Jürgen On 14.04.2014 08:46, Marcel Offermans wrote: > Hello Jürgen, > > Just to make sure: a service is unregistered when the bundle that > registered it is *stopped*, so not on shutdown (of the framework). > > Other than that, performing explicit cleanup in your code makes > perfect sense, and OSGi defines an explicit life cycle for those cases. > > Greetings, Marcel > > > On 14 Apr 2014, at 8:06 , Jürgen Rose <[email protected] > <mailto:[email protected]>> wrote: > >> I'm just a bit paranoid about disposing services correctly. Of course >> you are right about the automatic deregistration on shutdown. Its >> just out of habit that I deregister every service manually, but then >> in most of our code there is actually some dispose method which has >> to be called, so it is a necessity (mostly executors which have to be >> shut down). >> >> Jürgen >> >> Am 13.04.2014 11:50, schrieb Marcel Offermans: >>> On 13 Apr 2014, at 10:02 , Neil Bartlett <[email protected] >>> <mailto:[email protected]>> wrote: >>> >>>> On Sat, Apr 12, 2014 at 11:05 PM, Jürgen >>>> Rose <[email protected] <mailto:[email protected]>> wrote: >>>> >>>> True, but a lot of projects (Eclipse since the move to OSGi, >>>> Apache since Felix/Karaf etc became more prominent, Spring, >>>> ...) have now jars which have correct OSGi meta infos already. >>>> >>>> Dave Cramer from the postgres JDBC driver encouraged me to send >>>> a pull request and he will merge it onto the current branch. So >>>> we can use the JDBC driver directly in the future. I'm just >>>> waiting for a answer here on this list how to handle the >>>> properties correctly. >>>> >>>> I didn't talk about the DataSource, the OSGi service doesn't >>>> get unregistered, which is at least bad style. >>>> >>>> >>>> Can you clarify what you mean by this? The DataSourceFactory >>>> service doesn't get unregistered when the bundle stops? That's >>>> actually impossible... OSGi services are automatically unregistered >>>> when the publishing bundle stops. >>> >>> Unless, by accident or design, a bundle registers a service with the >>> BundleContext of a *different* bundle. So far that sounds unlikely >>> though. >>> >>> Greetings, Marcel >>> >>> >>> >>> _______________________________________________ >>> OSGi Developer Mail List >>> [email protected] >>> https://mail.osgi.org/mailman/listinfo/osgi-dev >> >> _______________________________________________ >> OSGi Developer Mail List >> [email protected] <mailto:[email protected]> >> https://mail.osgi.org/mailman/listinfo/osgi-dev > > > > _______________________________________________ > OSGi Developer Mail List > [email protected] > https://mail.osgi.org/mailman/listinfo/osgi-dev
_______________________________________________ OSGi Developer Mail List [email protected] https://mail.osgi.org/mailman/listinfo/osgi-dev
