Well, it turns out that what I did that made me encounter this problem in the first place (embedding the JAX-WS API classes within the bundle as opposed to using what was provided by the OSGI framework) was unnecessary for getting the bundle to work.
I only have access to the code for two days a week so I can't really justify experimenting with this anymore, at least for the time being. Glad to have helped find this oversight, in any case. On Tue, Feb 11, 2014 at 4:48 PM, David Bosschaert < [email protected]> wrote: > Hi, > > Just to let you know that this issue is now fixed on trunk. You should > be able to use either SPI-Consumer or Require-Capability headers with > static weaving. For more details see > https://issues.apache.org/jira/browse/ARIES-1156 and > https://issues.apache.org/jira/browse/ARIES-1157 > > Tovi, I would appreciate if you could check that it works in your > environment too. > > Cheers, > > David > > > On 5 February 2014 15:16, David Bosschaert <[email protected]> > wrote: > > On 5 February 2014 14:55, Tovi Almozlino <[email protected]> wrote: > >> I suppose I'll have to use the proprietary headers for now, unless you > guys > >> can have this fixed by next week! FYI, what I'm working on is a plugin > for > >> Cytoscape 3.0.x, which uses an old (pre-2.3) version of Karaf. > > > > I can have a look at it sometime in the near future, but if you need > > it earlier you can always propose a patch ;) > > > >> By the way, about providers: does the file in resources/services have to > >> point at an implementation class contained WITHIN the provider bundle? > Or > >> can it point at any class whatsoever? > > > > I'm pretty sure that the class can be any class that the bundle has > > visibility to, so that would include classes in the bundle, its > > fragments, packages imported via Import-Package and bundles wired via > > Require-Bundle... > > > > Cheers, > > > > David >
