Well, theoretically a Map<CL, CDIProvider> might be more correct. But there is already a static (!) setter for it. So I fear there is not much we can do. Plus the CL trick is broken in most containers anyway for such low-level things. Think about EARs. You would only be able to set it for the main CL...
It's also not practical to have different CDI containers which share the same API via a parent ClassLoader. If you have multiple CDI containers then they are fully separated - means including the API. Otherwise you are bound to a very specific version anyway. And thus it makes not much sense. LieGrue, strub > Am 27.08.2017 um 19:41 schrieb Jean-Baptiste Onofré <[email protected]>: > > Agree with Romain. > > Regards > JB > On Aug 27, 2017, at 17:48, Romain Manni-Bucau <[email protected]> wrote: > I don't like much this impl since it enforces a single CDI impl and flat > classloader for the container but if really in the API we must, if just a > leak from the javadoc we don't have to. > > > Romain Manni-Bucau > @rmannibucau | Blog | Old Blog | Github | LinkedIn | JavaEE Factory > > 2017-08-27 18:31 GMT+02:00 John D. Ament <[email protected]>: > Hmmm so you're thinking we need to keep a map of classloader/CDIProvider > pairs? > > > On Sun, Aug 27, 2017 at 12:16 PM Romain Manni-Bucau < [email protected]> > wrote: > Doesn't break anything so +1 > > Now a few side note on it: > > 1. private static volatile CDI shouldn't be static IMO since it is broken in > any container (including SE if not under a flat classloading of the cdi impl) > 2. i'm not sure how the patch changes much in practise (= maybe it does worth > mentioning it in the class?) > 3. just realized we are not aligned on https://docs.jboss.org/cdi/api/2.0/ > "protected" behavior. This looks like an abuse of the javadoc "discovered > API" but wonder if we want to be aligned on the spec or make the spec cleaned > up > > If unclear: none of these point block a release but since we revisit this "to > be enhanced" area I thought it was important to mention them > > > Romain Manni-Bucau > @rmannibucau | Blog | Old Blog | Github | LinkedIn | JavaEE Factory > > 2017-08-27 14:22 GMT+02:00 John D. Ament <[email protected]>: > Hi > > I pushed up a small fix to the CDI spec. Would it make sense to cut a > release of that? > > John > >
