On 07/10/15 18:01, Neil Bartlett wrote:
On 7 Oct 2015, at 13:56, Ferry Huberts <maili...@hupie.com> wrote:

Hi,

Cross-posting because I don't know if the situation/problem below is by design, 
a bug in bnd, or a bug in SCR


I have a R6 component that does
  @Reference(cardinality = ReferenceCardinality.AT_LEAST_ONE)

This appears to not bind new matching services, because the policy appears to 
be STATIC by default.
Yes, static policy has always been the default in DS. You can see this right back in 
the very first DS specification found in OSGi Compendium R4.0, section 112.3.3: 
"The static policy is the most simple policy and is the default policy”. Also 
section 112.10 that gives the XML schema, where the reference element has the policy 
attribute defined as:

     <attribute name="policy" type="scr:Tpolicy" default="static" 
use="optional"/>


I did expect the ReferenceCardinality.AT_LEAST_ONE to imply 
ReferencePolicy.DYNAMIC, just like the bnd annotations did.
Nope. These are new annotations in a new namespace. There is no reason to 
expect them to have the same behaviour as the bnd annotations.

However, I challenge you to give me _one_ use-case where ReferencePolicy.STATIC && ReferencePolicy.(MULTIPLE|AT_LEAST_ONE) makes sense...

That combination _creates_ a timing dependency, services might or might not be bound, depending on startup order.

A warning in bndtools like 'you probably meant to use ReferencePolicy.DYNAMIC' when havingReferencePolicy.MULTIPLE or ReferencePolicy.AT_LEAST_ONE would be nice IMHO


Changing it to
  @Reference(policy = ReferencePolicy.DYNAMIC, cardinality = 
ReferenceCardinality.AT_LEAST_ONE)

appears to make it work like I expected.


Is this by design?

Coming from bnd annotations this - to me - seems like another (breaking) 
subtlety.
It’s a subtlety perhaps because the standard annotations are just a direct, 
literal transformation to the XML, whereas the bnd ones tried to be “smart”. 
Yes, this is by design.

It’s certainly not breaking because, again, these are new annotations and there 
is no reason to expect them to do the same thing…


--
Ferry Huberts

--
You received this message because you are subscribed to the Google Groups 
"bndtools-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to bndtools-users+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--
Ferry Huberts

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscr...@felix.apache.org
For additional commands, e-mail: users-h...@felix.apache.org

Reply via email to