I agree with nearly everything you (Gareth) have said on this thread. I see you also suggested returning 3 strings from the parse method for the same reason I did :-) I was objecting to: - calling the generated karat-specific id <foo>-<bar> , where one of foo and bar is the factoryPid and the other is a user supplied id, a pid (or a factory pid). For a factory configuration the only pid is generated by config admin and the format is implementation specific. - confusing pid and factoryPid. It's a little hard to talk about…. however a factoryPid is never a pid, and the pid from calling configAdmin.createFactoryConfiguration(factoryPid, null) is not a factoryPid.
I think the original code had the correct effect but is hopelessly confusing because of not making these distinctions properly. thanks david jencks On Oct 7, 2014, at 8:35 PM, Gareth <[email protected]> wrote: > Hello John, > > It looks like it is fixed on the 3.0.x branch: > > https://github.com/apache/karaf/blob/karaf-3.0.x/features/core/src/main/java/org/apache/karaf/features/internal/FeatureConfigInstaller.java > > Guillaume reverted the offending change on July 11: > > https://github.com/apache/karaf/commit/c31edca04692d8127989acbcdadc8157d1390f2a > > The same revert was done on master but Achim appears to have remade the > change here: > > https://github.com/apache/karaf/commit/21089127c89ddae275d495607c422dda8ffec474 > > Achim - was that intentional? > > I am not sure if David was referring to me in some way when he said "there > is a bunch of misinformation in some of the comments on this thread". I > don't really care about the philosophical reasons for one particular factory > pid format or the other. I just know with this change my > ManagedServiceFactories and Declarative Service components could not find > their factory configurations anymore...which thus made it impossible for me > at least to upgrade to Karaf 3. > > <rant> > Karaf is a great project but the surprise regressions and the long variable > delays between versions (which include fixes for these regressions) is very > frustrating. What makes it worse for me is that in the version I was using > (Karaf 2.3.2) editing factory configurations (not stored in etc) from the > Karaf command line console is broken causing me hours upon hours of misery > as the workaround for it (copy the factory configuration to a file in the > etc directory, edit the file, delete the old factory configuration, restart > karaf) is very error prone (I can't ask people to use the web console > because I need to use Pax Web 3 with Karaf 2.3.2). I keep being told how > karaf is "too hard to use". Oh well... > </rant> > > thanks, > Gareth > > > > -- > View this message in context: > http://karaf.922171.n3.nabble.com/Problems-With-Factory-Configurations-In-Karaf-3-0-1-tp4033921p4035683.html > Sent from the Karaf - User mailing list archive at Nabble.com.
