Sorry if this is a duplicate. I never saw the first one go through. In Aries, we maintain and release separate bundles for the API and implementation. However, you could use the subsystems all in one bundle [1] to achieve the same effect.
[1] http://search.maven.org/#artifactdetails%7Corg.apache.aries.subsystem%7Corg.apache.aries.subsystem%7C2.0.2%7Cbundle >To: [email protected] >From: David Jencks <[email protected]> >Date: 09/10/2015 01:46PM >Subject: Re: Works on Laptop, Problem on Raspi > >For R6 (as well as earlier), you should expect that the spec >implementation bundle includes the api so you shouldn’t really be >using the individual api bundles there either. However apparently >not all implementations adhere to this best practice. > >thanks >david jencks > >> On Sep 10, 2015, at 2:23 PM, David Bosschaert ><[email protected]> wrote: >> >> Hi Paul, >> >> There are some interesting comments on the bug, especially the last >> ones from John Ross and David Jencks: >> >https://issues.apache.org/jira/browse/ARIES-1398?focusedCommentId=147 >39043&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-ta >bpanel#comment-14739043 >> >> It would be interesting to see whether this problem disappears if >you >> try your original scenario without the cmpn.jar bundle. >> >> (Note that the cmpn.jar bundle was not really designed for runtime >> use. As of the OSGi R6 releases the cmpn.jar bundle does not work >any >> more at runtime and there are individual runtime bundles for spec >APIs >> instead) >> >> Cheers, >> >> David >> >> On 9 September 2015 at 11:47, Paul F Fraser <[email protected]> >wrote: >>> Just as a matter of interest, >>> >>> The reason it turned up on the pi was that I was starting the >bundles from a >>> list of files in a directory. >>> >>> All laptops and desktops I tried kept the order of the files from >the >>> directory. >>> >>> The pi, running ubuntu, scrambled the order. >>> >>> Ubuntu on my laptops kept the order. >>> >>> I changed the system to read from a file and works OK everywhere. >>> >>> Paul >>> >>> >>> >>> On 9/09/2015 6:25 PM, David Bosschaert wrote: >>>> >>>> Thanks Paul! I have been able to reproduce this on my mac as >well, so >>>> it's not really Raspberry Pi specific. I've created >>>> https://issues.apache.org/jira/browse/ARIES-1398 >>>> >>>> Cheers, >>>> >>>> David >>>> >>>> On 9 September 2015 at 03:00, Paul F Fraser <[email protected]> >wrote: >>>>> >>>>> Hi David, >>>>> >>>>> I did some more tests on the pi and came up with the attached >results. >>>>> >>>>> Paul >>>>> >>>>> >>>>> >>>>> >>>>> On 9/09/2015 9:04 AM, Paul F Fraser wrote: >>>>>> >>>>>> Hi David, >>>>>> >>>>>> In my case, playing with the start order on the pi is a bit >messy and >>>>>> time >>>>>> consuming. >>>>>> A check of start order of just the Aries subsystems bundles on >my >>>>>> windows >>>>>> desktop indicates no problem with start order, even when >including my >>>>>> own >>>>>> bundles. >>>>>> >>>>>> I will create an issue if and when I can establish a cause of >this >>>>>> problem. >>>>>> >>>>>> Paul >>>>>> >>>>>> On 8/09/2015 6:09 PM, David Bosschaert wrote: >>>>>>> >>>>>>> Hi Paul, >>>>>>> >>>>>>> Start order within a single start level is undefined. If you >really >>>>>>> need a specific start order then you can use OSGi start levels >to >>>>>>> achieve this, but it is generally better to avoid that if >possible. >>>>>>> >>>>>>> The Aries Subsystem bundles should not require a particular >starting >>>>>>> order. Could you maybe create an Aries issue [1] for this? I >guess >>>>>>> it's easy to reproduce by just installing the subsystem >bundles and >>>>>>> then starting them in the 'wrong' order? >>>>>>> >>>>>>> Thanks, >>>>>>> >>>>>>> David >>>>>>> >>>>>>> [1] https://issues.apache.org/jira/browse/ARIES >>>>>>> >>>>>>> On 8 September 2015 at 07:27, Holly Cummins >>>>>>> <[email protected]> wrote: >>>>>>>> >>>>>>>> Interesting, thanks for sharing the solution! >>>>>>>> >>>>>>>> On 8 Sep 2015, at 05:53, Paul F Fraser <[email protected]> >wrote: >>>>>>>> >>>>>>>> Hi, >>>>>>>> >>>>>>>> Seems that bundle start order is important. Raspi scrambled >the start >>>>>>>> order >>>>>>>> of bundles. >>>>>>>> >>>>>>>> Forcing start order to be the same as other devices caused >raspi to >>>>>>>> create >>>>>>>> the synthesized bundle. >>>>>>>> >>>>>>>> Paul >>>>>>>> >>>>>>>> >>>>>>>> On 7/09/2015 8:07 PM, Paul F Fraser wrote: >>>>>>>> >>>>>>>> Hi, >>>>>>>> >>>>>>>> Using subsystems on osx, linux and windows (laptops),the >synthesized >>>>>>>> bundle >>>>>>>> is created. >>>>>>>> >>>>>>>> org.osgi.service.subsystem.region.context.0 (1.0.0) >>>>>>>> >>>>>>>> On a raspberry pi (armhf) this is not created. >>>>>>>> >>>>>>>> Can someone point me to the code I should be checking to see >where the >>>>>>>> problem could be. >>>>>>>> >>>>>>>> Or is there any known reason why this might be so? >>>>>>>> >>>>>>>> Regards >>>>>>>> >>>>>>>> Paul Fraser >>>>>>>> >>>>>>>> >>>>>> >>> > >
