#opendaylight-meeting: carbon release sync

Meeting summary

1. overview of status (colindixon, 15:01:49)
   1. https://docs.google.com/spreadsheets/d/1VcB12FBiFV4GAEHZSspHBNxKI_9XugJp-6Qbbw20Omk/edit#gid=921315511 blocking bugs (colindixon, 15:02:04)
   2. colindixon thinks most of our current issues are likely tied to the fact that no bundles/features in OpenDaylight are built to be reloaded, but Karaf 4 lifecycle management by default reloads them quite a bit because we have dependency=false by default (colindixon, 15:09:20)
   3. the issue with that is that the plugin is hosted in karaf not in ODL, so changing the karaf plugin is a longer-term strategy than we would otherwise have to go after (colindixon, 15:15:14)
   4. we could (in theory) patch features.xml files in autorelease, but it would require two build passes, one to generate the feature.xml files, and other to modify them and rebuild (colindixon, 15:16:08)
   5. skitt says in theory, we could probably hook in just after the karaf maven plugin and do it in a single pass, but we haven't done that yet (colindixon, 15:16:30)
   6. skitt also says we could in theory do it in the karaf 4 distribution in integration/distribution, which would work, but would ship something different that we would test (basically just manually tweak all the features.xml files in the distribution feature repo) (colindixon, 15:17:20)
   7. that would have the somewhat bad property that the integration/distribution artifacts would not match nexus, that makes it a good short-term candidate for testing if this fixes most of the issues and let us know if we need to work harder on things to get it working (colindixon, 15:20:22)
   8. colindixon asks about prerequisite=true as well, skitt says that will likely need to be added in some places, but adding it everywhere will result in deadlocks (colindixon, 15:23:14)
   9. the other key question is how are we doing with Karaf 3 in Carbon, the answer is CSIT running against it probably means it's in pretty good shape (colindixon, 15:26:21)
   10. skitt says that Karaf 3 still has security support from apache, but that doesn't totally save us as it could be that karaf 3 stops us from pulling in a dependency that would be critical for us, but doesn't matter to Karaf (colindixon, 15:31:13)
   11. lori says that he's seeing issues that look like "Unable to start… aaa …" when testing lispflowmapping (colindixon, 15:34:11)
   12. colindixon notes that we have a bug tracking something like this https://bugs.opendaylight.org/show_bug.cgi?id=8373 (colindixon, 15:37:28)
   13. lori points out that he ran into stack overflow errors early in the Karaf 4 testing and he thought it was because of dependency=true, but it might be because of prerequisite=true (colindixon, 15:38:19)
   14. skitt says his hope is that won't be the case for dependency=true, it was more for prerequisite=true, but we'll see (colindixon, 15:40:24)
   15. katie asks if this will eventually require changes in all projects, skitt and colindixon say they really hope not and shouldn't (colindixon, 15:40:48)

Action items

1. (none)

People present (lines said)

1. colindixon (18)
2. odl_meetbot (3)