https://git.opendaylight.org/gerrit/#/c/68789/

https://git.opendaylight.org/gerrit/#/c/68789/3

The question I guess is if we can just exclude Karaf's OSGi mySQL(TM)(R) JDBC driver support provided by Pax JDBC in odlparent? Out of curiosity, I did a (very) quick grep, and find JDBC usages at least in AAA (but that AFIK is by embedding h2, so not a problem here) and in TSDR (which I've no idea what it does; but seems again h2 only not mysql?) and something apparently related to Postgres in VTN (again, I've no clue about that ODL project). It's perhaps at least worth a question on wider lists? Not to forget all possible open and closed downstream in-house ODL applications... which this global exclusion of the mysql JDBC driver, theoretically, could break.

An alternative could be not to do the exclusion in odlparent/karaf/karaf4-parent/, but instead let projects which build local Karaf distribution, such as netvirt/distribution/karaf and the like, opt in to do such an exclusion themselves?

The more durable long term correct solution is probably https://ops4j1.jira.com/browse/PAXJDBC-128 ... if anyone from the ODL community wants to contribute to that, I'm happy to help with getting that done (but currently don't have enough reason to see to that from A-Z myself).

This is new as of Nitrogen (i.e., Karaf 4.0.10).  I highly doubt anything from ODL utilized it last release, but I could be wrong (I know AAA and TSDR did not).

Other downstream projects can still utilize the artifact by deploying it using out of band methods;  this simply prevents us from shipping it.

I can look into maven-enforcer-plugin instead, but this seemed simpler at the time.

rgoulding just to be clear, I'm not fundamentally against your c/68789 being merged if other odlparent commiters are OK with it; the review & comment above are feedback thoughts, not -1/-2 "I'm against c/68789".

Of course, I appreciate the feedback on it!  There are likely a few more sound ways we can do this long term, but they may require changes in upstream projects as you suggest.  If the odlparent committers like this approach, great.  If not, we will figure out what needs to be done in order to address the issue for this release.  It probably could also be solved with a simple release note, but I'm not an IP expert by any means!

It would be easier to ship a fix for this in Oxygen if we can deal with it in integration/distribution; I’m looking into that.

Ryan added a post-build task to remove the dependency, nicely dealing with the issue. The fix is https://git.opendaylight.org/gerrit/68844 on master, https://git.opendaylight.org/gerrit/68857 on Oxygen, and https://git.opendaylight.org/gerrit/68858 on Nitrogen.

The patches are merged in integration/distribution, so this isn’t a release blocker any more. It might still be worth adding a feature to odlparent to filter out undesirable GAVs, but I’ll mark this bug resolved for now...

Patch to be merged for oxygen: https://git.opendaylight.org/gerrit/#/c/68857/

Thanh has merged the oxygen patch.