While trying to reproduce scenario described by Colin Ma in
https://issues.apache.org/jira/browse/SENTRY-1422 I made an attempt to use
the same version of datanucleus parts as is used by Sentry and discovered a
few interesting things:
1. Inconsistency between javax.jdo - Sentry uses Apache javax.jdo while
DN has its own version which is similar but not exactly the same. Given
that DN tests with its version I think we should consider switching to DN
version of javax.jdo.
2. Inconsistency between versions:
1. datanucleus api - 3.2.6 for some parts and 4.1.6 for other parts
2. datanucleus core - 3.2.12 for some parts and 4.1.6 for other parts.
3. datanucleus rdbms - 3.2.12 for some parts and 4.1.7 for other
parts.
3. The maven datanucleus plugin is not compatible with 3.2.6
datanucleus API - I can't run enhancer, so I can't use exactly the same
version as upstream for non-Sentry code.
Datanucleus itself is now on version 5. I think that some parts of version
5 require Java 8, so we should, probably, stick for version 4.2 for now but
it definitely makes sense to move from 3.2 to 4.2 - and it would be great
to be consistent between CDH and upstream versions.
What are your thoughts on this?
- Alex Kolbasov.
