Hi, meanwhile I created an issue for the problem seen with the release candidate: OAK-4085
I still think we should cancel the release because the node type registry is malformed after an upgrade with reregistered node types. with the votes we currently have, the release would still go out. anyone with a +1 willing to change his mind? Regards Marcel On 03/03/16 11:25, "Marcel Reutegger" wrote: >Hi, > >I have to change my vote based on further testing to > >-1 > >As noted earlier I was looking into an upgrade issue >reported by Zygmunt Wiercioch (OAK-4077). This fix >is included in the 1.4.0 release candidate. However >there appears to be a more severe problem with node >type definitions in the repository. > >A while back OAK-3584 fixed the indexes for names of >item definitions. This change works well for new >repositories or newly registered node types, but it >makes the situation worse when an existing node type >is reregistered. The result after an upgrade may look >like this: > >my:type > + jcr:childNodeDefinition > + jcr:childNodeDefinition[1] > + jcr:childNodeDefinition[2] > + jcr:propertyDefinition > + jcr:propertyDefinition[1] > >The duplicate child nodes with explicit and implicit >index 1 is not the only issue. It may also happen that >there are duplicate item definitions after a node type >is reregistered. The NodeTypeDiff is also affected and >may report an item definition is removed even though >it is still there. > >In my view these are rather severe issues for users >upgrading from earlier Oak versions and we should not >release 1.4.0 with these kind of problems. > >Regards > Marcel > > > >On 03/03/16 09:47, "Marcel Reutegger" wrote: >>Hi, >> >>On 02/03/16 17:34, "Davide Giannella" wrote: >>>Please vote on releasing this package as Apache Jackrabbit Oak 1.4.0. >>>The vote is open for the next 72 hours and passes if a majority of at >>>least three +1 Jackrabbit PMC votes are cast. >> >>All checks OK. >> >>+1 Release this package as Apache Jackrabbit Oak 1.4.0 >> >>Regards >> Marcel >> >
