MODE-1965 Added explicit validation around mixin types not being able to inherit non-mixin types. #857
Conversation
|
|
||
| [ddl:operand] mixin abstract |
There was a problem hiding this comment.
Why are these 'abstract' keywords removed. It's important because we don't want these mixins to be used directly; instead, specific subtypes must be used.
|
The changes to the DDL sequencer are incorrect IMO. We need to keep the notion of some of the node types being 'abstract' so that they cannot directly be used as a node's primary type or mixin type. |
|
added back the |
| @@ -37,6 +37,9 @@ | |||
| [xmi:referenceable] mixin | |||
| - xmi:uuid (string) mandatory | |||
|
|
|||
| [xmi:model] > nt:unstructured, xmi:referenceable, mix:referenceable orderable | |||
| [xmi:model] > xmi:referenceable, mix:referenceable orderable mixin | |||
There was a problem hiding this comment.
Also not sure about this change. This wasn't a mixin, so it was okay for it to extend 'nt:unstructured' and other mixins at the same time. Remember, non-mixin node types can extend other non-mixins and mixins, but a mixin cannot extend a non-mixin.
There was a problem hiding this comment.
Okay, @hchiorean pointed out to me that xmi:model is extended by other mixins. For example: https://github.com/ModeShape/modeshape/blob/master/sequencers/modeshape-sequencer-teiid/src/main/resources/org/modeshape/sequencer/teiid/mmcore.cnd#L39
… to inherit non-mixin types.
|
Merged into the 'master' branch, and cherry-picked into the '3.3.x' branch. |
This exposed a number of problematic areas in some of our extensions, that had to be fixed:
NodeTypeTemplate) mixins that inherited from non-mixin types, but was really using them as primary types.