eclipse-qvtd-bot
commented
15 hours ago By Ed Willink on Mar 07, 2019 12:52
(In reply to Ed Willink from comment #0)
Should we require that at least the root objects have a supertype that is one of the non-abstract input metamodel elements?
We pretty much have to do a total model categorization to align the input elements to the Connections from which they propagate. Detecting stupid model elements shouldn't be too costly.
Unfortunately the CG and interpreted categoraizations are pretty different. Time to unify.
| --- | --- | | Bugzilla Link | 545181 | | Status | NEW | | Importance | P3 normal | | Reported | Mar 07, 2019 10:56 EDT | | Modified | Mar 07, 2019 12:52 EDT | | Reporter | Ed Willink |
Description
Work on Bug 537609 provides some launch tests that deliberately fail.
But specifying a stupid input file, e.g. a *.qvtr (Qhen not a QVTr2xxx tx) gives no error, just a silent nothing becuase no input elements match.
From one perspective this is correct, but it is really unhelpful.
Should we require that at least the root objects have a supertype that is one of the non-abstract input metamodel elements?