eclipse-qvtd-bot
commented
1 week ago By Adolfo Sanchez-Barbudo Herrera on Jun 08, 2015 11:03
Hi Ed,
When all the cs2as gen was not integrated in QVTd, the xsd component (as well as epsilon ones) were installed from qvtd-extras.setup. The xsd component installation should haven been ported to the qvtd.setup when the integration took place.
The dependency was needed in order to execute some qvts2graphml ETL transformation in the OCL2QVTi test suit (when the debug flag is activated). If XSD is not present, there is no compilation error: the ETL transformation fails at runtime.
I think that the new graphml generator doesn't need it anymore, so I agree that the dependency can be removed, and the OCL2QVTi test cases should only use the new graphml generator (currently both approaches are invoked). Provided I wanted to run the "old" qvts2graphml transformation, I could always locally install the XSD component and add the dependency in my own.
IIRC, Oomph doesn't make a kind of dependencies validation in your setup, you have to run it and check the resulting workspace by yourself.
Regards,\ Adolfo.
| --- | --- | | Bugzilla Link | 468102 | | Status | NEW | | Importance | P3 normal | | Reported | May 23, 2015 05:02 EDT | | Modified | Jun 08, 2015 11:03 EDT | | Reporter | Ed Willink |
Description
Installing OCL and QVTd using OOMPG from Mars latest at RC1+4 gives only 8 errors.
(7 are because the Xtext builder imposes a unified ResourceSet on all builds. Simple solution, eliminate the Xtext builder wherever there are *.ocl etc files.)
1 is because of a spurious XSD dependency. Trivial fix, the dependency isn't needed. But shouldn't something be diagnosing this at compile-time? Is there a missing OOMPH validate capability to locate missing dependencies? Is the capability there but we are not using it?