[scheduler] New scheduler doesn't work with example1

[eclipse-qvtd-bot]

| --- | --- | | Bugzilla Link | 485853 | | Status | RESOLVED FIXED | | Importance | P3 normal | | Reported | Jan 14, 2016 09:56 EDT | | Modified | Jan 17, 2016 08:28 EDT | | Reporter | Adolfo Sanchez-Barbudo Herrera |

Description

When incorporating to the OCL2QVTi test suite a new test case for the example1 so that it uses the new scheduler, both testNewExample1_Interpreted and testNewExample1_CG fail.

I'll push a new branch as soon as I get the new bug id.

[eclipse-qvtd-bot]

By Ed Willink on Jan 14, 2016 10:33

Example1 Example3 hstm2stm are on my to do list. Until I claim they work there is no point reporting bugs.

[eclipse-qvtd-bot]

By Ed Willink on Jan 16, 2016 09:10

(Using EObject as the 'ast' target is very provocative. EObject is not in any of the models so it forces a more generous interpretation that gets confused because EObject is in both models, forcing the DomainAnalysis to work very hard to make sense of the ambiguity. Using Visitable as the 'ast' target is much more credible.)

Ultimately the problem was inadequate discrimination between isNavigation() and isNavigable() edges. Need a snappier name for isNavigable() expressing the to-one navigable sub-region.

commit cab215ae504708c1ebe8547079b76639a7bbc0ed pushed to master for M5.

/org.eclipse.qvtd.cs2as.compiler.tests/src/org/eclipse/qvtd/cs2as/compiler/tests/models/example1/Source2TargetSchedule-9-final.graphml\ /org.eclipse.qvtd.cs2as.compiler.tests/src/org/eclipse/qvtd/cs2as/compiler/tests/models/example1/Source2TargetSchedule-9r-final.dot

look convincing but only 3 early merges were found. Must get round to re-enabling the late merges.

[eclipse-qvtd-bot]

By Adolfo Sanchez-Barbudo Herrera on Jan 17, 2016 08:14

Hi Ed,

Thanks for fix. Verified.

(In reply to comment #2)

(Using EObject as the 'ast' target is very provocative. EObject is not in any of the models so it forces a more generous interpretation that gets confused because EObject is in both models, forcing the DomainAnalysis to work very hard to make sense of the ambiguity. Using Visitable as the 'ast' target is much more credible.)

I understand your comments but unfortunately not every possible AS meta-model need to have a common meta-class for every of them. So, although me can tweak the CS for particular cases, I think that prohibiting generic cross-references using EObject as its type is an undesired limitation I'd like not to impose.

[eclipse-qvtd-bot]

By Ed Willink on Jan 17, 2016 08:28

(In reply to Adolfo Sanchez-Barbudo Herrera from comment #3)

I understand your comments but unfortunately not every possible AS meta-model need to have a common meta-class for every of them. So, although me can tweak the CS for particular cases, I think that prohibiting generic cross-references using EObject as its type is an undesired limitation I'd like not to impose.

It is the use of type that is not in any model that I find unnecessary and smelly.