eclipse-qvto / org.eclipse.qvto

Eclipse Public License 2.0
0 stars 1 forks source link

Application level JUnit tests #841

Open eclipse-qvt-oml-bot opened 3 days ago

eclipse-qvt-oml-bot commented 3 days ago

| --- | --- | | Bugzilla Link | 455946 | | Status | NEW | | Importance | P3 normal | | Reported | Dec 22, 2014 06:15 EDT | | Modified | Dec 22, 2014 10:08 EDT | | Reporter | Ed Willink |

Description

QVTo has a significant number of statement level unit tests, but they failed to pick up Bug 455917 and a previous regression.

Perhaps the test suite should be augmented by a few application level unit tests that demonstrate that the QVTo-under-test compiles and executes a non-trivial QVTo application on non-trivial model(s) to produce identical outputs to reference saved outputs. There are probably quite a few users who may be prepared to contribute applications for this purpose.

See org.eclipse.ocl.examples.xtext.tests.assertSameModel for a wrapper on org.eclipse.xtext.util.EmfFormatter.listToStr that eliminates irrelevant differences and gives a result that can be viewed directly within JUnit.

eclipse-qvt-oml-bot commented 3 days ago

By Sergey Boyko on Dec 22, 2014 06:41

Actually we have JUnit exactly for Bug 455917. \ The only difference is that in test we check linear dependency (project 'p1' references 'p2') while mentioned bug is caused by cyclic dependencies (project 'p1' references 'p2' and at the same time 'p2' references 'p1').\ And consequence of fixing Bug 455917 will be another JUnit to test the cyclic dependencies.

eclipse-qvt-oml-bot commented 3 days ago

By Ed Willink on Dec 22, 2014 10:08

Yes, but statement level tests cover the issues we know.

Application level tests while hideously redundant coverage wise can tackle issues we do not know, and when contributed by significant customers can avoid upsetting those customers.