[tests] Hidden load failures

[eclipse-qvtd-bot]

| --- | --- | | Bugzilla Link | 521077 | | Status | NEW | | Importance | P3 normal | | Reported | Aug 17, 2017 13:03 EDT | | Modified | Aug 30, 2017 05:16 EDT | | See also | 521577 | | Reporter | Ed Willink |

Description

A number of tests such as CS2AS testExample4_Interpreted fail to load adequately resulting in a failure to may EIntegerObject to a pivot type. The new xmi:id detects that as a bad opposite.

Temporarily worked around in PivotLUSSIDs.java.

[eclipse-qvtd-bot]

By Marcus Huewe on Aug 29, 2017 10:31

Maybe something related: the attached project, which contains two complete OCL\ files, can be used to provoke a

No opposite for $$::Tuple::xyz

error.

How to reproduce:

• import the attached project
• clean the imported project "Project -> Clean..." (this starts the\ XtextBuilder, which indirectly calls DefaultResourceDescriptionStrategy.createReferenceDescriptions(...),\ which eventually (indirectly) executes the related AS2ID code)

[eclipse-qvtd-bot]

By Marcus Huewe on Aug 29, 2017 10:33

Created attachment 270012 CompleteOCL reproducer for the "No opposite..." message

bug521077.tar

[eclipse-qvtd-bot]

By Ed Willink on Aug 29, 2017 10:57

(In reply to Marcus Huewe from comment #1)

executes the related AS2ID code)

AS2ID was first pushed to OCL master on 18-Aug-2017 and so there is no stable OCL build that includes it. Are you using a recent N-build or GIT master?

This is a QVTd bug not that obviously related to your possibly OCL issue.

I cannot reproduce the issue in my workspace, which probably differs in some significant way from your configuration.

[eclipse-qvtd-bot]

By Marcus Huewe on Aug 29, 2017 13:13

Created attachment 270014 Plugin test to provoke the "No opposite for..." message

:notepad_spiral: 0001-Plugin-test-to-provoke-the-No-opposite-for.-message.patch

[eclipse-qvtd-bot]

By Marcus Huewe on Aug 29, 2017 13:22

(In reply to Ed Willink from comment #3)

(In reply to Marcus Huewe from comment #1)

executes the related AS2ID code)

AS2ID was first pushed to OCL master on 18-Aug-2017 and so there is no stable OCL build that includes it. Are you using a recent N-build or GIT master?

git master (in a vanilla eclipse oxygen).

This is a QVTd bug not that obviously related to your possibly OCL issue.

Agreed. If you prefer, I can open a new bug report for the OCL product.

I cannot reproduce the issue in my workspace, which probably differs in some significant way from your configuration.

I just hacked a junit plugin test that does not pass, because\ the "No opposite for $$::Tuple::xyz" is printed.\ Can you reproduce it via this plugin test?

[eclipse-qvtd-bot]

By Ed Willink on Aug 30, 2017 05:16

(In reply to Marcus Huewe from comment #5)

Agreed. If you prefer, I can open a new bug report for the OCL product.

Thanks. Reproduced. Moved to Bug 521577 since this is nothing to do with the original and outstanding observation that there are some hidden load failures (observable if IOExceptions are breakpointed.)