[releng] Refresh the OOMPH setup

[eclipse-qvtd-bot]

| --- | --- | | Bugzilla Link | 564180 | | Status | RESOLVED FIXED | | Importance | P3 normal | | Reported | Jun 10, 2020 12:55 EDT | | Modified | Oct 22, 2021 04:47 EDT | | See also | 576546, 576593, 564135 | | Reporter | Ed Willink |

Description

qvtd.setup has gone stale.

Revamp.

[eclipse-qvtd-bot]

By Ed Willink on Oct 10, 2021 03:43

Latest target was 2020-03 when org.eclipse.emf.examples.library was still required.

No 2020-09 release allowed a careless copy and paste in ocl.oomph to go adrift.

---

After refreshing fails at at a double entry from a cached Orbit.

ERase .p2 cache totall.

Now fails on a bad Epsilon mirror. See Bug 576538.

'latest' ytarget is hard to use OCL nightlu - defer

2021-09 target is not too bad but Bug 576546

[eclipse-qvtd-bot]

By Ed Willink on Oct 10, 2021 05:03

Improved qvtd.setup pushed to master.

2021-09 platform/stream works. More work for latest / Sirius / ...

[eclipse-qvtd-bot]

By Ed Willink on Oct 12, 2021 07:22

Simplified to eliminate Modular Target. Now works for a variety of versions, but not for Latest.

---

The current qvtd.setup is an evolution from Adolfo's first attempt at ocl.setup and my hack-it-and-see instinct.

After starting to read the OOMPH Authoring Guide, clearly a lot has gone amiss. In particular the Outline shows nothing useful, rather than the state if variables that might enable me to debug the variant repo to make Latest work. Possibly calling it Master (Latest) was a bad idea.

The simplification to eliminate the Modular Target is probably a step in the wrong direction. Certainly duplication between P2 Director and Modular Target is stupid, but it is hard to persuade the P2 Director to use a stream-specific repo.

Probably the P2 Director should be an as-thin-as-possible bootstrap with only the Modular Target, dreadful name, being the functional installation that can more easily have project/stream sensitivities.

Anyway, kicking off again with the OOMPH new wizard gives something much more debuggable. The way forward must be to take the new-QVTd created by the wizard and then merge the relevant details from the old-QVTd.

[eclipse-qvtd-bot]

By Ed Willink on Oct 13, 2021 04:41

(In reply to Ed Willink from comment #1)

when org.eclipse.emf.examples.library was still required.

Eliination of org.eclipse.emf.examples.library unfortuantly created a confusing typo where by both

examples/org.eclipse.ocl.examples.extlibrary\ examples/org.eclipse.ocl.examples.library

have a.project whose name is org.eclipse.ocl.examples.extlibrary

The duplication gives an almost random choice as to which P2 imports giving unpredictable initial build errors.

Successfully moving examples/org.eclipse.ocl.examples.library to archive has taken much longer than it should.

[eclipse-qvtd-bot]

By Ed Willink on Oct 13, 2021 04:43

(In reply to Ed Willink from comment #3)

Probably the P2 Director should be an as-thin-as-possible bootstrap

Yes. But Bug 576593 requires UML2.

Xtext cannot be avoided since the Xtend bui;der must exist in the Running platform.

[eclipse-qvtd-bot]

By Ed Willink on Oct 13, 2021 08:01

(In reply to Ed Willink from comment #3)

Anyway, kicking off again with the OOMPH new wizard gives something much more debuggable.

The appearance of resolved/unresolved/undeclared variables in the Outline seems to be very dependent on the Outline toolbar settings, which perhaps have unexpected history, and just start good from the wizard. Select a stream and click away at the toolbar and eventually the variables appear. Even when they appear undeclared does not seem consistently diagnosed.

[eclipse-qvtd-bot]

By Ed Willink on Oct 16, 2021 12:09

Much better ocl/qvtd/qvto.setup pushed to master.