Supporting MSL 3.2.2

[Francesco Casella]

The release plan for MSL 3.2.2 is found here:

​https://trac.modelica.org/Modelica/ticket/1867

The next major deadline is Feb 19, when a RC will be released. The Beta 1 revision was tagged on Feb 12, and nothing substantial has been done on trunk since then, so the Beta 1 is roughly equivalent to the trunk.

I have done a quick regression between

​https://test.openmodelica.org/libraries/MSL_trunk/BuildModelRecursive.html

and

​https://test.openmodelica.org/libraries/MSL_3.2.1/BuildModelRecursive.html

The front end is ok, but there are some problems with back-end, code generation and simulation which affect 23 models (9 existing ones and 14 new ones). I haven't counted problems that already affected MSL 3.2.1, though it would be of course nice to solve them as well. But I think it is imperative that OMC doesn't show a worse performance on MSL 3.2.2 than it does on MSL 3.2.1.

You can find a summary of the analysis in the attached spreadsheet.

I would urge the developers to look at this problems and report if they are caused by bad Modelica code (that Dymola handles somehow, so developers don't care), or if they are caused by some omc bug.
In the first case, please report to me and/or open a ticket on trac.modelica.org.

In the second case, it would be nice to have the problem fixed by the time MSL 3.2.2 is released (planned date: March 11 2016), or even better for the 1.9.4 release.

I would also suggest that, once MSL 3.2.2 is out, we release a new version of OMC that uses it ASAP.

Note that DS (Dymola), Modelon (JModelica) and ITI (SimulationX) have taken a very active stand on ​https://trac.modelica.org/Modelica/ticket/1867. It would be weird if we didn't, as we are aiming at becoming the reference Modelica compiler.

[Francesco Casella]

Regression analysis

[Lennart Ochel]

I agree that we should fix these issues ASAP. It would be good to have one ticket for each type of error with a list of related models. That would make it easier to assign the issues properly. However, I will have a look at the back end issues after the OpenModelica workshop.

[Francesco Casella]

Replying to lochel:

I agree that we should fix these issues ASAP. It would be good to have one ticket for each type of error with a list of related models. That would make it easier to assign the issues properly.

I can try to manage this

[Francesco Casella]

As the final release of 1.9.4 is approaching, we have this situation:

MSL 3.2.1:
BuildModel Results: 278/278 succeeded (100.0%)
Simulation Results: 271/278 succeeded (97.4%)

MSL 3.2.2 (MSL_trunk)
BuildModel Results: 358/366 succeeded (97.8%)
Simulation Results: 345/366 succeeded (94.2%)

IMHO we cannot really release 1.9.4 as it is with these numbers, because it will look as if the performance of OMC in terms of MSL coverage is going down instead of going up.

OMC 1.9.4 is in Beta phase, and I would say supporting MSL 3.2.2 is definitely a new feature, so I would not attempt to include MSL 3.2.2 there. That would be too risky: I am not 100% sure MSL 3.2.2 will be released when currently planned, and also we don't know how much time will be needed to support it at the same level as MSL 3.2.1.

I would recommend that we release 1.9.4 ASAP and then immediately start work on supporting MSL 3.2.2 as satisfactorily as we supported MSL 3.2.1. Then, if we are still not ready for 2.0, we can release 1.9.5 with MSL 3.2.2 as soon as we have a coverage rate similar to what we had with MSL 3.2.1.

[Martin Sjölund]

Also note that even MSL C-sources are not stable yet. There were additional changes tonight...

[Rüdiger Franke]

Also note that from the perspective of most users it's just the patch level that increases from 3.2.1 to 3.2.2. You'll hardly notice a difference unless you are interested in specific new sublibraries, such as Modelica.Electrical.PowerConverters and Modelica.Magnetic.QuasiStatic.FundamentalWave.

[Martin Sjölund]

Milestone pushed to 1.9.5

[Martin Sjölund]

Milestone renamed

[Anton Haumer <a.haumer@…>]

MSL v3.2.2-build.3-release was released 2016-04-03, it's stable!

See discussion at #3499. To me it seems most of the problems are related to index reduction and overconstrained connection sets.

Modelica.Electrical.PowerConverters and Modelica.Magnetic.QuasiStatic.FundamentalWave are important extensions!
We (Christian and I) plan to use both libraries with OpenModelica for teaching.
Here at Technical University of Applied Sciences Regensburg I have at least 2 important industrial project partners that just started using OpenModelica for electric drives modeling according to my advice, and for them both libraries are essential!

[Francesco Casella]

The situation has improved, now OMC can simulate 373 models out of 385 in MSL 3.2.2, i.e. 96.9%. We still need to solve the problems with the remaining 12 models before the 2.0.0 release.

[Francesco Casella]

Update: 379 out of 388 models now simulate, i.e. 97.7%, see ​report. Only 338 pass the verification stage, these should be checked prior to the release of 2.0.0.

In fact, 2.0.0 will use the new front-end, so the ​coverage using the NF should be at least as good as this before releasing.

[Francesco Casella]

This ticket is now obsolete, please see #5288 instead