open-simulation-platform/libcosim

Clean up the test FMU set

kyllingstad opened this issue · 5 comments

The tests/data/fmi2 directory contains a couple of FMUs from the old official "Test FMUs" set:

  • RoomHeating_OM_RH.fmu
  • WaterTank_Control.fmu

These are no longer available on the FMI web site, and as far as I can tell, they're no longer easily available online at all. We probably shouldn't be hosting obscure binary blobs in our repo.

And there are other issues too:

  • There is no licence info for the abovementioned FMUs, which is bad form.
  • Even our own FMUs are just binary blobs with no information about their origin or licence.
  • Clock.fmu is our own, but it's not part of cpp-fmus. Where is the source for that?

We really should clean this up. I would understand any user who is sceptical of running tests where we use these FMUs.

We do need to test with real FMUs, though, so any suggestions for improvements are welcome.

Actually, the RoomHeating FMU doesn't seem to be used by our tests, so it can simply be removed.

I'm working on this, and it turns out that almost all of these FMUs can be removed. The only ones I haven't been able to sort out are tests/data/ssp/demo/CraneController.fmu and tests/data/ssp/demo/KnuckleBoomCrane.fmu. Judging from the source files inside them, it seems that both were generated by 20-sim using an NTNU Ålesund student license. Are they yours, @markaren?

If so, would you like to specify a licence for them? I'm trying to ensure that we provide the correct licence file alongside each FMU.

They were generated by @yingguangchu. Ideally the SSP tests should just perform the same tests as the osp config, allowing us to remove these FMUs.

The OSP config tests use the same FMUs, it seems. But since we've now confirmed that the FMUs were contributed by NTNU to the OSP project, I'll assume that they are covered by the MPL 2.0 licence like most of the other things in the repo.

The most pressing issues were fixed in #764, so I'm closing this. Feel free to reopen if anyone sees the need to get rid of binary FMUs entirely.