David Schneider

How is this issue different from #1707?

> @uekerman @Fujikawas do you think we can still improve something here, while still allowing simulations to continue? I think we can still do better. The problem you describe above...

> Using a preconditioner when there is only one data vector in the acceleration is anyway useless. Maybe we could forbid this or switch off automatically. Yes, switching off automatically...

That's a very good idea. I have a pending PR for missing initializations since clang complains about them and this would actually be a perfect use-case. In my opinion, we...

> Strong types could have saved us here: https://github.com/precice/precice/pull/1590 (I mixed up a mesh ID with a size in the test) With the recent changes in #1461, these kind of...

Config of the coupling scheme was wrong. This version works: ```xml ```

> I guess a proper fix would be to create (yet another) interface in the mapping methods which enable each mapping to modify the bounding boxes in the partitioning class....

Also relevant here: The `geometric-filter=no-filter` silently doesn't work for 2LI. The combination doesn't make sense, but we should probably raise an error in the configuration, see also https://github.com/precice/precice/pull/1912.

> @davidscn should GinkGo also be in the dependencies page? We also list optional dependencies there. I would say no, as we are not compatible with a released version.

As part of https://github.com/precice/precice/issues/1608 we (or probably I) anyway have to rewrite the mapping documentation, where I would mention this feature. The Ginkgo part is anyway experimental at the moment,...