Andy Porter

> I don't think it is, no. However, this is the change that fixed that problem with the examples failing (I don't know how, it just did). I'm afraid that...

> Is there a possibility that when ACC transformations happen, the `KernCallArgList` is remade with no checks for where the arguments originate from? I'm not entirely sure what's happening I...

> That is exactly what's happening. Does this mean that the `isinstance` check is justified, or that another change needs to be made? It probably means that you need to...

Unfortunately the compilation checks failed:  Looks like you just need proper declarations for the arguments to the test kernel. You can test this locally by doing `pytest --compile ...`.

Integration tests were fine (modulo the existing race-condition problem in threaded NEMOv4).

This all sounds sensible and good to me. However, I would avoid *multiple* inheritance (i.e. where a class inherits from more than one parent class) if possible as that can...

> What about using mixins? They seem like an obvious solution for the problems of similar or identical methods between some - but not all - kernel classes. Yes, a...

While having this information available would be great, I'm not that keen on the idea of depending on Fab to do it. Perhaps we could have some mechanism that says...

OK, so the crux of it is that we need to re-write the tests to purely cover the corresponding classes. As Sergi points out, the implementation of these is short...

Using `master` of PSyclone with this test (thanks for supplying it) now, we raise the correct TransformationError: ``` Transformation Error: The supplied node is passed as an argument to a...