Andy Porter
Andy Porter
The problem is in the Routine handler in the Fortran backend: we flatten all tables by populating a brand-new one. However, the brand-new one is not (initially) connected into the...
Finally got all of the integration tests to pass. Really ready for review now!
CodeCov really isn't happy so I've checked the coverage for `symbol_table.py` locally and all is fine. I've tried firing off the CI again so maybe CodeCov will sort itself out....
Should be ready for another look now, integration tests permitting.
Hi @JulienRemy, I can't remember what you said about PRs on Monday. Was this the one that's ready for a review?
Integration tests were green.
Thanks @tommbendall. Those kernels that "loop to some depth in the halo regions" - is that depth set at compile-time or run-time? To summarise, we have: * Kernels that write...
Currently, the docs for `operates_on` say:  We need to extend this to support the two further cases given above. Really, `CELL_COLUMN` is effectively `OWNED_CELL_COLUMN` and then we would naturally...
Thanks Tom. Yes, your last two options could be captured in meta-data but that would require more engineering. However, if they are going to be the bulk of the use...
Right, so now I have:  and  does that look OK @tommbendall ?