Also note i've added C++ spans to the NodeGeometrySimulationOutput now to make this a bit nicer.
Downside of putting this in RNA is that we don't have access to the C++ functions like ntree->nodes_by_type there.
Sorry, it's kinda hard to figure out the correct behavior here. I think i've fixed it now by syncing only the output field states. This should work, because the requires_single flag on inputs…
I'm not sure what you mean here. I'm checking the iterator here to avoid comparing it to itself, to prevent fake name collision. Otherwise having an existing item "XY" and entering the name "XY"…
Now this socket seems to be wrong. It should be a field. Currently it would be possible to connect this socket to e.g. a cube node without any visible warning in the ui. ![image](/attachments/…
Am I missing something or is this now breaking simulations completely?
Should be fixed, now geometry sockets should be working again. Other data types just use a dummy state item that doesn't…
Hah, i always forget how this number-to-string stuff works in C++, somehow find it easier to remember old BLI.
Yeah i removed all the copying of state item data in the process of removing the templated types. Could add that back for the geometry i suppose.
Adding a "color" property to SimulationStateItem really doesn't feel like the right way to draw them in the UI. That information should probably only be accessible from the socket types. Could…