Communication between the GUI and the simulation involves a FIFO in each direction and will probably grow to involve some shared memory too.
Initially it supports 3 use cases:
- Simulation started from command line connects to existing GUI
- Simulation started from command line starts new GUI and connects to it
- Simulation started from within GUI
At present this is done in a fairly ad hoc way. It would be a good idea to think it through and make sure it can do everything we want robustly without any risk of deadlocks or other problems.
I can see all three as being useful. In terms of importance though, for debugging the sim itself, the first option is very useful, with option 2 as a handy thing for when things go wrong.
Option 3 is probably where we are aiming for for a "user product"
Ability to use other GUIs, either from the command line or otherwise is probably something we will want in the future.
We will certainly want the GUI to talk back to the sim in a reliable way.
I will think some more