Create an explicit immutable and cown region.

[mjp41]

No description provided.

[mjp41]

I need to fix the cown stuff, before this change can land as I have made it stricter, and properly broken valid_01.vpy, rather than just in Asan

[mjp41]

I've created a cown region, and it fixes the Asan issue in the cowns/valid_01.vpy test. But I am not sure if it is the right thing from the point of view of understanding.

Pros

• Naturally captures parenting of the region, into a cown.

Cons

• Looks messy as all the cowns get grouped together

Alternatives

Could create a new region for each cown. This would look nicer, but requires more complex checking.

Questions

• Is Opaque the same as Cown? What other flexibility are we expecting here?
• Should string, true, false, be initially allocated in the immutable region?

[xFrednet]

Is Opaque the same as Cown? What other flexibility are we expecting here?

In our prototype they are currently the same. I added the is_opaque as a separate method to is_cown to better indicate why an access may be invalid. The methods act the same in this prototype, since we can't acquire a cown. But if we ever add something like acquisition to the prototype, the two functions might return different values.

Another thing I could think of would be closed regions. They would be opaque, until they're opened. (This opening could be implicit though)

[xFrednet]

Looks messy as all the cowns get grouped together

We could try not putting cowns into a sub graph, that would have them floating around, like the global objects initially did. Then we can just say that everything outside a "drawn" region is the cown region. Finding a good visualization will probably also require a bit of try and error :D

We shouldn't block this PR on it, I can try to figure something out next week, if you don't want to :D

[xFrednet]

Very nice changes! Makes the whole thing a lot cleaner :D

[xFrednet]

This means that cowns can't be frozen. It makes sense, but we should check that this is inline with what is written up. If not, we probably want to correct the text on this one, as I like this implementation

[mjp41]

I guess I viewed them as already in the "shared" world, so freeze doesn't do anything. A frozen cown can still be mutated in "my view" of the model.

[xFrednet]

Yep, I agree. This was just meant as a reminder to check/sync the docs