How to handle [possible] negative VM refcounter

[AnnaShaleva]

Summary or problem description
Due to a possible bug in the node implementation, VM refcounter may become negative (e.g. like it happened in #3301 (comment)). We need to decide how to handle negative refcounter.

Quick suggestions were:

1. FAULT the transaction immediately if refcounter becomes negative (implemented in [Neo VM Bug]Fix negative counter issue #3304).
2. Stop the node immediately if refcounter becomes negative (implemented in [Neo VM Bug] Kill the process on negative reference counter #3324).
3. Adjust the logic of references counting (implemented in [Neo VM Bug] Fix negative counter issue #3327).

We need to discuss and choose the best solution (or think about other possible solution) how to deal with negative refcounter.

Where in the software does this update applies to?

• VM

[shargon]

I think that 3 is the best solution for now, we need to check it deeply, but the main problem it was not DeepCopy, it was when is removed a CompundType that contains entries that was already present in the stack, the current implementation removed all of them, and it require a check for all the entries.