`accept` analysis improvement

[anton-trunov]

There is an analysis of accept statements in the Scilla codebase (see Accept.ml). The analysis checks

• if there is at least one accept statement in the contract;
• if there are code paths with more than one accept statement in them.

We can improve accept-checking in Scilla. For example, to remove false positives in case of non-native token contracts we can check that all send statements have the _amount message field equal to zero and turn this warning off in this case. (This is not going to be precise, but for many practical contracts it's easy to establish the _amount field is zero). See, for instance, https://github.com/Zilliqa/ZRC/blob/main/reference-contracts/FungibleToken.scilla

[jubnzv]

For example, to remove false positives in case of non-native token contracts we can check that all send statements have the _amount message field equal to zero and turn this warning off in this case. (This is not going to be precise, but for many practical contracts it's easy to establish the _amount field is zero)

How accurate should such an analysis be?

If we just use a zero constant, it could be trivially implemented. But in more complex cases we need some kind of constant propagation and maybe some other dataflow analyses.

[anton-trunov]

@jubnzv It's usually a Uint128 0 directly or just a constant that unfolds to Uint128 0.