Support for e2e tests in cairo with devnet

[cptartur]

On Starknet, when calling contracts using starknet_call RPC method (without an account) info methods like get_caller_address etc. return default valeus e.g. 0. There is no transaction and block context.

However this is impossible to do in Starknet Foundry. It doesn't matter if the method accepts @ContractState or ref ContractState, we are always executing the method in cheated context - all cheatcoded values are present,get_caller_address returns non 0 value etc.

It should be possible to emulate starknet_call in Foundry. We should either change the behavior of Dispatchers or introduce a separate method for just calling contracts.

[MaksymilianDemitraszek]

It should be possible to emulate starknet_call in Foundry.

I think it shouldn't, at least in the standard test target. Standard test target and dispatchers should operate on contract execution layer not on the transaction layer. Why? Because of use-cases like account testing (transaction layer assumes existence of an account) and testing contract designed to be called from other contracts.

We could consider introducing additional target #devnet-tests (name TBD) which could provide call and invoke dispatchers and take targetted devent url as argument which those be executed on.

[cptartur]

I think it should be at least noted somewhere in the docs that calling in forge never behaves like starknet_call rpc method because it's not obvious. And I think protostars call did behave like an rpc call, so it may be confusing for the users migrating from that tool.

We could consider introducing additional target #devnet-tests (name TBD) which could provide call and invoke dispatchers and take targetted devent url as argument which those be executed on

We should definitely support that somehow.

[MaksymilianDemitraszek]

And I think protostars call did behave like an rpc call, so it may be confusing for the users migrating from that tool.

Cairo 1 Protostar did, Protostar 0 didn't

[Arcticae]

Isn't this related to #18 somehow?
We could run deployment scripts on a local node and achieve the behavior you mean @cptartur

[Arcticae]

Moving this to general

[cptartur]

This task is mostly about the whole discussion about the call/invoke architecture in tests as well as stateful tests. Not sure what component it should be attached to, perhaps both cheatnet and forge?

[piotmag769]

Is it still relevant? @cptartur @MaksymilianDemitraszek

[cptartur]

I think it kind of still is? In case we want to add some tests in the future that are not in the invoke transaction context.

[MaksymilianDemitraszek]

Renaming it to be more accurate

[cptartur]

Duplicate of #1033