Fix named error parameter encoding order in IR codegen

[nikola-matic]

Closes #16452

[cameel]

This needs a lot more tests and a bug list entry.

[cameel]

As a part of due diligence for the fix we must make sure that similar functionality in other places is not affected. In this case this means checking named parameter support in language constructs. I'd especially verify events, as these share a lot of code with errors, but are more:

• emitting events
• function calls
  • internal
  • external
  • library
  • bound
  • virtual
    • Especially when the names in the base function are in a different order.
  • getter
• bare calls
• struct constructor invocations
  • https://github.com/argotorg/solidity/blob/develop/test/libsolidity/semanticTests/structs/struct_named_constructor.sol

For these named parameters are not allowed, but they should be still covered with (failing) tests using them:

• builtin calls
  • ABI encoding builtins
• inherited constructor calls
• modifier invocations
• calls via function pointers
• type conversions (these are not calls, but look the same to the parser)

Is there anything else I'm missing?

[nikola-matic]

A couple of questions:

1. When you say getters - do you mean auto generated ones, and how would named parameters fit here? My assumption is that getters don't take any parameters?
2. What do you mean by bare calls? call, staticcall, delegatecall? Can you give an example?

[cameel]

My assumption is that getters don't take any parameters?

They do on array types. Though, now that I think of it, I'm not sure we support named arguments there. Please check in any case.

What do you mean by bare calls? call, staticcall, delegatecall?

Yes. Not sure what the name of the data argument is (and if we actually support named arguments there) but it would be something like this:

a.call({data: ""});