using `x` as a field name causes an compilation error

[Y-Nak]

When x is used in the contract field, the compiler emits an error.

PoC

Code:

import std;

contract PoC {
    x : word;
}

Error:

tcExp not implemented for: this.x
FieldAccess Nothing x
 - in:Typedef.rep(this.x)
 - in:function toWord (x : ?$5999) -> word {
   return Typedef.rep(this.x);
}
 - in:?$5999 : Add, ?$5999 : Sub, ?$5999 : Bounded, ?$5999 : Eq, ?$5999 : Ord, ?$5999 : Typedef (word) => default instance ?$5999 : Num {
   function maxVal () -> ?$5999 {
      return Bounded.maxVal();
   }
   function toWord (x : ?$5999) -> word {
      return Typedef.rep(this.x);
   }
   function fromWord (x : word) -> ?$5999 {
      return Typedef.abs(this.x);
   }
   function add (x : ?$5999, y : ?$5999) -> ?$5999 {
      return Add.add(this.x, y);
   }
   function sub (x : ?$5999, y : ?$5999) -> ?$5999 {
      return Sub.sub(this.x, y);
   }
   function gt (x : ?$5999, y : ?$5999) -> bool {
      return Ord.gt(this.x, y);
   }
}

If x is renamed to x1, for example, the compilation succeeds.

[mbenke]

This seems similar to #237: if x is a contract field then occurrences of x in library functions get incorrectly replaced by this.x.

[mbenke]

Minimal reproducer without importing the whole std:

forall self fieldType offsetType.
class self:StructField(fieldType, offsetType) {}

data StructField(structType, fieldSelector) = StructField(structType);
data ContractStorage(cxt) = ContractStorage(cxt);
data storage(t) = storage(word);

function foo(x:word) -> word { return x; }

contract PoC {
    x : word;
}

[mbenke]

It seems #279 was not properly merged, main now misses a crucial check in ElabTree:

        isVar <- isDefinedVar n
        if isF && not isVar then do

and has the old and buggy version instead:

        if isF then  -- TODO: check if not shadowed
          pure $ FieldAccess me' n

However, fixing this does not resolve the issue, since isDefinedVar checks the variables field of Env, which holds type variables.

Anyway, elaboration probably shoudl not introduce FieldAccess outside of contract.

@rodrigogribeiro can you have a look at this?

[rodrigogribeiro]

@rodrigogribeiro can you have a look at this?

On it.