Incorrect warning: nonexistent __x variable is shadowed

[brainiacfive]

I get a compiler warning

The definition of `__x` shadows an older definition at (builtin location).

I am not using __x in my source. Tried to eliminate the String include from it, only Option left, but it throws the same warning; so it would not be include conflict. No other includes.

Looks like a compiler issue to me. Will provided a minimal source example later. Maybe you have an idea already from this stub notice.

[brainiacfive]

(builtin location)

is verbatim the warning, not a placeholder I inserted.

[marc0olo]

@brainiacfive this is happening in a (currently) private contract, right?

if you can't provide a minimal example right now you could maybe invite @ghallak and/or @radrow to the repo to check with the contract where the compiler currently produces this warning

[hanssv]

I think the compiler uses the variable __x when it desugars some of the Map-syntax - could be something 'interesting' going on there.

[hanssv]

contract C =
  record r1 = { f : bool, g : bool }
  record state = { m : map(int, r1) }

  let default1 = { f = false, g = false }

  entrypoint init() = {m = {}}

  stateful entrypoint f(the1 : int) =
    put(state{ m[the1 = default1].f = true })

is a small example capturing the same problem - the compiled code is correct in this case, but that might be luck (cc @radrow and @ghallak )

[brainiacfive]

Much appreciated. I saw more warnings using __x meanwhile; in folds.

Do you know which variable in your example is the first use of __x that would get shadowed by which other variable that is the second use? I'd be checking for patterns of that in the contract source to make sure it is harmless.

[radrow]

It's a bug caused by running code analysis on desugared code. Nothing to worry about. We'll take a look and fix it soon.

[radrow]

If you are worried though, or if it may impact something serious, you can use aesophia_cli with pp_asm to check if the function is compiled correctly. The assembler should be fairly readable.

[brainiacfive]

Thank you, will do!

[brainiacfive]

The asm code is fascinating but not something to crack on the fly. Can you @radrow @hanssv share what sugar is causing the use of __x so I can change the code to avoid it? I assume a nesting of scopes is a prerequisite for the effect to happen. So I should be able to eliminate the select places this could be happening. I'd rather have this evened out, thanks!

[brainiacfive]

Tried with replacing

put(state{ used[account = unused].abstraction = true })

with

let u = used[account = unused]{ abstraction  = true }
put(state{  used[account] = u })
              

On @hanssv recommendation, but so far no luck.

[hanssv]

You need to change all five of them, and indeed it does workaround the problem.

[hanssv]

I think the concrete issue has been worked around - as for the compiler, I think using __x everywhere might be a bit too simplistic 🙈 - and it should be fixed eventually.