A wrong type is reported in a type mismatch error

[auduchinok]

Consider this code:

let a, b: int = ()

The error is produced on the unit expression, but it says that the type is int:

Error FS0001 : This expression was expected to have type
    ''a * 'b'    
but here has type
    'int'

This may be happening due to some expressions rewriting during the analysis.

[brianrourkeboll]

Right, if I remember correctly, the compiler basically uses the type annotation on a binding to wrap the expression being bound in a SynExpr.Typed.

AST

expr =
    SynExpr.Typed(
        expr = SynExpr.Const(constant = SynConst.Unit, range = R("(1,16--1,18)")),
        targetType = SynType.LongIdent(SynLongIdent.SynLongIdent(id = [ Ident("int", R("(1,10--1,13)")) ], dotRanges = [], trivia = [ None ])),
        range = R("(1,16--1,18)")
    ),

It gives the exact same error for this:

let x, y = () : int

But the SynBindingReturnInfo is still available in the AST after that transformation, so we could use that to improve the error message here.

[travv0]

Is it alright if I try my hand at this?

[vzarytovskii]

Is it alright if I try my hand at this?

Absolutely, would be great!

[majocha]

let a, b: int = ()

This declares a destructuring of a primitive int type. Isn't it the actual error that should be reported, regardless of RHS?

A nice error message could suggest for example

Did you mean a, (b: int) ?

[travv0]

Right now it shows the error on the (), which surely can't be right. With my PR it still shows the error, but on the int.

[majocha]

Right now it shows the error on the (), which surely can't be right. With my PR it still shows the error, but on the int.

Yep, that's much better. At least not misleading.
Now I wonder, shouldn't the error be on the invalid a, b expression and not on the type annotation?