Referring to a static nil objects's field is impossible, even when it should

[dawkot]

Example

type T = ref object
  x: int

proc f(arg: static T) = discard arg.x

f nil

Current Output

/usercode/in.nim(9, 3) template/generic instantiation of `f` from here
/usercode/in.nim(7, 36) Error: undeclared field: 'x' 

Expected Output

The standard error message for when trying to access a nil ref

Additional Information

$ nim -v
Nim Compiler Version 1.0.99 [Windows: amd64]
Compiled at 2019-10-10

[dawkot]

It's worse than that:

Example

import sequtils

type T = ref object
  x: int

proc foo(arg: static T) =
  static:
    assert arg == nil or arg.x > 0
 
foo nil

Current Output

/usercode/in.nim(10, 5) template/generic instantiation of foo from here
/usercode/in.nim(8, 29) Error: undeclared field: 'x'

[kraptor]

Isn't static T referring to the type of T, which is not an instance and, thus, it has not x defined? (which is correct, in my opinion).

Also, if instead of a proc, you define a macrom then it is catched correctly:

type 
  T = ref object
    x: int

macro f(arg: static[T]) = 
  discard arg.x

f nil

Then you get:

/usercode/in.nim(6, 14)  f
/usercode/in.nim(8, 3) template/generic instantiation of `f` from here
/usercode/in.nim(6, 14) Error: attempt to access a nil address

Afaik both results (undefined for proc and nil address for macro) are correct, in my humble opninion.

[dawkot]

Well, if it's a wont-fix, is there any other way to do compile-time checking on arguments like this?

[Araq]

It's not a wont-fix, it looks like a real bug to me.

[bung87]

should do f cast[T](nil) to make it work.

[metagn]

This is an instantiation bug, the type of static T is inferred as typeof(nil) with value nil, related to #21331, #12552, #4858

Nope, just static not allowing conversions