Mirror of forum.nim-lang.org

7005 :: Subrange check during initialisation

[2020-10-29T09:02:42+01:00]

michy (orginal) [2020-10-29T09:02:42+01:00] view original
This compiles and prints 0 :
```
type DiceVal = range[1..6]
var d : DiceVal
echo d                   # -> 0
```
Obviously the 0-initialisation is not checked. Compiler bug?

haxscramper_ (orginal) [2020-10-29T09:05:34+01:00] view original

Not this is not a bug. All objects are zero-initialized upon construction.

Stefan_Salewski (orginal) [2020-10-29T09:14:22+01:00] view original

It is a bug. I have seen a similar discussion some years ago for enums I think. Maybe we have a needsInitialization pragma which can catch this, but I can not find it yet.

haxscramper_ (orginal) [2020-10-29T09:17:50+01:00] view original

No, currently this is just how default initialization is implemented. See RFC #252 and RFC #126, specifically this comment

xigoi (orginal) [2020-10-29T09:35:09+01:00] view original

I swear I did something like this not so long ago and Nim did complain.

lscrd (orginal) [2020-10-29T10:18:00+01:00] view original
The check is done for the implicit variable “result” and we get the usual warning: Warning: Cannot prove that 'result' is initialized. This will become a compile time error in the future. [ProveInit]

Why not for a normal variable?
```
type DiceVal = range[1..6]

proc throwDice(): DiceVal =
  discard

echo throwDice()
```

Araq (orginal) [2020-10-29T12:24:36+01:00] view original
The compiler currently doesn't try the init property for global variables.

This produces the warning:
```
type DiceVal = range[1..6]
proc m =
  var d: DiceVal
  echo d

m()
```