Re: [Vala] Change in 0.23.1 for array ownership and .length parameter

[Luca Bruno <lethalman88 gmail com>]

On 07/01/2014 06:46, Tal Hadad wrote:
> This case trigger me a question I wanted to ask before.
> Why transforming ownership is nulling the original variable?
>
> Instead of nulling, maybe just change variable to behave as unowned.
>
> You might say that there is a problem in my solution, like this code:
>
> uint8[] ar = new uint8[10];
> if (some_method() != NULL)
>      another_method((owned)ar);
> ...
>
> Since after the "if" case, valac can't determinate if ownership was given(at compile time).
>
> Despite of this, nulling "ar" variable is even worse in my opinion and it's error prone.
>
> The user might use "ar" varible later and it will crush he's application.
> If he was lucky enough, he would realize that ar is null.
> It's not easy to find that this line causing it.
>
> I suggest solution for the last problem from valac side - treat ar as unowned after the
> (owned) operator, even if this operator was in "if" block!
>
> That way, "ar" won't be null, and the user could use it later only as unowned.
>
> And for class/struct fields - I do accept that they must be nulled when transferring ownership.

What about:

if (foo) {
  bar = (owned) ar;
}
// what now

This behavior is confusing.