I have some experimental code I'm writing, and it seems when i return a structure it is either corrupted, or should be disallowed all together. A postblit is suggested this(this), but since I'm returning a structure that isn't tied to anything else it doesn't need to duplicate it.

Example:

struct A{
int b;
char c[];
}

void fun1() {
  A a;
  a = fun2();
//array a.c corrupted, causes range exception
//int a.b corrupted?!?
}

A fun2(){
  A t;
  t.c.length = 100;
/*do something*/
  return t;  //valid before returning.
}

Era Scarecrow:

>   I have some experimental code I'm writing, and it seems when i return a structure it is either corrupted, or should be disallowed all together. A postblit is suggested this(this), but since I'm returning a structure that isn't tied to anything else it doesn't need to duplicate it.

If possible please show a complete minimal program that contains a main() that shows your problem. I have written this, but I don't see the problem:

import std.stdio: writeln, write;

enum int N = 100;

struct Foo {
    int x;
    char[] arr;
}

Foo fun2() {
    Foo f1;
    f1.arr.length = N;
    f1.arr[] = 'x';
    f1.x = 25;

    writeln(f1.arr.length);
    write(">");
    foreach(i; 0 .. N)
        write(f1.arr[i]);
    writeln("<");
    return f1;
}

void main() {
    Foo f2;
    f2 = fun2();
    writeln(f2.x);
    writeln(f2.arr.length);
    write(">");
    foreach(i; 0 .. N)
        write(f2.arr[i]);
    writeln("<");
}

Bye,
bearophile

== Quote from bearophile (bearophileHUGS@lycos.com)'s article
> Era Scarecrow:
> >   I have some experimental code I'm writing, and it seems when i return a
structure it is either corrupted, or should be disallowed all together. A postblit is suggested this(this), but since I'm returning a structure that isn't tied to anything else it doesn't need to duplicate it.
> If possible please show a complete minimal program that contains a main() that
shows your problem.

 I know i've been going a different route with the code since so i am not sure
where exactly it was, and can't duplicate the problem now.

 However single-lining through it brings up a question of assumption. Aren't
array's (dynamic and otherwise) filled with 0's by default?

Str_BigNum add(in Str_BigNum rhs) {
	Str_BigNum tmp;
	char res[];
	res.length = _result.length + 1;
	res[1 .. $] = _result[];
//res[0] should be null, right??
	_addsub!("+")(res, rhs._result);
	tmp._result = res;
	return tmp;
}

On Fri, 27 Aug 2010 16:04:50 -0400, Era Scarecrow <rtcvb32@yahoo.com> wrote:

> == Quote from bearophile (bearophileHUGS@lycos.com)'s article
>> Era Scarecrow:
>> >   I have some experimental code I'm writing, and it seems when i  
>> return a
> structure it is either corrupted, or should be disallowed all together. A postblit
> is suggested this(this), but since I'm returning a structure that isn't tied to
> anything else it doesn't need to duplicate it.
>> If possible please show a complete minimal program that contains a main() that
> shows your problem.
>
>  I know i've been going a different route with the code since so i am not sure
> where exactly it was, and can't duplicate the problem now.
>
>  However single-lining through it brings up a question of assumption. Aren't
> array's (dynamic and otherwise) filled with 0's by default?

No, arrays are filled with the .init value.  For char, that's 0xff.

-Steve

== Quote from Steven Schveighoffer (schveiguy@yahoo.com)'s article
> > array's (dynamic and otherwise) filled with 0's by default?
> No, arrays are filled with the .init value.  For char, that's 0xff. -Steve

 Ahh that's right. It's 0xFF which is a illegal character for UTF-8 encoding. That
does help explain why my asserts involving 0 to 9 kept failing, since my checks
were for 0x0, and 0x20.

 If i happen to duplicate the original problem i'll post back again.