http://d.puremagic.com/issues/show_bug.cgi?id=5906



--- Comment #10 from bearophile_hugs@eml.cc 2013-03-07 19:25:12 PST ---
(In reply to comment #9)

> struct S(int a)
> {
>     void fun(int b) in { assert(a != b); } body {}
> }
> 
> void main()
> {
>     foreach(i; TypeTuple!(1, 2, 3, 4))
>     {
>         auto s = S!i();
>         if (i != 4)
>             s.fun(4);
>     }
> }
> 
> This code would error 'cannot call S.fun with b == 4' etc even though S.fun never actually gets called with 4.

I see, thank you for the answer.
If the pre-condition analysis (constant folding) is done after a normal step of
dead branch removal, then maybe that error will not be shown.

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------

http://d.puremagic.com/issues/show_bug.cgi?id=5906



--- Comment #11 from bearophile_hugs@eml.cc 2013-03-07 19:29:40 PST ---
(In reply to comment #10)

> If the pre-condition analysis (constant folding) is done after a normal step of dead branch removal, then maybe that error will not be shown.

Currently it doesn't happen, and this generates an error (you need a "static if" to make the error go away):


import std.typetuple;
void main() {
    int[4] a;
    foreach (i; TypeTuple!(1, 2, 3, 4)) {
        if (i != 4)
            a[i]++;
    }
}

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------

http://d.puremagic.com/issues/show_bug.cgi?id=5906



--- Comment #12 from yebblies <yebblies@gmail.com> 2013-03-08 14:33:58 EST ---
(In reply to comment #10)
> 
> I see, thank you for the answer.
> If the pre-condition analysis (constant folding) is done after a normal step of
> dead branch removal, then maybe that error will not be shown.

> Currently it doesn't happen, and this generates an error (you need a "static if" to make the error go away):

That is an interesting point.

I seriously doubt that would be enough anyway, I think you would need full flow analysis to work out which paths can actually be taken.

Another good example is this:

int div(int a, int b)
in
{
    assert(b != 0, "Division by zero!");
}
body
{
    return a / b;
}

void main(string[] args)
{
    auto i = 3 / 0; // Fails at compile time
    auto j = div(3, 0); // Would fail at compile time if this was implemented
}

So maybe it is acceptable.

-- 
Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email
------- You are receiving this mail because: -------