Is that a bug?

The code below returns:

"main.A!(double).A"

if we uncomment "auto c=make!A(1.0);", we get:

"main.a!(double).A"

----
module main;
import std.stdio;
class A(T){
	T x;
	this(T x){this.x=x;}
}
auto make(alias a,T...)(T args){
	return new a!T(args);
}
void main(){
	//auto c=make!A(1.0);
	auto b=new A!double(1.0);
	writeln(typeid(b));
}
----

If this behavior is not a bug, I wonder what's the rationale, but also, how can we have such a "make" function that doesn't mess up the template class name depending on the order of which is used first (the make!T variant or the new A!T) ?

On 09/16/2012 05:08 PM, timotheecour wrote:
> Is that a bug?
>
> The code below returns:
>
> "main.A!(double).A"
>
> if we uncomment "auto c=make!A(1.0);", we get:
>
> "main.a!(double).A"
>
> ----
> module main;
> import std.stdio;
> class A(T){
> T x;
> this(T x){this.x=x;}
> }
> auto make(alias a,T...)(T args){
> return new a!T(args);
> }
> void main(){
> //auto c=make!A(1.0);
> auto b=new A!double(1.0);
> writeln(typeid(b));
> }
> ----
>

This bug must be related:

  http://d.puremagic.com/issues/show_bug.cgi?id=8579

Other bugs that are linked from that one mention an internal hash table. Apparently different objects resolve to the same key, and the value of the first key gets used from that hash table.

I don't know whether Kenji Hara's fix for the bug above also fixed this issue. I say, open a bug anyway.

Ali

done: http://d.puremagic.com/issues/show_bug.cgi?id=8674