getOverloads __traits does not seem to include instances of the generic members

[monkyyy]

On Monday, 2 March 2026 at 11:48:09 UTC, user1234 wrote:
> On Monday, 2 March 2026 at 11:23:25 UTC, monkyyy wrote:
>> On Monday, 2 March 2026 at 10:53:30 UTC, user1234 wrote:
>>>
>>> ```d
>>> void v2(T)(T t){}
>>> void v2(){}
>>> alias v2 = v2!int;
>>>
>>>> Error: template instance `v2!int` `v2!int` forward 
>>>> references template declaration `v2(T)(T t)`
>>
>> I think thats the wrong error message but Im not convinced its 
>> wrong to give an error
>>
>> ```d
>> void v1(T)();
>> alias v1= bool;//error
>> ```
>
> For this case the compiler can check whether that's for a 
> function... oh crap we also have eponymous templates...
>
> Really the problem is that we cannot grow the set, while the v0 
> and v1 sets are equivalent. Or maybe in the compiler internals, 
> the set is well updated, problem is that then when you iterate, 
> the count is wrong...
>
> An example more closed to a real use-case:
>
> ```d
> void v3(T)(T t){}
> void v3(){}
>
> enum IncludeTemplate = true;
> static if (__traits(compiles, mixin("v3(0)"))) // should add an 
> instance to the set
> {
>     pragma(msg, "yeah were are in the branch...");
>     static assert (__traits(getOverloads, mixin(__MODULE__), 
> "v3", IncludeTemplate).length == 3);
> }
> ```
>
> According to me the test on the mixin should have for effect to 
> update the overload set.

Overload sets are headers not initializations and the trait is 
basicly useless

```d
import std;
static struct S{
     static void v4(T)(){}
}
unittest{
     S.v4!int;
     S.v4!bool;
     S.v4!float;
     __traits(getOverloads,S,"v4").stringof.writeln;//guess what 
this prints
}
```

The compiles mixin *shouldnt* have an effect(it would tho) it in 
theory suppose to unwind its effects