Maybe i´m tired already, but whats wrong here:

pragma(msg, __traits( compiles, mixin("int x") ) );
//output: false

On Wednesday, 25 October 2017 at 19:12:02 UTC, SrMordred wrote:
> Maybe i´m tired already, but whats wrong here:
>
> pragma(msg, __traits( compiles, mixin("int x") ) );
> //output: false

Or the original case I found:

struct T{}

pragma(msg, __traits( compiles, T() ) ); //true
pragma(msg, __traits( compiles, mixin("T()") ) ); //true
mixin( "T();" ); Error: `structliteral` has no effect in expression `T()`

On Wednesday, 25 October 2017 at 19:12:02 UTC, SrMordred wrote:
> Maybe i´m tired already, but whats wrong here:
>
> pragma(msg, __traits( compiles, mixin("int x") ) );

You are missing a ;

The mixin must compile as a full thing in context. Variable declarations need the ; to be complete.

On Wednesday, 25 October 2017 at 19:25:01 UTC, Adam D. Ruppe wrote:
> On Wednesday, 25 October 2017 at 19:12:02 UTC, SrMordred wrote:
>> Maybe i´m tired already, but whats wrong here:
>>
>> pragma(msg, __traits( compiles, mixin("int x") ) );
>
> You are missing a ;
>
> The mixin must compile as a full thing in context. Variable declarations need the ; to be complete.

it returns false anyway.

void F(){
  writeln("inside");
}

struct T{}

void main()
{
        pragma(msg, __traits( compiles, mixin("int x;") ) );//false

	pragma(msg, __traits( compiles, mixin("F();") ) );//false
	pragma(msg, __traits( compiles, mixin("F()") ) );//true
	mixin( "F();" ); //compiles
	
	pragma(msg, __traits( compiles, mixin("T();") ) ); //false
	pragma(msg, __traits( compiles, mixin("T()") ) ); // true
	mixin( "T();" ); // Error: `structliteral` has no effect in expression `T()`
}

On Wednesday, 25 October 2017 at 19:21:27 UTC, SrMordred wrote:
> mixin( "T();" ); Error: `structliteral` has no effect in expression `T()`

The semicolon there indicates it is a complete statement that does nothing, and that's no error.

If there's no ;, it is just an expression that must be somewhere else - and the compile error is deferred until higher in the call chain.

> The semicolon there indicates it is a complete statement that does nothing, and that's no error.

so why this line resolves to false?
void F(){}
pragma(msg, __traits( compiles, mixin("F();") ) );//false

On Wednesday, 25 October 2017 at 19:50:31 UTC, SrMordred wrote:
> so why this line resolves to false?

Because it is illegal to put a statement or declaration inside __traits(compiles). sorry, I should have said that before... even though the mixin can be legal in another context, it won't be in the __traits context due to this:

https://dlang.org/spec/traits.html#compiles

"Returns a bool true if all of the arguments compile (are semantically correct). The arguments can be symbols, types, or expressions that are syntactically correct. The arguments cannot be statements or declarations. "


When there's a closing ;, it is a mixin statement.

> void F(){}
> pragma(msg, __traits( compiles, mixin("F();") ) );//false

On Wednesday, 25 October 2017 at 20:04:47 UTC, Adam D. Ruppe wrote:
> On Wednesday, 25 October 2017 at 19:50:31 UTC, SrMordred wrote:
>> so why this line resolves to false?
>
> Because it is illegal to put a statement or declaration inside __traits(compiles). sorry, I should have said that before... even though the mixin can be legal in another context, it won't be in the __traits context due to this:
>
> https://dlang.org/spec/traits.html#compiles
>
> "Returns a bool true if all of the arguments compile (are semantically correct). The arguments can be symbols, types, or expressions that are syntactically correct. The arguments cannot be statements or declarations. "
>
>
> When there's a closing ;, it is a mixin statement.
>
>> void F(){}
>> pragma(msg, __traits( compiles, mixin("F();") ) );//false

Oh, now thats explains. I thought that a "mixin statement" was equal to the argument type that it compiles to. Thanks!