opCall/ctor partially sorted out

[Artur Skawina]

On 10/07/12 04:24, bearophile wrote:
> Recently one of the most important bugs was mostly fixed, beside Win64 support this is one of the most important changes in dmd 2.061:
> 
> http://d.puremagic.com/issues/show_bug.cgi?id=6036
> 
> 
> Do you think this has to be correct code?
> 
> struct Adder {
>     int v;
>     int opCall(int x) { return x + v; }
> }
> void main() {
>     auto a = Adder(5);
> }

Yes, with resulting a.v==5.

This should result in an int == 42:

    auto a = Adder(37)(5);

'static' either needs to be handled properly or another op needs to be
introduced - opStaticCall(). Which might look like a hack, but doing it
like that may be simpler to implement and relatively harmless (as the 
aggregate.op* namespace has to be treated as reserved in practice).
Also, the type and instance methods shouldn't form an overload set, so
separating them is a good idea (a matching non-static opCall must always
take precedence).

Calling the static-opCall (either opStaticCall or the "normal" 'static
opCall', if/when that one works properly) with an instance should work.
(not doing this would need changes to how the null-checking is done
and likely cause other problems that i can't think of right now)

artur