Bug 23779 – UFCS using pointer should be allowed and should automatically take a reference on values

This code: ``` struct Data { int value; } void something(Data* data) { data.value++; } void main() { Data d; Data* ptr = &d; ptr.something(); // works (&d).something(); // works, but the syntax looks bad d.something(); // doesn't work, compiler should do what i do above and allow it } ``` It is similar to the request here: https://issues.dlang.org/show_bug.cgi?id=8597 But it's the opposite, if the function is declared by accepting a pointer, then this example should be allowed

The UFCS implementation is already very complex, why would we add another special case? This needs to be a accompanied by strong rationale to have any chances of being accepted. If we were to implement anything, I would rather go the other way around and implement automatic dereferencing with UFCS not automatic address taking. However, I would argue that we should not complicate UFCS any further.

what is requested here is an implict conversion from a `D` to a `D*` based on lvaluness. It is not very well described, and maybe a second attempt would be better, so close as "invalid" for the moment.