[enh] fix issue 10491 - Type inference for function arguments with default value

[hpohl]

http://d.puremagic.com/issues/show_bug.cgi?id=10491

void foo(i = 10, immutable f = 10.0f, const d = 10.0)
{
    static assert(is(typeof(i) == int));
    static assert(is(typeof(f) == immutable float));
    static assert(is(typeof(d) == const double));
}

struct S(T) { T val; }

void bar(T)(x = S!T("str"))
{
    static assert(is(typeof(x) == S!T));
}

[9rnsr]

I have negative opinions against doing this. See my comment in bugzilla.
http://d.puremagic.com/issues/show_bug.cgi?id=10491#c2

[hpohl]

I'd like to hear what @andralex or @WalterBright think about this before we close this down.

[andralex]

@hpohl it's a nice idea but on the wrong side of the cost/benefit analysis. Will close this now. Keep them coming though!

[ntrel]

This feature is now pre-approved BTW (so we can use it for const e = Enum.member without verbosity):
https://issues.dlang.org/show_bug.cgi?id=17186

I don't think void foo10491(a = 10, b = null) should be allowed, a storage class should probably be required.