New (C) function syntax about gmqcc HOT 7 CLOSED

I'd like to also add an "auto" keyword for function syntax, which has the compiler infer the type (if possible). This would be a gmqcc only feature and would error in any other standard.

from gmqcc.

Doable. How about 'auto' only for function return-types, and let for variable declarations (or var, see #4, we just need something)

auto foo() {
    let e = self.attacker; // an entity
    e.foocount++;
    return e;
}

Thoughts?
(We should open a separate ticket for this probably)

from gmqcc.

My problem with the let keyword is it has a certain instilling characteristic of a typical lambda construct. So using is as a variant type like this seems a bit intrusive (if we have auto hinting the compiler to infer, as well as let). I'd imagine they'd do the same thing anyways.

from gmqcc.

Right, the above code should definitely use auto e = self.attacker. However when using struct types which have not yet been declared, auto might not be the best choice. var would just make more sense, and it already exists in original QC.

auto foo() {
    auto e = self.attacker;
    e.foocount++;
    return e;
}

auto foo() {
    var AStruct x = self.somestruct;
    if (x.something) whatever();
} // foo becomes 'void()'

struct AStruct {
    finally the type gets defined
}

from gmqcc.

new-syntax branch contains the new implementation - I rewrote the whole field/global parsing code. It's much cleaner and unified. It's not "strict" enough yet though (ie. doesn't error properly on new-style functions with -std=qcc, but it is ready to be merged once we're aiming for 0.2)

from gmqcc.

now errors on -std=qcc, and also contains the var keyword to create global field-pointers without creating fields, AND fixes a bug which caused global/local vector field pointers to only have 1 component generated. Now behaves correctly, and fteqcc-compliant, whereas -std=gmqcc autoenabled -fadjust-vector-fields

from gmqcc.

Closing this - keep in mind it's not merged to master yet.

from gmqcc.