Track parser position for error reporting

[joelreymont]

No span tracking is being done, just position tracking for error reporting.

[joelreymont]

Why do tests keep failing in CI?

[Hejsil]

Why do tests keep failing in CI?

No idea. I'll take a proper look this weekend 🙏

[Hejsil]

Why do tests keep failing in CI?

You have these lines:

var res: Res = undefined;
var pos: usize = 0;
res.ok.rest = str;

This is illegal behavior. You're initializing Res to undefined and then accessing through the union with res.ok. res.ok is not the active field. Nothing is the active field.

I don't see why you need to initialize a Res right here. Just keep track of Res.Ok

var res: Res.Ok = undefined;
var pos: usize = 0;
res.rest = str;

[joelreymont]

I wonder if this is a compiler regression because my 2 Macs have different versions of the Zig master and they haven't caught this.

[joelreymont]

Is anything else missing here?

Can you accept the PR?

[Hejsil]

Is anything else missing here?

Can you accept the PR?

There are some changes I want to make, which have ended up taking a bit. Hopefully this will be merged in some modified form within this week

[joelreymont]

What changes are you planning?

[Hejsil]

My plan is to make Result this:

pub fn Result(comptime T: type) type {
    return struct {
        index: usize,
        value: union(enum) {
            ok: T,
            err,
        },

        pub fn ok(index: usize, value: T) @This() {
            return .{ .index = index, .value = .{ .ok = value } };
        }

        pub fn err(index: usize) @This() {
            return .{ .index = index, .value = .err };
        }
    };
}

Which results in a lot of small changes. There where also some bugs I'm ironing out which hopefully should yield even better error positions (in the json example, the error position is now rarely 0)

[Hejsil]

Thanks for the initial implementation. It has been merged with my modifications: 7100cb3