Interaction between comptime var and runtime struct creation

[justin-geopipe]

Zig Version

0.11.0 (playground)

Steps to Reproduce and Observed Behavior

This code runs without error:

pub fn main() !void {
    comptime var MyType: type = i32;
    const MyStruct = struct { v: MyType };

    //_ = MyStruct { .v = 5 }; // line A

    MyType = [2]f32;
    _ = MyStruct { .v = [2]f32 {1.4, 2.4} }; // line B
}

However, if you uncomment "line A", then "line B" errors. That is, the introduction of the runtime code _ = MyStruct { .v = 5 }; changes the type of MyStruct.

Expected Behavior

My guess as to what's going on is that this is an interaction between mutation (comptime var) and laziness. Mutation and laziness should never mix, so I'd expect either:

• Prevent the laziness: the above code (with line A still commented out) produces a type error on line B, because MyStruct contains an i32 at the time it is defined.
• Prevent the mutation: the above code (with line A still commented out) produces an error saying that comptime var is not allowed in type definitions.

Related to #7396, though so far as I've seen this isn't exactly covered by any of the other issues it cites. For example, there are no closures involved.

[mlugg]

This design issue is solved on master. The value of MyType is resolved eagerly at the point of the struct declaration, so the same error occurs regardless of whether line A is commented.