#[cube] impl blocks do not use spans to show where problems are

[torsteingrindvik]

Example:

If possible spans from the proc macro should be used such that:

1. The problematic method is highlighted
2. (if possible) the problematic statement is highlighted

It can get difficult to know which method actually causes compile problems when doing larger refactors.

[nathanielsimard]

When the cube function compiles, but not the expand function, it's hard to display good error messages! Not sure what we can do about it except try to bring those errors in the cube functions.

[torsteingrindvik]

When the cube function compiles, but not the expand function, it's hard to display good error messages! Not sure what we can do about it except try to bring those errors in the cube functions.

Hm- this sounds like a situation that shouldn't happen then.
Could it then be considered a bug that the cube function compiled in such a case?

[nathanielsimard]

There are stuff that we can't detech from the cube function only. We rely on the type system to also validate the expand function. Things like does the expand method exist, do comptime branches return the same type, etc.