Simon Krajewski

The error literally tells you where to look though: `source/Main.hx:3: characters 13-17 : Redefinition of variable code in subclass is not allowed. Previously declared at haxe._Exception.NativeException`. However, I suppose we...

> The correct behaviour is for haxe to call haxelib once per compilation to resolve all libraries at once, which gives haxelib the chance to complain if library versions are...

I imagine that in that case whoever provides the compiler with the arguments will have to check this. That's basically what this `HighLevel` module is about anyway, it's the pre-processing...

> This already doesn't happen. Within the context of a single hxml, as soon as haxe encounters a `-lib`, it inserts the flags for all libraries within that hxml: https://github.com/HaxeFoundation/haxe/blob/b0b299781a22cb3ad1fc8460d7cd3af9530a22e4/src/compiler/compiler.ml#L656C21-L656C41....

I think that would still not be a general solution because a `-lib` can lead to more `-lib` after the haxelib call.

FWIW I'm in favor of this change and think we should move forward with this. I think we all agree that this should be done, the only remaining question is...

> Caused by [#12049](https://github.com/HaxeFoundation/haxe/pull/12049) That PR just made it more consistent across targets. I don't know why we would want the native _exception_ type to be `String` because that's not...

A simple `trace(e == e.native)` will tell you that the answer is yes, they are identical.

Circular references are not inherently bad. Many graphs have them, if not reflexively like here (node connected to itself) then by transitive property (A -> B -> A). You can...

I have implemented a sort of low-level global import in https://github.com/HaxeFoundation/haxe/commit/0008ff9ad3e72663b87b47dbc4c670132135c7cf, but I don't particularly like that. Will have to think about how to handle this.