Static inline variables can be initialized to non-constants again

[Simn]

class Main {
    static inline var test = [1];
    static public function main() {
        test;
    }
}

That should be forbidden (again).

[nadako]

I wonder if it's possible to write unit tests on compiler errors for that?

[ncannasse]

in macros you should be able to try/catch a Context.typeOf, but that would only work for exception. So it would require to create a set of "should error" cases and run the haxe compiler on each.

@Simn maybe that's a good idea for the long term Haxe dev

[Simn]

Yes, something like that could go to tests/errors. We'll have to make sure that we don't get just any error, but the correct one.

We could have a distinct directory per test-case which has a .hxml and some code in .hx files. A (macro?) application could then read these directories individually and invoke the Haxe compiler on the .hxml with the cwd set correctly.

It should be possible to encode expected error messages into the .hxml as well, e.g. as `-D error_inline1=Inline variable initialization must be a constant value". The running application can then compare these with the actual output and report differences accordingly.

[Simn]

On the actual subject, does this look correct to you?

https://github.com/Simn/HaxeManual/blob/development/assets/graphics/var-initialization.png

[ncannasse]

looks ok, but shall we allow to access member/super class variables in initialization code ?

[Simn]

Uhm, you were the one who didn't want that.

[ncannasse]

Yes, I wanted to confirm that it's not entirely "anything" at the bottom right of the schema ;)

[Simn]

That's the static = yes branch. ;)

[ncannasse]

Ah, right sorry. There's no "this" access in the static branch either, hence the confusion