Emily

Ok, so if the file uses a builtin directly, it needs a forward declaration, otherwise it doesn't? In this case I don't see why levels of churn would be the...

> My guess is that we should probably permit forward declarations anywhere within the syntax tree This is my guess as well.

> we still need to work out what the semantic integrity check should be for a module with a forward declaration Hm. Isn't it just "however they are serialized"? I...

In my mental model, you can have the same declaration in two different files and then both declarations will be interpreted as the same builtin. How does your mental model...

About semantic hashes – another option is treating forward declarations like implicit lambdas and keep them unevaluated when hashing. If we evaluate them eagerly, we would have to invalidate the...

Okay. @sjakobi @f-f since you ❤️-ed the proposal: * What usecases do you have in mind? * How do you think cases above should be handled? (duplicate forward declarations in...

If we standardize this, I think we should also require that unused builtins don't have to be provided, i.e. `let foo : X in True` should work fine without providing...

Is normal react-icons as tree-shakable as the all files version?11.01.2024 kl. 5:01 am skrev kamijin_fanta ***@***.***>: It seems to be a problem with other file systems as well. However, @react-icons/all-files...

Oh okay. I’ll try the normal react-icons eventually and see how it goes. This said, the original issue doesn’t affect me anymore because I switched from yarn to pnpm.

What’s worse, a line comment inside an OPTIONS pragma also negates the closing #-}, so the rest of the file becomes mis-highlighted. NB: if you special-logic the OPTIONS pragma, please...