Vladislav Zavialov

Alright. Only 16 and 32 bit are realistic scenarios. I threw in 8 and 64 just for good measure. > I honestly don't see the appeal in that; when all...

> I'm convinced that it doesn't matter much. Alright. Perhaps someone someday will measure the actual effect of going from 16-bit to 32-bit, but it won't be me 😄 Correctness...

I agree with @Zemyla that `rnf` for functions should force everything inside the function's closure. > I think I understand what you are suggesting, but I doubt that it is...

Although I'm not sure how @Zemyla imagines running `rnf` for the data captured in the function closure. Where would the `NFData` instances for the captured data come from? This suggests...

> and thus the RTS would have to know about how to NFData everything underneath the function itself. Thats a pretty hard ask Maybe it's not hard – I believe...

I'd like to offer the following specification: * `rnf :: a -> ()` should be a function such that for any `f` and `cont`, the following program prints `A` before...

> In particular, 0-ary parameterised productions are like the declaration-site %inline pragma. Are you sure about this? My mental model for parameterised productions was that they were expanded to top-level...

> Cabal doesn't realize it should use happy-boot for `.ly` files Could this be resolved by having a custom Setup.hs and calling happy-boot by hand?

@phadej OK. How do you imagine it should be done? Should there be a field in the `.cabal` file specifying that `.y` and `.ly` files are to be processed with...

I would like us to take a critical look at what we're trying to achieve. Why wouldn't a single library `happy-lib` suffice?