Ryan Scott

Implementor: Solution implemented, review requested.

I tried out that branch, but I still witness the buggy behavior on the program above. (Looking at the [commit on that branch](https://github.com/ivanperez-keera/copilot/commit/048b8897cdf2afbcd4f84b87abe2e572d568372f), all it does is remove some constraints—are...

Thanks. I was able to build this branch after adding the following to my `cabal.project.local`: ```cabal source-repository-package type: git location: https://github.com/fdedden/language-c99-simple tag: 1482a2a4cc104ce6d3efb4f84e5685d6ecc9979b ``` ----- Something strikes me as suspicious...

I also tried out some variations of the program above using nested arrays and structs, and they also produce invocations of `memcpy` that appear valid. I did uncover one compilation...

> I think that a solution to this would be to pass the direction of `s0_tmp` as an argument to `s0_gen` for it to store the result in. Yes, that...

> Weird though how the optimisation level of GCC makes a difference. I always thought that up to `-O2`, semantics didn't change. I'm fairly certain that the program in https://github.com/Copilot-Language/copilot/issues/314#issuecomment-1225858862...

I really don't feel qualified to make a judgment about whether these should be _the_ laws for these classes, which is why I haven't taken action. It might be worthwhile...

Is there an upstream GHC issue about this?

Solution (2) would have to be implemented carefully. Here is a tricky corner case: ```hs {-# LANGUAGE ExistentialQuantification #-} {-# LANGUAGE PolyKinds #-} {-# LANGUAGE StandaloneKindSignatures #-} module Foo where...

See goldfirere/singletons#466 for an example of where this limitation occurs "in practice". In particular, https://github.com/goldfirere/singletons/issues/466#issuecomment-646117067 contains a laundry list of obstacles that I ran into when trying to implement option...