Troels Henriksen

I don't think *normal* users would know how to do that. I'm also skeptical about putting things in `~/.mlkit`. I would recommend creating an MLB folder next to the `.mlb`...

I volunteer to help maintain the package if it is moved.

I am not any particular haskeline expert or enthusiast, but I maintain various other community packages (like [fgl](https://github.com/haskell/fgl) or [haskell-src-meta](https://github.com/haskell-party/haskell-src-meta/issues)) that are dependencies of programs I care about.

Immutable bytes sound good to me. I agree that it is best to let Futhark handle the allocation, but note that Python is still responsible for deallocation.

This is because the ISPC kernels are in a separate `.o` file. futhark-pycffi will have to be modified to also include that in the `.so` that gets built. (I don't...

Futhark's ISPC backend uses its own scheduler, not `tasksys.cpp`.

Using `cmake -DCMAKE_BUILD_TYPE=Release -DENABLE_TESTS=OFF ..` skips the problematic package (GTest) and allows `cmake` to finish.

Of course. `smlpkg` just requires that the repository directory structure is of a specific form ([example](https://github.com/diku-dk/sml-md5)).

@apaszke Is the failure here related to my PR?

[Looks like an upstream issue.](https://github.com/haskell-gi/haskell-gi/issues/318)