So, it seems random if Hugs need the patch or not. Which suggests some undefined behaviour.

But if it works for you without the patch, so much the better.

I have no idea what it would take to compile GHC with MicroHs. I've not tried.

I think it might not be too far away. MicroHs will probably need to get unboxed types (or at least fake them). GHC is written fairly conservatively and is not using all the latest features.

If somebody wants to try, I'm willing to add needed features to MicroHs if they are not too crazy.

On Sat, Jan 25, 2025 at 3:28 PM Lars-Dominik Braun <lars@6xq.net> wrote:

Hi Lennart,

> If someone fixes all the warnings when compiling Hugs, then I bet this
> problem goes away. But I am not going to do that. In its current state Hugs
> can compile a fully functional MicroHs, and that's enough for me.

actually, I can build the MicroHs package from
https://issues.guix.gnu.org/75778 just fine with our default, unpatched
Hugs. It can also compile/run the example from
https://github.com/augustss/MicroHs?tab=readme-ov-file#example.

Can we therefore just drop this patch? It causes Hugs to die with a
SIGFPE while evaluating the example expression `fromInteger (9 ^ 100)
:: Int` from above, which is rather ugly behavior.

And on an unrelated note: Would it be possible to bootstrap (a possibly
very, very old version of) GHC with MicroHs? (As far as I see our GHC
7 is currently bootstrapped via binaries.)

Lars