rakuPackage fails with error about sub MAIN

[lilyball]

I don't know if something changed in Raku recently (I'm brand new to it), but rakuPackage is failing to compile with the message

MAIN must be a 'sub' to allow it to be called as a CLI handler

It's referring to this line:

https://github.com/chloekek/raku-nix/blob/28820950819949e933c17cb274d1c0ffaa9bc5be/tools/install.p6#L3

Experimentally, a subroutine with a return type constraint appears to be of type Sub+{Callable[Nil]} rather than just being Sub, and it seems this isn't allowed. Removing the --> Nil should fix this.

[zopsicle]

Hi Lily,

Must have changed then. It always worked for me. Seems like a bug, as sub MAIN(--> Nil) being disabled is less than awesome.

---

General service announcement:

I saw you forked this repository. Thank you for continuing work on raku-nix! I lost interest in Raku some time ago so I stopped maintaining the Nix infrastructure.

I also took CRAI offline because I didn’t think anyone was using it. If you (or anyone else) want to host it yourself then that shouldn’t be too difficult if Nix-savvy; all the deployment code is there, as well as the server configuration. This includes Terraform configuration (may need to be adapted if you don’t want to host on Hetzner Cloud, and change the public key) and systemd service file installation. After updating the Terraform configuration and deleting terraform.tfstate, nix-build -A deploy && result/bin/deploy will take care of everything, including Let’s Encrypt.

[zopsicle]

If you want I can give you direct access to this repository.

[lilyball]

Must have changed then. It always worked for me. Seems like a bug, as sub MAIN(--> Nil) being disabled is less than awesome.

I'm told it's a bug in rakudo v2020.10 and it should be fixed in v2020.11. It looks like nixpkgs has already been updated for v2020.11 at this point.

---

I lost interest in Raku some time ago so I stopped maintaining the Nix infrastructure.

To be honest, while I think Raku is very neat, I've been running into frustrations that make me feel very iffy about it myself, such as a massive performance cliff when trying to solve Advent Of Code 2020 day 11, no clear way to get good errors back from grammars, concerns about how the module loading system hasn't been designed to ensure that distributions favor themselves when loading modules¹, and lots of confusion over containerized lists and implicit containerization and whatnot (e.g. (1, 2; 3, 4).flat eqv (@ = (1, 2; 3, 4)).flat # => False).

I'd like to see better Raku support in Nix, but that's more for the sake of Nix than for the sake of Raku. And my free time is pretty limited these days, so I don't know if I'm going to continue to invest any effort in this for the time being.

_{¹if I have two distributions for the same package available, and I load a module while specifying an explicit version or auth string, when it loads subsequent modules Raku does not give priority to its own distribution. This means I could load e.g. Foo:ver<1.0> and it could in turn load Foo::Bar and end up loading a module from the distribution for Foo:ver<1.5> instead. Given how much the module system has been designed to allow for inclusion of the same package in separate distributions, I'm pretty shocked at this oversight.}

[lilyball]

Feel free to close this ticket. I'm just leaving it open at the moment because of the other discussion in here.