[WIP] New package: shellcaster-1.0.1

[daniel-eys]

Terminal-based podcast manager built in Rust

https://github.com/jeff-hughes/shellcaster

Unfortunately nocross because ncurses-rs's build.rs tries to run a target executable on the host.

[ericonr]

See #23972 , ncurses-rs is a bad dependency.

[daniel-eys]

Exactly, not cross-friendly at all. That's why I eventually went for nocross.

[ericonr]

It's also arch restricted, due to their incorrect usage of signed/unsigned chars.

[daniel-eys]

So basically archs="i686* x86_64*"?

[fosslinux]

A better construct here is:

case "${XBPS_TARGET_ARCH}" in
    somearch) broken="blah" ;;
    *) ;;
esac

because archs is meant to be used for packages that explicitly only support particular archs, this is unintended breakage.

[ericonr]

I'm not sure I agree, because ncurses-rs is supposedly kinda dead iirc, so it won't ever not be broken.

There is, however, a chance that upstream changes the dependency they are using, so I guess broken could work just as well.

[fosslinux]

I see it as an unintended consequence by the maintainer of shellcaster.

If we were packaging ncurses-rs archs would be the right thing to do but I don't think it is here.

[daniel-eys]

Well, given the fact that the dependency ncurses-rs is quite troublesome and their upstream seems kinda inactive, I'm probably dropping this pr.

[fosslinux]

Nah, keep it open, I'm working on something for ncurses-rs. I like this application, TBH

[kotajacob]

Hey I also use this and would be happy to see it packaged. I opened a ticket on shellcaster to make the author aware of the issue. Also it seems like shellcaster uses pancurses which uses ncurses-rs. Both projects haven't been updated since 2019 so if ncurses-rs gets forked pancurses will probably also need to be forked.

I looked into #23972 and it seems like upstream dijo switched out their pancurses dep for termion.

[jeff-hughes]

Hi there -- shellcaster maintainer here. I'd love to make it available for Void Linux and for ARM as well. Yes, in hindsight relying on pancurses and ncurses-rs was a bad idea; I didn't notice that the projects are not very active. Long-term I may try to move to termion instead to remove the ncurses dependency, but that would require a fairly large rewrite.

Forking the ncurses-rs library to merge the open PR for building on arm64 would be an option, and I'm happy to jump to this dependency as long as it doesn't impact building on x86. However, to publish shellcaster on crates.io (which allows for easy installation with cargo install shellcaster), all the deps need to be on crates.io as well. I don't feel comfortable becoming the maintainer of a forked ARM-compatible ncurses-rs/pancurses library on crates.io if I don't even have an ARM device myself to test things on. If there is someone willing to do that, that would be great, and I'd happily switch dependencies.

The other option is that I can add a git dependency in my Cargo.toml, which will mean it will build ncurses-rs from the crates.io version when installing from crates.io, and build it from the Github repo when compiled from source. It's a little bit of a hack, but if I end up being the one forking those projects, that's ideal for me, as it doesn't set me up as some sort of "official" ncurses fork -- the scope is restricted to shellcaster.

Anyway, I wanted to check in here to figure out what would work best. Thanks, @kotajacob, for letting me know this existed. @fosslinux, you mention you're doing something with ncurses-rs -- what does that entail? If it's something I can help with in some way, I'd be happy to.

[fosslinux]

I haven't been able to do very much work on that yet, but it would basically involve adding a ncurses-rs build_helper so it patches ncurses-rs.