Jason Conroy <jcon...@tscripta.net> writes: > Hi Hilton, thanks for the response. I am aware that the existing > approaches have this feature gap in `guix shell`, but the question is > whether, in your opinion, the recent changes make it any easier or > harder to close the gap in the future.
It would be harder. > Going back a while, Efraim shared a preliminary patch [1] that > generated a local cargo registry via a profile hook. It seemed like a > viable proof of concept: walking recursively through cargo-inputs and > cargo-development-inputs provided enough to reconstruct the registry > metadata and install the sources. Do you think that this strategy is > still possible with the new data model? I'm experimenting on the Guix Rust Registry, currently at <https://codeberg.org/hako/guix-rust-registry/>. It's basically managing rust-crates and rust-sources modules in a separate repository. It should be possible to write a script to turn definitions in the modules into an on-disk registry. And we can have some policy, for example to allow contribution of libraries without it being a dependency and check in pruned versions of modules into Guix.
