@Nadrieril, thanks for your PR! By analyzing the history of the files in this pull request, we identified @garbas, @nckx and @manveru to be potential reviewers.

When using the neovim configuration options, the neovim derivation gets replaced by a wrapper that links to the actual neovim derivation output. In doing so, it hides the normal neovim derivation.

In my opinion a wrapper should always be returned for the sake of consistency.

The problem is that if one wants to override an attribute of the underlying neovim (say, to patch it), using overrideAttrs will instead modify the wrapper, which is not the desired behavior.

overrideAttrs should override the call to stdenv.mkDerivation of the derivation that is returned. If it returns the wrapped one, then overrideAttrs should override that.

This PR makes it so that calling pkgs.neovim.overrideAttrs always overrides the expected derivation, even when neovim gets configured before of after.

Maybe its better to have an overrideNeovim method then instead for overriding the underlying derivation.

Anyway, I don't use neovim and don't know how its build.

I agree that it would be better to always have a wrapper, but as it is now changing it may break configurations. Actually the best way would be for the wrapper to use pkgs.neovim instead of the locally declared neovim, but this seems tricky to me.

Probably a better way would be to have two separate packages, like what firefox does. This would be very breaking though.

I guess #32657 might interest you. It would be nice if you could give it a spin.

Fixed by #32657.