stylix/testbeds: don't unnecessarily use an IFD

[0xda157]

inspired by #1985 (comment), cc @MattSturgeon @danth @trueNAHO

---

• I certify that I have the right to submit this contribution under the MIT license
• Commit messages adhere to Stylix commit conventions
• Theming changes adhere to the Stylix style guide
• Changes have been tested locally
• Changes have been tested in testbeds
• Each commit in this PR is suitable for backport to the current stable branch

[MattSturgeon]

Initial review from my phone, so I can't do any testing or verification. But here are my thoughts and questions:

[MattSturgeon]

Although this is technically importing from an input derivation, IIUC it doesn't really count as IFD because flake inputs are special and get made available in the nix store before evaluation starts.

So unless flake-compat and flake.parts partitions are breaking that assumption, I don't think this is a problematic import. Maybe I was wrong to call out IFD in your testbeds?

IFD is normally an issue because the derivation has to get built part-way through the eval.

Looking elsewhere, it looks like base16Scheme gets used in:

    lib.stylix.colors = (cfg.base16.mkSchemeAttrs cfg.base16Scheme).override cfg.override;

And that is a function that also comes from a flake input. So if you completely wanted to avoid IFD bootstrapping I think you'd need to see lib.stylix.colors directly.

Does cfg.base16 itself do any IFD? Does it behave differently when given a yaml file path vs an attrset? Did you notice any performance differences with this change?

[trueNAHO]

This essentially just inlines the theme file, making future theme changes like in #1857 harder. IMHO this is not a worthy trade-off.

Ideally, we disable IFD in CI, as previously done in #1124.

[0xda157]

Although this is technically importing from an input derivation, IIUC it doesn't really count as IFD because flake inputs are special and get made available in the nix store before evaluation starts.

indeed, nix build github:nix-community/stylix#testbed:gnome:dark --option allow-import-from-derivation false builds fine, so this pr can be closed.

Does cfg.base16 itself do any IFD? Does it behave differently when given a yaml file path vs an attrset? Did you notice any performance differences with this change?

with my config replacing pkgs.base16-schemes with an attrset of colors allows it to build without an IFD, but seems like that's not the case if a flake input is used.

Ideally, we disable IFD in CI, as previously done in #1124.

I'm not really sure what this means... we can't just disable IFDs because the palette generator uses an IFD.