Cleanup nix build

[bendi-github]

Motivation: Cleanup nix build and add fields required for inclusion in nixpkgs

Things done:

• rename vicinae.nix to package.nix
• Remove unnecessary dependencies
• Remove unnecessary build steps
• Add meta field @Vortriz
• add version field
• use new finalAttrs-style mkDerivation

This builds without issues. I have tested the resulting binary and it seems to work as well as one built from main. We also get proper build logs again, the hand-rolled build phase caused them to not be output to the log for some reason.

Why was I able to remove so many deps and build steps, am I missing something here?

When opening the launcher window WARN - Could not find any platform plugin is printed to the console, although this also happens on main. This does not seem to impede functionality however.

[aurelleb]

Not sure about
WARN - Could not find any platform plugin

Is the clipboard history working on your build?

[Vortriz]

it would be nice to not use flake-utils and instead use an eachSystem function

edit: thanks for that meta attribute!

[bendi-github]

Not sure about WARN - Could not find any platform plugin

Is the clipboard history working on your build?

Yes, clipboard history works fine

[TomRomeo]

Great work, thank you for cleaning up the nix build!

Why was I able to remove so many deps and build steps, am I missing something here?

I think many of these might have been leftovers from me trying things out that did not work / were unnecessary. (E.g.: I think I did not import cmake as a build input in the beginning because of which the native build steps did not work).
I also did not know about npmHooks, this is great!

One thing I am weary about though is the removal of the wayland binaries in the postFixup step. For me, the binary produced by the flake did originally not work without them. Maybe something changed since then? Did you build the flake on a system using a wayland compositor?

Btw, I think if we set pname and version, we can also remove the name attribute, right?

[bendi-github]

One thing I am weary about though is the removal of the wayland binaries in the postFixup step. For me, the binary produced by the flake did originally not work without them. Maybe something changed since then? Did you build the flake on a system using a wayland compositor?

I am running Hyprland (a wayland-only compositor) and haven't had any issues with removing the wayland binaries. Maybe this only becomes a problem on non-NixOS? I have also checked echo $PATH | grep wayland, that doesn't output anything.

The reference I am using (https://github.com/NixOS/nixpkgs/blob/master/pkgs/README.md) doesn't mention anything about removing name, am I missing something?

[whispersofthedawn]

The reference I am using (https://github.com/NixOS/nixpkgs/blob/master/pkgs/README.md) doesn't mention anything about removing name, am I missing something?

This is a feature of stdenv.mkDerivation. See https://nixos.org/manual/nixpkgs/unstable/#sec-using-stdenv

For convenience, you can also use pname and version attributes and mkDerivation will automatically set name to "${pname}-${version}" by default. Since RFC 0035, this is preferred for packages in Nixpkgs, as it allows us to reuse the version easily:

[bendi-github]

@dawnofmidnight Hey, thanks for the info! I have removed the name attribute.

[aurelleb]

the conflicts are from me merging your other pull request

[aurelleb]

the existing nix checks are failing, is this normal @bendi-github ?

[bendi-github]

Seems like the error occurs because I renamed vicinae.nix to package.nix. The workflow uses both the newer and the older nix CLI with the latter making some implicit assumptions about filenames. I don't understand what running nix-shell --run "echo OK" is supposed to accomplish. @TomRomeo what was your intention with this workflow step? Should we make it work with the new CLI? Shouldn't the artifact we want to build be specified clearly instead of relying on nix build automatically choosing .#packages.<currentSystem>.default as the default target? You seem to have copied the github action from nix.dev almost 1:1, with the only change being the switch from nix-build to nix build. Looking at previous runs of this workflow, it seems like the invocation of nix-shell --run "echo OK" builds protoc-gen-ts_proto, which I removed in this PR. Is this intended?

Edit 2: Originally, there was a paragraph of me being confused about github actions here, which was woefully uninformed, therefore I removed it.

[idlip]

Hi,

Since you are at it, can you please bump the nixpkgs revision (flake.lock)

I use unstable branch, and qutebrowser (from unstable) uses 6.9.2 QT library, but vicinae flake.lock gets 6.9.1 version
I think this issue occurs due to vicinae having overlay here?

I'm not very sure, please let me know if this can have some workaround.

14:39:50 CRITICAL: Cannot mix incompatible Qt library (6.9.2) with this library (6.9.1)

FYI: some old ref nixos discourse

[bendi-github]

Hi,

Since you are at it, can you please bump the nixpkgs revision (flake.lock)

I use unstable branch, and qutebrowser (from unstable) uses 6.9.2 QT library, but vicinae flake.lock gets 6.9.1 version I think this issue occurs due to vicinae having overlay here?

I'm not very sure, please let me know if this can have some workaround.

14:39:50 CRITICAL: Cannot mix incompatible Qt library (6.9.2) with this library (6.9.1)

FYI: some old ref nixos discourse

I use

  vicinae = {
      url = "github:vicinaehq/vicinae";
      inputs.nixpkgs.follows = "nixpkgs";
    };

[idlip]

I use

  vicinae = {
      url = "github:vicinaehq/vicinae";
      inputs.nixpkgs.follows = "nixpkgs";
    };

Yea, that should solve it. But I think I'll miss out cachix from vicinae, so I'd have to compile vicinae myself...
May I know how long it takes to compile this on avg?

Edit: Solved, I built it. It took ~ 8mins

[Leka1B]

I feel like this is still good to include into the flake, as vicinae is actively worked on and nixpkgs get updated much slower.

Is the aforementioned error possible to fix?

[aurelleb]

closing as #858 got merged