I have an ErgoDox EZ which I haven’t flashed for years, but am now thinking of fiddling with the layout again. Looking at the notes I made describing the various hoops that need to be jumped through in order to compile a new layout from C sources and then flash the keyboard, gives me a strong urge to wrap it all up in a flake.
Does anyone have any wisdom to offer in this area?
It seems that the recommended procedure has changed significantly since I last looked. Very briefly, it goes something like this
qmk setup (Clones qmk/qmk_firmware repo with submodules, and verifies that everything looks OK.)qmk compile (ignoring injection of my own keymaps, for now)qmk flashCrucially, the ‘check that everything looks OK’ part in step 1 requires udev rules to be set. On NixOS this can be done by including pkgs.qmk-udev-rules in services.udev.packages, but I have no idea how to approach this in the sandboxed nix build environment.
Adding
qmk-firmware = {
flake = false;
url = "https://github.com/qmk/qmk_firmware.git";
ref = "refs/tags/0.24.6";
type = "git";
submodules = true;
};
as a flake input and using it as the src of a derivation, seems to get the git repo and all its submodules (with version pinning in flake.lock) but the verification step of qmk setup fails when pointed at this clone. Furthermore, I don’t see how I would inject my own keymap into this clone which would be in the immutable Nix store.
I have an xbows nature and compile qmk for it, but despite using nix for as much as possible, I simply took the container route as I had to wait forever for the compilers to be compiled. The qmk tool itself is fine - it’s everything around compiling for the MCU on your keyboard that’s painful.
Take a look at qmk_firmware/flake.nix at b9d97d2b49995a5647baade7cb343d7ee611d3e3 · RyanGibb/qmk_firmware · GitHub
Interesting.
How does it work without any flake inputs?