Every fleet wrapper in this repo assumes the closed-source ix CLI is available at ~/.local/bin/ix (see lib/fleet.nix userLocalBinPath). There is no documented install path, no nix run fallback, and the only failure mode is "ix binary not found in PATH" with a generic "install it" message.
What needs deciding
- Where does
ix come from in a fresh contributor's workflow? Cargo install from the (private) ix repo? A private flake input? A pre-built binary download? Something the operator just installs out of band?
- Should
health-checks and ix-fleet wrappers fail with an actionable install command (the canonical install line for this workflow), or do they just check and exit with a link to wherever the install docs end up?
- Is there value in shipping a
nix run .#ix that builds from a flake input pointed at the ix repo, gated on a contributor having a checkout? Or does that just paper over a real missing-step in the contributor flow?
This is a coordination question more than a code one — the answer probably names a path in this repo's docs (CONTRIBUTING.md?), pinpoints the install command, and updates the wrapper error message to reference it. But "what is the install command" needs a human call.
Refs #71.
Every fleet wrapper in this repo assumes the closed-source
ixCLI is available at~/.local/bin/ix(seelib/fleet.nixuserLocalBinPath). There is no documented install path, nonix runfallback, and the only failure mode is "ix binary not found in PATH" with a generic "install it" message.What needs deciding
ixcome from in a fresh contributor's workflow? Cargo install from the (private) ix repo? A private flake input? A pre-built binary download? Something the operator just installs out of band?health-checksandix-fleetwrappers fail with an actionable install command (the canonical install line for this workflow), or do they just check and exit with a link to wherever the install docs end up?nix run .#ixthat builds from a flake input pointed at the ix repo, gated on a contributor having a checkout? Or does that just paper over a real missing-step in the contributor flow?This is a coordination question more than a code one — the answer probably names a path in this repo's docs (
CONTRIBUTING.md?), pinpoints the install command, and updates the wrapper error message to reference it. But "what is the install command" needs a human call.Refs #71.