pixcode Files

Fixes the Linux in-place update flow crashing the server.

What was happening: clicking "Update Now" on a Linux npm-global install ran npm install -g against the running server's own package files. npm rewrote them while the process still had them open — the daemon crashed mid-install, the child install died with it, and users ended up with a half-installed package and no running server. The UI just showed "Failed to fetch".

What changes in 1.31.13:

• If your server is already on the latest version, the npm path now short-circuits entirely. No install attempt, no restart, no chance of self-crash. (The desktop runtime-dir path already had this guard — parity restored.)

• The install child process is now spawned detached so it survives the parent being killed mid-swap. npm finishes, systemd brings the server back on the new code, and the update UI's /health probe picks it up as a successful self-restart.

• The update modal now recovers from mid-stream network errors (the "Failed to fetch" case), not just clean stream closes. Any stream interruption during an update now triggers the same /health version check instead of surfacing a scary error.

If your server is already dead from a previous failed update: run "sudo systemctl restart pixcode" (or "pkill -f pixcode && pixcode"), then refresh the browser. The update files are already on disk from the crashed attempt.