dsh-simple-restart

dsh-simple-restart addresses a very specific pain point: after you modify a client-side plugin bundle for dsh web, the change only takes effect once the host process restarts. Doing that from a terminal means hunting for the right window or wrestling with a systemd unit. This plugin adds a single row to Settings → General: a button that, after a confirming second click within five seconds, relaunches dsh web with the exact same boot arguments—entirely from the browser.

The implementation is worth a closer look. A process cannot replace itself; it can only stop, and the successor must start only after the listening port is genuinely free, or it dies with EADDRINUSE. So the restart is a two-process handshake: the host spawns a detached helper via node -e, hands it the full boot invocation (executable, argv array, working directory), and exits. The helper confirms the port is free by connecting to it—never by binding, which would itself hold the port at the exact moment the replacement needs it—then starts the replacement and watches for up to twenty seconds to confirm a successful bind. If anything goes wrong after the handoff, the helper writes diagnostics to .err.log and .out.log under $TMPDIR.

Security is enforced with several layers of fencing. The single HTTP route accepts only same-origin POST from loopback peers; any forwarding header, cross-origin Origin, or non-127.0.0.1 socket results in a 403. When a systemd supervisor is detected, restart is refused by default because KillMode=control-group would kill the helper along with the unit. An attached debugger is likewise detected and blocked. At most one restart may be in flight at a time; concurrent attempts receive a 409.

The plugin fits well for engineers actively developing dsh client-side plugins—edit, save, click, done, no terminal hunt. It also serves anyone who runs dsh web as a long-lived background process or under a service manager and finds manual restarts tedious or error-prone.