mirror of
https://github.com/multica-ai/multica.git
synced 2026-08-13 11:30:58 +02:00
e583c149ef
Three defects in the documented local flow, plus the bootstrap collapse that keeps them from coming back. 1. `make daemon` produced a daemon that could not run any task. The chain ended in `go run`, and the daemon records its own executable path at startup to re-exec as the execution-environment helper for every task. `go run` deletes that temp binary when the launcher exits, so the daemon started, registered, heartbeat — then failed every task with "fork/exec .../go-build.../exe/multica: no such file or directory". `make daemon` now builds `server/bin/multica` and runs that (`ARGS` forwards any daemon subcommand), and `daemon start` refuses an ephemeral binary up front instead of deferring the failure. 2. `make stop` killed the daemon. `lsof -ti:PORT` matches every process holding a socket on the port, including CLIENTS — and the daemon holds a long-lived connection to the backend. Stop now goes through scripts/port-guard.sh, which filters to `-sTCP:LISTEN` and sends TERM before KILL. scripts/port-guard.test.sh pins both halves against a real listener and a real connected client. 3. Restarting over a running instance failed invisibly: the new process died on the bind, the old one kept serving, and /health answered 200 throughout. `make dev` now preflights both ports and names the process that owns them, and /health reports `pid`, `started_at`, and `commit` so a caller can prove which process answered. 4. `make dev-bootstrap` collapses the ~12-step manual sequence into one command: env file, fixed verification code before first launch, DB via DATABASE_URL (not docker exec), port preflight, detached services, health verified against process identity, single-shot login, PAT, workspace, CLI profile, built daemon binary. It pins a durable TMPDIR so an agent-created environment outlives the run that created it. `make dev-bootstrap-stop` tears it down. Measured on a clean checkout: 11.7s to a logged-in, daemon-attached environment. Co-authored-by: multica-agent <github@multica.ai>