Refactor the post-receive deployment hook from a linear script into
discrete functions (init_log_file, write_metadata_header,
copy_env_by_branch, run_build, stop_existing_next, launch_next) for
better readability, maintainability, and error isolation.
Key changes:
- Remove `set -e` so deploy errors don't block git push; each step
handles its own failures (missing env, build failure
- Select env template per branch: main → .env.production, test/dev → .env.local
- Skip env copy on unknown branches with warning log
- Revert pnpm run start output to /dev/null (debugging complete)
- Reorder LOG_FILE definition before env-copy block so case branches can log
- Add per-environment AUTH_SECRET (dev/test/prod) for next-auth v4 JWT signing
- Temporarily redirect next start stdout/stderr to LOG_FILE in post-receive hook for server error debugging
Move repository-managed git hooks from `git_hooks/` to `.githooks/` to follow the conventional hidden-directory naming, and add a graceful shutdown step in `post-receive` that terminates any running `next start` / `next-server` processes (SIGTERM → wait 2s → SIGKILL → wait 1s) before relaunching, avoiding EADDRINUSE port conflicts so the freshly built code actually takes effect. README updated to reference the new path.
admin