darling-boot

name	darling-boot
description	Cleanly tear down and boot a Darling prefix without wedging launchd. Use before any live boot/smoke test in the Darling workspace (guest launch, shellspawn, clang-in-guest, DCC2 smoke). Prevents the perf#23a wedge where rapid boot/kill churn hangs launchd in __skb_wait_for_more_packets and a leftover orphan mldr/launchd silently blocks the next boot.

darling-boot: clean teardown + single boot

Darling boots via a setuid launcher; a botched teardown leaves an orphan launchd/mldr process tree that silently blocks the next boot (shellspawn.sock never appears). Rapid boot/kill churn additionally wedges launchd in __skb_wait_for_more_packets (perf#23a). This skill enforces the only reliable sequence: full teardown → kill orphans by PID → settle → ONE background boot → poll.

When to use

Before ANY live boot or in-guest smoke test: darling shell, shellspawn, /usr/bin/true in guest, clang-in-guest, DCC2 live smoke, closure boot tests.

Preconditions (verify first)

• west darling-doctor is green (manifest/build/deploy aligned). If it fails, STOP and fix drift first — do not boot a mismatched prefix (that was #89/#90).
• Know your prefix: DPREFIX (default ~/.darling); the launcher lives at ~/work/darling-prefix/bin/darling.

Protocol (do these in order — do NOT skip the orphan kill)

1. Graceful shutdown (ignore failure — the point is best-effort):

   darling shutdown 2>/dev/null || true
   

2. Kill the process tree by PID — server, vchroot, AND any orphan mldr/launchd. This is the step people skip; the orphan is the real confound.