Neil
efbf651c7b
* Let scheduled CI warmers wait and measure WebRTC startup * Measure a smaller daemon shutdown fixture image * Counterbalance WebRTC startup and verify retained fixture files * Record CI fixture measurements and remove temporary pilots * Clarify fixture build dependency cleanup evidence * Make coalesced snapshot fixture delivery deterministic * test: type the PTY write delay observer
Daemon shutdown descendant regression
Run from the repository root with Docker running:
ORCA_BACKGROUND_LAUNCH=1 node config/scripts/run-daemon-shutdown-descendants-docker.mjs
The runner bundles the production TerminalHost and native PTY wrapper from the working tree. By default it runs the candidate only. Pass --baseline <git-ref> to add a red/green run; the baseline loads the five teardown modules from that ref without changing the checkout. Both run in Linux with node-pty 1.1.0.
A live PTY shell owns a child that called setsid and ignores SIGTERM. The fixture verifies those relationships, calls host.dispose(), and immediately exits its process. A separate supervisor checks that the child has stopped, an unrelated canary is still running, and shutdown finished within the daemon's five-second limit. The baseline must leave the child running; the candidate must stop it.
ORCA_DOCKER overrides the Docker executable. ORCA_DOCKER_PLATFORM overrides the Linux architecture. The test creates a temporary image and removes it afterward.
This proves cleanup of descendants still parented beneath a live shell at shutdown. It does not prove cleanup of processes that had already reparented before shutdown, or reproduce systemd-oomd policy decisions.