fix(guanghu-os): decouple release origin from tunnel retries

product-source/hololake-platform/guanghu-os/scripts/hololake-release-bridge.sh

@@ -15,13 +15,13 @@ readonly FRONT_DOOR_HOST=43.139.217.141
 readonly FRONT_DOOR_USER=hololake-tunnel
 readonly FRONT_DOOR_PORT=19440
 
-declare -a CHILDREN=()
+SERVER_PID=
+TUNNEL_PID=
 
 cleanup() {
   trap - EXIT INT TERM
-  for pid in "${CHILDREN[@]:-}"; do
+  [[ -z ${TUNNEL_PID:-} ]] || kill "$TUNNEL_PID" 2>/dev/null || true
-    kill "$pid" 2>/dev/null || true
+  [[ -z ${SERVER_PID:-} ]] || kill "$SERVER_PID" 2>/dev/null || true
-  done
   wait 2>/dev/null || true
 }
 trap cleanup EXIT INT TERM
@@ -40,32 +40,37 @@ require_exact_runtime() {
   [[ $(stat -c '%a' "$TUNNEL_KEY") == 600 ]]
 }
 
-start_pair() {
+start_server() {
-  CHILDREN=()
   /usr/bin/setpriv --reuid=989 --regid=989 --init-groups --inh-caps=-all --reset-env \
     /usr/bin/env HOME=/nonexistent USER=hololake-release LOGNAME=hololake-release \
     PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin \
     HOLOLAKE_RELEASE_HOST=127.0.0.1 HOLOLAKE_RELEASE_PORT="$RELEASE_PORT" \
     HOLOLAKE_RELEASE_STATE_ROOT="$RELEASE_STATE_ROOT" \
     /usr/bin/node "$RELEASE_SERVER" &
-  CHILDREN+=("$!")
+  SERVER_PID=$!
+}
 
+start_tunnel() {
   /usr/bin/ssh -N \
     -i "$TUNNEL_KEY" -o IdentitiesOnly=yes -o BatchMode=yes \
     -o UserKnownHostsFile="$TUNNEL_KNOWN_HOSTS" -o StrictHostKeyChecking=yes \
     -o ExitOnForwardFailure=yes -o ServerAliveInterval=20 -o ServerAliveCountMax=3 \
     -R "127.0.0.1:${FRONT_DOOR_PORT}:127.0.0.1:${RELEASE_PORT}" \
     "${FRONT_DOOR_USER}@${FRONT_DOOR_HOST}" &
-  CHILDREN+=("$!")
+  TUNNEL_PID=$!
 }
 
 main() {
   require_exact_runtime
+  start_server
   while :; do
-    start_pair
+    kill -0 "$SERVER_PID" 2>/dev/null || return 1
-    wait -n "${CHILDREN[@]}" || true
+    start_tunnel
-    for pid in "${CHILDREN[@]}"; do kill "$pid" 2>/dev/null || true; done
+    wait -n "$SERVER_PID" "$TUNNEL_PID" || true
-    wait 2>/dev/null || true
+    kill -0 "$SERVER_PID" 2>/dev/null || return 1
+    kill "$TUNNEL_PID" 2>/dev/null || true
+    wait "$TUNNEL_PID" 2>/dev/null || true
+    TUNNEL_PID=
     sleep 2
   done
 }

product-source/hololake-platform/guanghu-os/scripts/test-hololake-release-bridge.sh

@@ -0,0 +1,16 @@
+#!/usr/bin/env bash
+set -Eeuo pipefail
+
+source_root=$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)
+subject=${source_root}/scripts/hololake-release-bridge.sh
+
+bash -n "$subject"
+grep -Fq 'start_server' "$subject"
+grep -Fq 'start_tunnel' "$subject"
+grep -Fq 'wait -n "$SERVER_PID" "$TUNNEL_PID"' "$subject"
+grep -Fq 'kill -0 "$SERVER_PID"' "$subject"
+if grep -Fq 'for pid in "${CHILDREN[@]}' "$subject"; then
+  echo 'tunnel failure must not kill the healthy local release server' >&2
+  exit 1
+fi
+echo HOLOLAKE_RELEASE_BRIDGE_LIFECYCLE_OK

product-source/hololake-platform/guanghu-os/world-seed/scripts/run-guanghu-native-quality-gate.sh

@@ -96,6 +96,7 @@ run_gate pncc_runtime bash -c '
   node --test "$1/pncc-runtime/pncc-runtime.test.mjs"
   "$1/scripts/test-install-jd-pncc-runtime.sh"
   "$1/scripts/test-guanghu-master-init.sh"
+  "$1/scripts/test-hololake-release-bridge.sh"
 ' _ "${source_root}"
 run_gate linux_subcontrol_docker_backend \
   "${source_root}/scripts/test-linux-subcontrol-docker-backend.sh"