Merge pull request #127 from outis1one/claude/laughing-sagan-2oo2mx

Claude/laughing sagan 2oo2mx

iso/firstboot/kiosk-firstboot

@@ -22,6 +22,10 @@
 #   KIOSK_OFFLINE_WAIT=N on an ISO built with --offline: seconds to wait
 #                        for internet before installing from the bundled
 #                        packages instead (default 90).
+#   KIOSK_STALL_TIMEOUT=N   on an --offline ISO, unattended: if the install
+#                        prints nothing for N seconds (a stalled download on
+#                        a flaky link), stop it and finish from the bundled
+#                        packages instead (default 600).
 ################################################################################
 
 set -uo pipefail
@@ -40,6 +44,7 @@ OFFLINE_FLAG="$STATE_DIR/offline-mode"
 
 KIOSK_UNATTENDED=0
 KIOSK_OFFLINE_WAIT=90
+KIOSK_STALL_TIMEOUT=600
 # shellcheck source=/dev/null
 [[ -f "$CONF" ]] && source "$CONF"
 
@@ -61,6 +66,21 @@ if [[ "${1:-}" == "--cleanup" ]]; then
     exit 0
 fi
 
+have_bundle() { [[ -f "$OFFLINE_BUNDLE/apt/Packages" ]]; }
+
+# Finish this install from the bundled packages instead of the internet
+# (cleanup() turns it back off). install.sh notices via lib/config.sh's
+# offline_mode_active.
+switch_to_offline() {
+    touch "$OFFLINE_FLAG"
+    if ! "$OFFLINE_APT" on; then
+        echo "ERROR: couldn't switch apt to the offline bundle."
+        rm -f "$OFFLINE_FLAG"
+        return 1
+    fi
+    return 0
+}
+
 # Make sure tty1 gets its login prompt back. getty@tty1 is ordered after
 # this unit and normally starts by itself once we exit; this just covers
 # a getty that was stopped or never queued (e.g. started by hand later).
@@ -83,13 +103,15 @@ banner() {
 # drop to a shell as the admin user to fix it, or give up until next boot.
 # Returns 0 to retry, 1 to give up.
 offer_retry() {
-    local choice
+    local choice offline_opt=""
+    have_bundle && [[ ! -f "$OFFLINE_FLAG" ]] && offline_opt="   [O] Install from the offline bundle on this disk"
     while true; do
         echo
-        echo "  [R] Retry   [S] Shell (as $ADMIN_USER, exit to come back)   [Q] Quit to login prompt"
+        echo "  [R] Retry   [S] Shell (as $ADMIN_USER, exit to come back)   [Q] Quit to login prompt$offline_opt"
         read -r -p "Choice [R]: " choice
         case "${choice,,}" in
             ""|r) return 0 ;;
+            o) [[ -n "$offline_opt" ]] && switch_to_offline && return 0 ;;
             s) runuser -l "$ADMIN_USER" ;;
             q)
                 echo "Setup will run again on next boot, or run it now with:"
@@ -227,6 +249,39 @@ run_install() {
 
     # runuser rather than sudo -u: sudo would set SUDO_USER=root, which
     # config.sh reads as "the admin running this tool".
+    # A stalled download on a flaky connection would otherwise hang an
+    # unattended install forever. With the offline bundle as a safety net,
+    # watch the log: no new output for KIOSK_STALL_TIMEOUT seconds means
+    # stuck, so the whole process group is stopped (the caller then
+    # finishes from the bundle).
+    if [[ "$KIOSK_UNATTENDED" == "1" ]] && have_bundle && [[ ! -f "$OFFLINE_FLAG" ]]; then
+        touch "$LOG"
+        setsid bash -c '
+            runuser -u "$1" -- env -i "${@:4}" bash -c "cd \"\$1\" && exec ./install.sh" _ "$2" 2>&1 | tee -a "$3"
+            exit "${PIPESTATUS[0]}"
+        ' _ "$ADMIN_USER" "$REPO_DEST" "$LOG" "${envs[@]}" &
+        local pid=$!
+        (
+            while kill -0 "$pid" 2>/dev/null; do
+                sleep 20
+                age=$(( $(date +%s) - $(stat -c %Y "$LOG") ))
+                if (( age > KIOSK_STALL_TIMEOUT )); then
+                    echo
+                    echo "No output for ${age}s - the online install looks stuck; stopping it."
+                    kill -TERM -- "-$pid" 2>/dev/null
+                    sleep 30
+                    kill -KILL -- "-$pid" 2>/dev/null
+                    break
+                fi
+            done
+        ) &
+        local watcher=$!
+        wait "$pid"
+        local rc=$?
+        kill "$watcher" 2>/dev/null
+        wait "$watcher" 2>/dev/null
+        return "$rc"
+    fi
     runuser -u "$ADMIN_USER" -- env -i "${envs[@]}" \
         bash -c 'cd "$1" && exec ./install.sh' _ "$REPO_DEST" \
         2>&1 | tee -a "$LOG"
@@ -249,6 +304,15 @@ while true; do
     echo
     echo "install.sh did not finish (exit code $rc). Full output: $LOG"
     if [[ "$KIOSK_UNATTENDED" == "1" ]]; then
+        # Online attempt failed or stalled: the bundle on this disk is
+        # always going to work, so use it rather than retry the same thing.
+        if have_bundle && [[ ! -f "$OFFLINE_FLAG" ]]; then
+            echo "Online install didn't finish - continuing from the offline bundle on this disk."
+            if switch_to_offline; then
+                attempt=$((attempt + 1))
+                continue
+            fi
+        fi
         if (( attempt >= 3 )); then
             echo "Giving up after $attempt attempts; will try again next boot."
             release_tty

lib/provision.sh

@@ -127,7 +127,7 @@ provision_install_nodejs() {
             # The bundle carries NodeSource's own nodejs package.
             sudo apt install -y nodejs
         else
-            curl -fsSL https://deb.nodesource.com/setup_22.x | sudo -E bash -
+            curl -fsSL --connect-timeout 20 --max-time 300 https://deb.nodesource.com/setup_22.x | sudo -E bash -
             sudo apt install -y nodejs
         fi
     fi