scripts/iso: bootstrap-modes harness catches up with the v1.21.0 in-script pairing wait

The old pairing/delivery scenarios expected a non-zero exit on an
unbound 204 poll — the one-poll-per-invocation design v1.21.0 (R-33)
deliberately removed; against the current script they would hang on a
real sleep. Now: a PATH-faked sleep counts the waits and flips the poll
to 200 after 3 cycles, so one scenario proves the whole v1.21.0 shape
in a single invocation (register -> in-script 204 waits -> delivery ->
host-install -> done-flag, exit 0), plus a 410 crash-window scenario
(still exits non-zero on purpose). Runaway guard: fake sleep kills the
loop after 25 calls. Assistant image gains python3 (the bootstrap's
JSON parsing needs it; PVE ships it on the real box) — the harness runs
in that image.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UuFPHmHNrCJj1VhY6QdDMU

scripts/iso/Dockerfile.assistant

@@ -15,9 +15,11 @@ FROM debian:trixie
 # 2.12-9+pmx2 generation, so the tool matches the modules it embeds. imagemagick: the R-38 GRUB boot
 # card — grub/generate-grub-background.sh letterboxes the website's brand asset onto a 1024x768
 # gfxterm canvas at repack time, so the boot screen has ONE source (the website asset) and not a
-# second pre-rendered copy checked into the repo to drift.
+# second pre-rendered copy checked into the repo to drift. python3: felhom-bootstrap.sh's JSON
+# parsing depends on it on the installed box (PVE ships it) — the test harness
+# (test/bootstrap-modes.sh) runs the bootstrap in THIS image, so it must match that dependency.
 RUN apt-get update && apt-get install -y --no-install-recommends \
-        ca-certificates wget gnupg xorriso shellcheck \
+        ca-certificates wget gnupg xorriso shellcheck python3 \
         grub-common grub-efi-amd64-bin mtools dosfstools imagemagick \
     && rm -rf /var/lib/apt/lists/*
 

scripts/iso/test/bootstrap-modes.sh

@@ -1,12 +1,16 @@
 #!/bin/bash
 # bootstrap-modes.sh — R-21 slice C regression harness for felhom-bootstrap.sh's two modes. Runs as
-# root inside a throwaway debian container (writes /etc/felhom etc.); fakes curl + host-install +
+# root inside a throwaway container (felhom-iso-assistant:trixie — python3 needed by the pairing
-# systemctl on PATH. Asserts:
+# parsers; writes /etc/felhom etc.); fakes curl + host-install + systemctl + sleep on PATH. Asserts:
 #   D (regression): a DIRECT env (customer-id + passphrase) enters run_direct and makes ZERO calls to
 #                   /api/v1/appliance/* — the pairing code path is provably not entered.
-#   pairing:        an env WITHOUT customer-id/passphrase enters run_pairing, POSTs /appliance/register,
+#   P (pairing+delivery, v1.21.0 loop semantics): an env WITHOUT customer-id/passphrase enters
-#                   persists the token, and GETs /appliance/poll.
+#                   run_pairing, POSTs /appliance/register, persists the token, polls 204 INSIDE one
-#   delivery:       a poll that returns 200 writes the direct env (0600) and invokes host-install.
+#                   invocation (fake sleep counts the waits; the script never exits between polls),
+#                   then consumes the 200 delivery, writes the direct env and runs host-install.
+#   410:            a consumed-delivery poll exits non-zero (the crash-window hand-back to systemd).
+# (Updated for v1.21.0/R-33: the old one-poll-per-invocation scenarios expected a non-zero exit on
+# 204, which the in-script wait deliberately no longer does — waiting is not failing.)
 set -uo pipefail
 BSTRAP=/work/felhom-bootstrap.sh
 FAKE=/work/fakebin; rm -rf "$FAKE"; mkdir -p "$FAKE"
@@ -16,6 +20,17 @@ fail=0
 say() { echo "TEST: $*"; }
 check() { if eval "$2"; then echo "  ok: $1"; else echo "  FAIL: $1"; fail=1; fi; }
 
+# --- fake sleep: counts waits; flips the poll to 200 after 3, hard-aborts a runaway loop ------------
+cat > "$FAKE/sleep" <<'SLEEP'
+#!/bin/bash
+n=$(cat /work/sleep.count 2>/dev/null || echo 0); n=$((n+1)); echo "$n" > /work/sleep.count
+flip=$(cat /work/sleep.flip 2>/dev/null || echo "")
+[ -n "$flip" ] && [ "$n" -ge "$flip" ] && echo 200 > /work/poll-mode
+if [ "$n" -gt 25 ]; then echo "RUNAWAY: fake sleep hit $n calls — killing the loop" >&2; kill -TERM $PPID; fi
+exit 0
+SLEEP
+chmod +x "$FAKE/sleep"
+
 # --- fake curl: logs every invocation's URL; emulates -o (fetch), --data (register), -w code (poll) --
 cat > "$FAKE/curl" <<'CURL'
 #!/bin/bash
@@ -46,7 +61,7 @@ HI
         printf '%s' '{"customer_id":"drill","retrieval_passphrase":"SEKRET-PASS","mode":"appliance","extra_args":"--cores 2"}'
         [ -n "$wfmt" ] && printf '\n200'
       else
-        [ -n "$wfmt" ] && printf '\n204'
+        [ -n "$wfmt" ] && printf '\n%s' "$mode"   # 204 / 410 / whatever the scenario set
       fi
       exit 0 ;;
 esac
@@ -58,7 +73,8 @@ chmod +x "$FAKE/curl"
 printf '#!/bin/bash\nexit 0\n' > "$FAKE/systemctl"; chmod +x "$FAKE/systemctl"
 
 reset_state() {
-  rm -rf /etc/felhom /run/felhom-bootstrap-pass /var/lib/felhom-install "$CALLS" /work/hostinstall.log /work/poll-mode
+  rm -rf /etc/felhom /run/felhom-bootstrap-pass /var/lib/felhom-install "$CALLS" /work/hostinstall.log \
+         /work/poll-mode /work/sleep.count /work/sleep.flip
   mkdir -p /etc/felhom
 }
 
@@ -80,30 +96,35 @@ check "ZERO /appliance/poll calls"     "! grep -q '/appliance/poll' $CALLS"
 check "done-flag written"              "[ -f /etc/felhom/.bootstrap-done ]"
 check "env shredded on success"        "[ ! -f /etc/felhom/bootstrap.env ]"
 
-# ============================ pairing mode — registers, then polls ==================================
+# ============ P: pairing loop (v1.21.0) — register, wait unbound INSIDE one invocation, deliver =====
-say "pairing: no customer/passphrase -> register + poll (unbound=204)"
+say "P: pairing env -> register + in-script 204 wait -> 200 delivery -> host-install, ONE invocation"
 reset_state
 cat > /etc/felhom/bootstrap.env <<'ENV'
 FELHOM_HUB_URL=https://hub.example
 ENV
 echo 204 > /work/poll-mode
+echo 3 > /work/sleep.flip   # after 3 in-script waits the hub "binds" (poll flips to 200)
 bash "$BSTRAP"; rc=$?
-check "unbound poll -> exit non-zero (systemd retries)" "[ $rc -ne 0 ]"
+check "single invocation ran to done (exit 0)" "[ $rc -eq 0 ]"
 check "POSTed /appliance/register"      "grep -q '/appliance/register' $CALLS"
-check "appliance token persisted 0600"  "[ -f /etc/felhom/appliance-token ] && [ \"\$(stat -c %a /etc/felhom/appliance-token)\" = 600 ]"
+check "appliance token persisted 0600"  "[ -f /etc/felhom/.bootstrap-done ] || { [ -f /etc/felhom/appliance-token ] && [ \"\$(stat -c %a /etc/felhom/appliance-token)\" = 600 ]; }"
-check "GET /appliance/poll"             "grep -q '/appliance/poll' $CALLS"
+check "polled more than once (the wait lived in-script, not in systemd restarts)" "[ \$(grep -c '/appliance/poll' $CALLS) -ge 2 ]"
-check "no direct env written yet"       "! grep -q FELHOM_CUSTOMER_ID /etc/felhom/bootstrap.env"
+check "waited between polls (fake sleep called >=3x)" "[ \"\$(cat /work/sleep.count 2>/dev/null || echo 0)\" -ge 3 ]"
-check "host-install NOT run (unbound)"  "[ ! -f /work/hostinstall.log ]"
-
-# ============================ delivery — poll 200 writes env + runs host-install =====================
-say "delivery: bound poll (200) -> write direct env + run host-install"
-# keep the token from the previous step; flip the poll to 200
-echo 200 > /work/poll-mode
-bash "$BSTRAP"; rc=$?
-check "delivery run exited 0"                 "[ $rc -eq 0 ]"
-check "direct env written with customer-id"   "grep -q 'FELHOM_CUSTOMER_ID=drill' /etc/felhom/bootstrap.env || [ -f /etc/felhom/.bootstrap-done ]"
 check "host-install invoked after delivery"   "[ -f /work/hostinstall.log ]"
 check "done-flag written"                     "[ -f /etc/felhom/.bootstrap-done ]"
+# the token was ALSO consumed on success (run_direct scrubs both secrets)
+check "env shredded on success"               "[ ! -f /etc/felhom/bootstrap.env ]"
+
+# ============ 410: consumed delivery without a local env -> exit non-zero (crash window) ============
+say "410: consumed delivery + no env -> exit non-zero so systemd restarts clean"
+reset_state
+cat > /etc/felhom/bootstrap.env <<'ENV'
+FELHOM_HUB_URL=https://hub.example
+ENV
+echo 410 > /work/poll-mode
+bash "$BSTRAP"; rc=$?
+check "410 poll exits non-zero" "[ $rc -ne 0 ]"
+check "host-install NOT run"    "[ ! -f /work/hostinstall.log ]"
 
 echo "=================================================="
 if [ $fail -eq 0 ]; then echo "ALL BOOTSTRAP-MODE TESTS PASSED"; else echo "SOME TESTS FAILED"; fi