test(build): add make watchdog, the glance memory budget and the glance screenshot

- make watchdog (in make ci): a stress build encodes, packs, turns into
  runs and draws the largest code of every 2D symbology at two steps per
  event on the Forerunner 255S and 70; it fails on a watchdog trip, and
  the last code left on the screen must decode (bitmap path on the 70).
- make memtest checks the glance (DIAG glance, budget 70 % of 64 KB).
- screenshots.sh launches the app again after the flow and compares the
  glance with the favourite's code (glance_code.png on 7 watches); the
  About goldens show 0.8.0.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

3 files changed, 99 insertions, 13 deletions

diff --git a/docker/scripts/app-diag.sh b/docker/scripts/app-diag.sh
index 804d68a..025fd49 100755
--- a/docker/scripts/app-diag.sh
+++ b/docker/scripts/app-diag.sh
@@ -1,21 +1,74 @@
 #!/usr/bin/env bash
 # Author:  cirakg & claude ai
-# Summary: Run the debug app on each device in the headless simulator and check start-up time or memory budgets.
+# Summary: Run the debug app on each device in the headless simulator and check start-up time, memory, or the watchdog stress test.
 #
-# Usage:   app-diag.sh perf|mem DEVICE [DEVICE ...]
+# Usage:   app-diag.sh perf|mem|watchdog DEVICE [DEVICE ...]
 # Output:  build/diag/<DEVICE>.log (console with DIAG and HARNESS lines) and
 #          build/diag/<perf|mem>-summary.json; verdict from tools/check_budgets.py.
+#          watchdog: the stress build (source/util/StressTest.mc) on each
+#          device, build/diag/watchdog-<DEVICE>.log; it fails when the
+#          simulator reports an error (such as "Watchdog Tripped"), the
+#          test doesn't finish, or its last code (left on the screen)
+#          doesn't decode.
 set -euo pipefail
 # shellcheck source=ciq-env.sh
 source "$(dirname "${BASH_SOURCE[0]}")/ciq-env.sh"
 
-kind="${1:?usage: app-diag.sh perf|mem DEVICE [DEVICE ...]}"
+kind="${1:?usage: app-diag.sh perf|mem|watchdog DEVICE [DEVICE ...]}"
 shift
-[[ "${kind}" == perf || "${kind}" == mem ]] || die "unknown kind '${kind}' (use perf or mem)"
-(($# > 0)) || die "usage: app-diag.sh perf|mem DEVICE [DEVICE ...]"
+[[ "${kind}" == perf || "${kind}" == mem || "${kind}" == watchdog ]] || die "unknown kind '${kind}' (use perf, mem or watchdog)"
+(($# > 0)) || die "usage: app-diag.sh perf|mem|watchdog DEVICE [DEVICE ...]"
 ciq_require_sdk
 ciq_require_key
 
+# The stress test takes a few minutes per device (about 3,500 job steps at
+# 50 ms on the 512 KB watches).
+WATCHDOG_TIMEOUT_S=1200
+STRESS_ITEMS=10
+
+if [[ "${kind}" == watchdog ]]; then
+  for device in "$@"; do
+    build-app.sh stress "${device}"
+  done
+  ciq_start_display
+  ciq_start_simulator
+  mkdir -p "${BUILD_DIR}/diag"
+  failed=0
+  for device in "$@"; do
+    log="${BUILD_DIR}/diag/watchdog-${device}.log"
+    log "watchdog stress test on ${device}"
+    ciq_launch_app "${device}" "${log}" stress || log "no first frame on ${device}"
+    for _ in $(seq 1 "${WATCHDOG_TIMEOUT_S}"); do
+      grep -q "DIAG stress all done\|Error" "${log}" && break
+      kill -0 "${CIQ_MONKEYDO_PID}" 2>/dev/null || break
+      sleep 1
+    done
+    # The last code stays on the screen; it must decode (on most watches
+    # it was drawn from a bitmap).
+    decoded=1
+    if grep -aq "DIAG stress all done" "${log}"; then
+      sleep 1
+      python3 /work/tools/ui_flow.py --decode "OTPW stress" --device-dir "${CIQ_HOME}/Devices/${device}" \
+        --window "$(ciq_simulator_window)" --out "${BUILD_DIR}/diag/watchdog-${device}" >&2 || decoded=0
+    fi
+    ciq_stop_app
+    grep -a "DIAG stress" "${log}" | sed "s/^/[${device}] /" >&2 || true
+    done_count="$(grep -ac "DIAG stress done name=" "${log}" || true)"
+    if grep -aq "Error" "${log}"; then
+      grep -a "Error" "${log}" | sed "s/^/[${device}] /" >&2
+      failed=1
+    elif ! grep -aq "DIAG stress all done" "${log}" || ((done_count != STRESS_ITEMS)); then
+      log "${device}: the stress test did not finish (${done_count} of ${STRESS_ITEMS} items)"
+      failed=1
+    elif ((decoded == 0)); then
+      failed=1
+    fi
+  done
+  ((failed == 0)) || die "watchdog stress test failed; logs in build/diag/"
+  log "watchdog stress test passed on: $*"
+  exit 0
+fi
+
 for device in "$@"; do
   build-app.sh debug "${device}"
 done
diff --git a/docker/scripts/build-app.sh b/docker/scripts/build-app.sh
index 1a7d59a..5ead272 100755
--- a/docker/scripts/build-app.sh
+++ b/docker/scripts/build-app.sh
@@ -1,21 +1,22 @@
 #!/usr/bin/env bash
 # Author:  cirakg & claude ai
-# Summary: Compile the watch app for one device: debug, release, unit-test or UI-test build.
+# Summary: Compile the watch app for one device: debug, release, unit-test, UI-test or watchdog stress build.
 #
-# Usage:   build-app.sh debug|release|test|uitest DEVICE
+# Usage:   build-app.sh debug|release|test|uitest|stress DEVICE
 # Output:  build/<mode>/<DEVICE>.prg, and the compiler log next to it.
 #
 # Policy (README section 17.7): strict type checking (-l 3), and any compiler
 # warning fails the build. UI-test builds add uitest.jungle (fixed clock,
-# fake entries; see source/util/TestMode.mc). Test builds add test.jungle, so test code never
-# reaches debug or release builds; release builds (-r) also drop every
+# fake entries; see source/util/TestMode.mc), stress builds stress.jungle (the
+# watchdog stress test, source/util/StressTest.mc). Test builds add
+# test.jungle, so test code never reaches debug or release builds; release builds (-r) also drop every
 # (:debug) declaration, such as the diagnostics in source/util/Diag.mc.
 set -euo pipefail
 # shellcheck source=ciq-env.sh
 source "$(dirname "${BASH_SOURCE[0]}")/ciq-env.sh"
 
-mode="${1:?usage: build-app.sh debug|release|test|uitest DEVICE}"
-device="${2:?usage: build-app.sh debug|release|test|uitest DEVICE}"
+mode="${1:?usage: build-app.sh debug|release|test|uitest|stress DEVICE}"
+device="${2:?usage: build-app.sh debug|release|test|uitest|stress DEVICE}"
 
 ciq_require_sdk
 ciq_require_key
@@ -29,11 +30,12 @@ case "${mode}" in
   debug) ;;
   release) args+=(-r) ;;
   uitest) jungles="${jungles};/work/uitest.jungle" ;;
+  stress) jungles="${jungles};/work/stress.jungle" ;;
   test)
     jungles="${jungles};/work/test.jungle"
     args+=(-t)
     ;;
-  *) die "unknown build mode '${mode}' (use debug, release, test or uitest)" ;;
+  *) die "unknown build mode '${mode}' (use debug, release, test, uitest or stress)" ;;
 esac
 
 out_dir="${BUILD_DIR}/${mode}"
diff --git a/docker/scripts/screenshots.sh b/docker/scripts/screenshots.sh
index 414af52..91477c4 100755
--- a/docker/scripts/screenshots.sh
+++ b/docker/scripts/screenshots.sh
@@ -9,7 +9,9 @@
 #
 # Uses the UI-test build (uitest.jungle): fake entries, PIN 123456 and a
 # frozen clock, so every screenshot is reproducible. tools/ui_flow.py
-# drives the app and checks the screens.
+# drives the app and checks the screens. Then the app is launched again and
+# its glance, which now shows the favourite's code (the first run sealed it),
+# is checked as glance_code.png.
 set -euo pipefail
 # shellcheck source=ciq-env.sh
 source "$(dirname "${BASH_SOURCE[0]}")/ciq-env.sh"
@@ -26,6 +28,34 @@ ciq_start_display
 ciq_start_simulator
 
 out="${BUILD_DIR}/screenshots"
+
+# A new launch starts in the glance; screenshot it before anything opens the app.
+shoot_glance() {
+  local device="$1" glance_log="${out}/$1/glance.log"
+  : >"${glance_log}"
+  monkeydo "${BUILD_DIR}/uitest/${device}.prg" "${device}" >>"${glance_log}" 2>&1 &
+  CIQ_MONKEYDO_PID=$!
+  local shown=0
+  for _ in $(seq 1 600); do
+    if grep -q "DIAG shown glance_code" "${glance_log}"; then
+      shown=1
+      break
+    fi
+    sleep 0.05
+  done
+  if ((shown == 0)); then
+    ciq_stop_app
+    log "${device}: the glance with the favourite's code didn't appear; see ${glance_log}"
+    return 1
+  fi
+  sleep 1
+  local status=0
+  python3 /work/tools/ui_flow.py --shot glance_code --device-dir "${CIQ_HOME}/Devices/${device}" \
+    --window "$(ciq_simulator_window)" --out "${out}/${device}" --golden "/work/test/golden/${device}" \
+    "${approve[@]}" || status=1
+  ciq_stop_app
+  return "${status}"
+}
 mkdir -p "${out}"
 approve=()
 [[ "${APPROVE:-0}" == 1 ]] && approve=(--approve)
@@ -43,6 +73,7 @@ for device in "$@"; do
     --window "$(ciq_simulator_window)" --out "${out}/${device}" --golden "/work/test/golden/${device}" \
     "${approve[@]}" || failed=1
   ciq_stop_app
+  shoot_glance "${device}" || failed=1
 done
 
 ((failed == 0)) || die "screenshot checks failed; see build/screenshots/"