fix(e2e): headless Graph construction + keep simulator cmdloop alive

Run 10537 reached node boot but the simulator crashed again at Graph
construction: lib/gui.py move_figure() unconditionally touches the Tk window
manager (canvas.manager.window), which does not exist under Agg. Patch it to a
no-op guard alongside the TkAgg->Agg backend swap. Also feed the simulator's
interactive cmdloop from a FIFO held open until teardown — CI stdin is EOF, so
the sim would otherwise exit immediately after booting the nodes.

scripts/meshtasticator_e2e/README.md

@@ -85,11 +85,16 @@ simulator. Nothing is forwarded when the daemon is local (unset/unix
 on the forwarders themselves.
 
 Other headless adaptations the script makes automatically:
-* patches `lib/gui.py` `matplotlib.use("TkAgg")` → `"Agg"` in the pinned
-  Meshtasticator checkout — the TkAgg call crashes headless runs even under
-  `MPLBACKEND=Agg` (only the interactive GUI mode is lost),
+* patches `lib/gui.py` in the pinned Meshtasticator checkout: forces the Agg
+  backend (the code calls `matplotlib.use("TkAgg")` at import) and makes
+  `move_figure()` a no-op when no Tk window manager exists (its unconditional
+  `canvas.manager.window` access crashes under Agg at Graph construction).
+  Only the interactive GUI modes, never used by this harness, are lost,
 * warms the `meshtastic/meshtasticd` image (`docker pull`) before booting the
-  simulator so the pull is not on the simulator's node-boot critical path.
+  simulator so the pull is not on the simulator's node-boot critical path,
+* feeds the simulator's interactive `cmdloop` from a FIFO held open until
+  teardown — an EOF stdin (e.g. CI) would otherwise make it exit right after
+  booting the nodes.
 
 Without `MESHTASTICATOR_E2E=1` the e2e module skips, so ordinary unit/CI runs
 stay fast and gateway-free.

scripts/meshtasticator_e2e/run_e2e.sh

@@ -109,14 +109,33 @@ else
   PROBE_HOST="$HOST"
 fi
 
-# The simulator forces TkAgg at import (lib/gui.py: matplotlib.use("TkAgg")),
-# which crashes headless runs ("Tkinter is needed") even under MPLBACKEND=Agg —
-# an explicit use() overrides the env var. Neutralize it in the pinned
-# checkout; only the interactive GUI mode (never used by this harness) is lost.
-if grep -q 'matplotlib.use("TkAgg")' "$SIM_DIR/lib/gui.py"; then
-  echo "== patching meshtasticator lib/gui.py: TkAgg -> Agg (headless) =="
-  sed -i.bak 's/matplotlib.use("TkAgg")/matplotlib.use("Agg")/' "$SIM_DIR/lib/gui.py"
-fi
+# The simulator assumes an interactive Tk desktop: lib/gui.py forces
+# matplotlib.use("TkAgg") at import and move_figure() touches the Tk window
+# manager at Graph construction — both crash headless runs even under
+# MPLBACKEND=Agg. Neutralize them in the pinned checkout (only the interactive
+# GUI modes, never used by this harness, are lost).
+echo "== patching meshtasticator lib/gui.py for headless runs =="
+"$VENV/bin/python" - "$SIM_DIR" <<'PY'
+import sys
+from pathlib import Path
+
+gui = Path(sys.argv[1]) / "lib/gui.py"
+src = gui.read_text()
+
+src = src.replace('matplotlib.use("TkAgg")', 'matplotlib.use("Agg")')
+
+old = 'def move_figure(fig, x, y):\n    fig.canvas.manager.window.wm_geometry("+%d+%d" % (x, y))'
+new = ('def move_figure(fig, x, y):\n'
+       '    # Headless (Agg) backends have no Tk window manager.\n'
+       '    try:\n'
+       '        fig.canvas.manager.window.wm_geometry("+%d+%d" % (x, y))\n'
+       '    except AttributeError:\n'
+       '        pass')
+assert old in src, "move_figure source no longer matches the pinned meshtasticator"
+src = src.replace(old, new)
+gui.write_text(src)
+print("patched lib/gui.py")
+PY
 
 if [ "$MODE" = docker ]; then
   # Warm the node image on the (possibly remote) daemon first: the simulator
@@ -127,14 +146,22 @@ if [ "$MODE" = docker ]; then
 fi
 
 echo "== booting Meshtasticator ($MODE) with $NODES node(s) from $SIM_DIR =="
+# The simulator ends in an interactive cmdloop reading stdin; an EOF there
+# (e.g. CI) would make it exit right after booting the nodes. Feed it from a
+# FIFO whose write end this shell keeps open until cleanup.
+SIM_FIFO="$(mktemp -u /tmp/meshtasticator-XXXXXX.fifo)"
+mkfifo "$SIM_FIFO"
 (
   cd "$SIM_DIR"
-  "$SIM_PYTHON" interactiveSim.py "${SIM_ARGS[@]}" >"$SIM_LOG" 2>&1
+  "$SIM_PYTHON" interactiveSim.py "${SIM_ARGS[@]}" <"$SIM_FIFO" >"$SIM_LOG" 2>&1
 ) &
 SIM_PID=$!
+exec 9>"$SIM_FIFO"   # hold the write end open -> sim stdin never hits EOF
+rm -f "$SIM_FIFO"
 
 cleanup() {
   echo "== tearing down simulator (pid $SIM_PID) =="
+  exec 9>&- 2>/dev/null || true   # release the FIFO write end (sim stdin EOF)
   kill "$SIM_PID" 2>/dev/null || true
   wait "$SIM_PID" 2>/dev/null || true
   for pid in "${FWD_PIDS[@]:-}"; do