lab: a crashing ble suite still writes its report

The report is written at the end of the run, so an exception took the
whole run's evidence with it. A MacBook pass on 2026-08-28 reached its
eighth scenario, raised on a missing mpremote and left nothing at all -
not even the seven scenarios that had already run against the phone,
which were the ones worth reading. An hour of bench time producing no
record of what the phone did.

The crash is now caught, printed with its traceback, and recorded as a
finding per device, so the run finishes and writes what it has.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

Author: Staacks

tools/lab/README.md

@@ -111,8 +111,10 @@ one is missing - a host that was never set up looks nothing like a board
 fault, but discovered per scenario it arrives as two failed flashes and a
 board declared dead. On the MacBook on 2026-08-28 a missing `mpremote`
 was worse still: the run raised on the first MicroPython scenario, and
-the operator watched an idle phone until the run ended before seeing
-why. pyserial is deliberately NOT in requirements.txt: the docs build and app CI install
+the operator watched an idle phone until the run ended before seeing why
+- and because the report is written only at the end, the seven scenarios
+that HAD already run against the phone were lost with it. A crashing
+suite now still produces a report, with the exception as its finding. pyserial is deliberately NOT in requirements.txt: the docs build and app CI install
 that file and never open a serial port. It is needed only for the three
 scenarios where the data goes phone -> board and the board's own printout
 is the evidence.

tools/lab/run.py

@@ -473,7 +473,25 @@ def main():
             # orchestration and only hands back the per-device results.
             if suite == "ble":
                 from lab import ble
-                for dev_id, r in ble.run_suite(devices, args).items():
+                try:
+                    per_device = ble.run_suite(devices, args)
+                except Exception as e:
+                    # A crashed suite must still produce a report. The
+                    # write happens at the end of this function, so an
+                    # exception used to take the whole run's evidence with
+                    # it: on 2026-08-28 a MacBook pass reached its eighth
+                    # scenario, raised on a missing mpremote, and left
+                    # nothing at all - not even the seven scenarios that
+                    # had already run against the phone, which were the
+                    # ones worth reading.
+                    import traceback
+                    traceback.print_exc()
+                    per_device = {
+                        dev_id: {"passed": False, "warnings": [],
+                                 "findings": [f"the ble suite crashed: "
+                                              f"{type(e).__name__}: {e}"]}
+                        for dev_id, _entry, _dev in devices}
+                for dev_id, r in per_device.items():
                     report["devices"][dev_id]["ble"] = r
                     state = "ok" if r.get("passed") else "FAIL"
                     print(f"== ble @ {dev_id}: {state}"