mirror of
https://github.com/ArduPilot/ardupilot.git
synced 2026-10-06 19:00:27 +08:00
Peter BarkerandClaude Opus 5
e98af74970
The FLAP_ACTUAL_SPEED subtest sets FLAP_1_SPEED=26 "above typical cruise of
~22 m/s", then commands RC3=1700 - which is not cruise. The aircraft settles
at 26.6 m/s, *above* the threshold it just set, so the flaps deploy on the way
up through 26 and retract again a few seconds later:
t=52.9 airspeed 24.37 flap 1380 (deployed)
t=55.4 airspeed 26.08 flap 1380
t=56.2 airspeed 26.31 flap 1200 (retracted)
t=59.6 airspeed 26.60 flap 1200 ... for the rest of the test
Both assertions can therefore only hold during that 3.3s transient, and
wait_airspeed(18, 24) needs its first sample inside the ~0.3s the aircraft
spends between 22 and 24 while accelerating at 7.7 m/s/s. On a slower
machine the first sample landed at 24.15 - 0.15 m/s late - and the test
failed; on faster ones it landed at 23.2-23.4 and passed. Every pass so far
has been won on that transient, on a premise which is false at equilibrium.
Fly the subtest at cruise throttle instead, so measured airspeed settles
below FLAP_1_SPEED and the flaps stay deployed, and require the airspeed to
be *held* in the band rather than merely pass through it - so if a future
change moves the settled speed back out of the band the test says so instead
of quietly going back to sampling a transient.
Settled airspeed at RC3=1500, 6 runs: 22.4 - 22.8 m/s, 6/6 pass.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>