autotest

name	autotest
description	Run ArduPilot SITL autotests (integration/behavior tests) and inspect their results. Use when the user asks to run autotests, vehicle tests, specific test methods, or to examine why an autotest failed.
argument-hint	<vehicle> [test_name]
disable-model-invocation	true
allowed-tools	Bash(python3 *autotest*), Bash(python3 *autotest_results.py*), Bash(./waf *), Read, Grep

Run ArduPilot Autotests

Autotests are Python-based integration tests that run vehicles in SITL simulation.

Argument parsing

Parse $ARGUMENTS for vehicle and optional test name:

• /autotest Copter — build and run all Copter tests
• /autotest Copter AltHold — run specific Copter test
• /autotest Plane QuadPlane — run specific Plane test
• /autotest --list Copter — list available Copter tests

Workflow

Step 1: Build the vehicle first

Always build separately using /build or ./waf — do NOT use build.<Vehicle> in autotest.py:

./waf configure --board sitl    # if not already configured
./waf copter                    # or plane, rover, sub, etc.

Step 2: List available tests (optional)

python3 Tools/autotest/autotest.py --list-subtests-for-vehicle=<Vehicle>

Vehicle names: Copter, Plane, Rover, Sub, Tracker, Helicopter, QuadPlane, BalanceBot, Sailboat, Blimp

Step 3: Run tests

# Run all tests for a vehicle
python3 Tools/autotest/autotest.py test.<Vehicle>

# Run a specific test method
python3 Tools/autotest/autotest.py test.<Vehicle>.<TestMethod>

Useful options

# Show test timing info
python3 Tools/autotest/autotest.py --show-test-timings test.Copter.AltHold

# Run with debug build (build with --debug first via /build)
python3 Tools/autotest/autotest.py --debug test.Copter.AltHold

Bounding a run that might hang

Use the runner script - it wraps autotest.py with a wall-clock timeout, streams output, and warns about a stale lock, all under one pre-authorized script (so we grant the script, not a blanket timeout/python3):

python3 .claude/skills/autotest/run_autotest.py test.Copter.<Test>
python3 .claude/skills/autotest/run_autotest.py --timeout 1200 test.Plane.QuadPlane

It exits 124 on timeout and kills the SITL children. Build first with /build.

Do NOT reach for & + pkill, nohup, launching build/sitl/bin/arducopter directly, or an env-var prefix like PYTHONUNBUFFERED=1 python3 .... They are the wrong tools (drive SITL through autotest.py / sim_vehicle.py), and the env prefix also changes the command's first token so it no longer matches the pre-authorized script and prompts needlessly. If a previous run left a live lock, clear the wedged process (with the user's ok) - don't work around it by hand. Inspect results with autotest_results.py (below), not tail/grep over the logs.

Test file locations

Vehicle	Test file
Copter	Tools/autotest/arducopter.py
Plane	Tools/autotest/arduplane.py
Rover	Tools/autotest/rover.py
Sub	Tools/autotest/ardusub.py
Helicopter	Tools/autotest/helicopter.py
Tracker	Tools/autotest/antennatracker.py

Report results

Use the autotest_results.py helper at .claude/skills/autotest/autotest_results.py to inspect results — do not grep / tail / head per-test files by hand. The helper parses the per-test buildlogs into structured output.

# Pass/fail summary across all tests in buildlogs
python3 .claude/skills/autotest/autotest_results.py summary

# Filter by vehicle
python3 .claude/skills/autotest/autotest_results.py summary --vehicle ArduCopter

# Show every failing test with reason + exception block
python3 .claude/skills/autotest/autotest_results.py failures

# Full failure context for one test (last N lines + exception)
python3 .claude/skills/autotest/autotest_results.py failure AltHold --lines 150

# List .BIN / .tlog logs produced by the run (feed these into /log-analyze)
python3 .claude/skills/autotest/autotest_results.py logs

Default --buildlogs is $BUILDLOGS or ../buildlogs (matches the autotest harness default). Override with --buildlogs <dir> if needed.

When reporting back to the user:

• Lead with the pass/fail counts.
• Quote the failure reason and exception line for each failing test (from failures).
• Mention any .BIN / .tlog logs available for further analysis with /log-analyze.
• Don't paste hundreds of lines of raw test output — extract the relevant signal.

Common failure patterns

• Timeout waiting for message — vehicle didn't reach expected state in time
• Altitude/position check failed — vehicle didn't hit waypoint or target
• Mode change rejected — arming checks or pre-conditions not met
• Build failure — fix build first before running tests