fix: harden task recovery and DR contract handling

This commit is contained in:
2026-09-20 13:36:48 +08:00
parent 492676344a
commit f9d8feb6f0
49 changed files with 2083 additions and 165 deletions
+3 -2
View File
@@ -9,7 +9,8 @@ flags=(-std=c++17 -O2 -Wall -Wextra -Werror -pthread -Icore/include)
"$cxx" "${flags[@]}" -c core/src/workflow.cpp -o build/workflow.o
"$cxx" "${flags[@]}" -c core/src/sim_driver.cpp -o build/sim_driver.o
"$cxx" "${flags[@]}" core/examples/demo.cpp build/core.o build/workflow.o build/sim_driver.o -o build/bt_demo
for test_name in core_test workflow_test journal_failure_test preflight_test settlement_test proof_regression_test readiness_regression_test scenario_test lifecycle_test; do
"$cxx" "${flags[@]}" "core/tests/$test_name.cpp" build/core.o build/workflow.o -o "build/$test_name"
for test_source in core/tests/*_test.cpp; do
test_name="$(basename "$test_source" .cpp)"
"$cxx" "${flags[@]}" -UNDEBUG "$test_source" build/core.o build/workflow.o -o "build/$test_name"
"./build/$test_name"
done
+18 -5
View File
@@ -16,8 +16,14 @@ import trace
import unittest
ROOT = Path(__file__).resolve().parents[1]
CORE_TESTS = ('core_test', 'workflow_test', 'journal_failure_test', 'preflight_test',
'settlement_test', 'proof_regression_test', 'readiness_regression_test', 'scenario_test', 'lifecycle_test')
CORE_TESTS = tuple(path.stem for path in sorted((ROOT / 'core/tests').glob('*_test.cpp')))
def executable_lines(path):
# Python 3.14 can include artificial bytecode positions without a source
# line; these are not executable source statements and cannot be sorted.
return {line for line in trace._find_executable_linenos(str(path))
if type(line) is int and line > 0}
def production(path):
@@ -35,13 +41,20 @@ def python_lines(destination):
# are deliberately not counted as covered without an observed trace event.
result = tracer.runfunc(unittest.TextTestRunner(verbosity=1).run, suite)
counts = tracer.results().counts
observed = {}
resolved = {}
for (filename, line), count in counts.items():
if count <= 0:
continue
if filename not in resolved:
resolved[filename] = Path(filename).resolve()
observed.setdefault(resolved[filename], set()).add(line)
files = []
for path in sorted(ROOT.rglob('*.py')):
if not production(path):
continue
statements = set(trace._find_executable_linenos(str(path)))
hit = {line for (filename, line), count in counts.items()
if Path(filename).resolve() == path and count > 0} & statements
statements = executable_lines(path)
hit = observed.get(path, set()) & statements
files.append(dict(path=str(path.relative_to(ROOT)), statements=len(statements),
covered=len(hit), missing=sorted(statements-hit)))
data = dict(metric='Python executable source lines; main test thread only',
+71
View File
@@ -0,0 +1,71 @@
#!/usr/bin/env python3
"""Measure fresh-connection task admission bursts against an isolated local API.
The planner is deliberately idle: receipts must not wait for planning. Latency
is reported as evidence, never asserted as a machine-independent unit-test SLA.
"""
import argparse
from concurrent.futures import ThreadPoolExecutor
import http.client
import json
import math
from pathlib import Path
import sys
import tempfile
import threading
import time
ROOT=Path(__file__).resolve().parents[1]
sys.path.insert(0,str(ROOT/'coordinator'))
from robot_bt_coordinator.backends import ManualBackend
from robot_bt_coordinator.http_api import make_server
from robot_bt_coordinator.service import Coordinator
def probe(clients=10,rounds=3):
if not 1<=clients<=128 or not 1<=rounds<=100:
raise ValueError('clients must be 1..128 and rounds 1..100')
with tempfile.TemporaryDirectory(prefix='bt-http-probe-') as folder:
coordinator=Coordinator(str(Path(folder)/'tasks.db'),ManualBackend(),{'probe_robot'})
server=make_server(coordinator,'127.0.0.1',0,'local-probe-token')
thread=threading.Thread(target=server.serve_forever,daemon=True)
thread.start()
samples=[]
try:
for iteration in range(rounds):
gate=threading.Barrier(clients)
def submit(index):
connection=http.client.HTTPConnection('127.0.0.1',server.server_port,timeout=10)
request=dict(client_request_id=f'probe-{iteration}-{index}',robot_id='probe_robot',instruction='local admission probe',known_info={})
gate.wait(timeout=15)
started=time.perf_counter()
try:
connection.request('POST','/v1/tasks',json.dumps(request),{'Authorization':'Bearer local-probe-token','Content-Type':'application/json'})
response=connection.getresponse()
body=json.loads(response.read())
return dict(elapsed_ms=(time.perf_counter()-started)*1000,http_status=response.status,task_id=body.get('task_id'),status=body.get('status'))
finally:connection.close()
with ThreadPoolExecutor(max_workers=clients) as pool:
samples.extend(pool.map(submit,range(clients)))
successful=all(s['http_status']==202 and s['status']=='QUEUED' for s in samples)
successful=successful and len({s['task_id'] for s in samples})==clients*rounds
durations=sorted(s['elapsed_ms'] for s in samples)
return dict(clients=clients,rounds=rounds,requests=len(samples),all_accepted=successful,
p50_ms=durations[math.ceil(len(durations)*.5)-1],
p95_ms=durations[math.ceil(len(durations)*.95)-1],max_ms=max(durations),
listen_backlog=server.request_queue_size,planner_started=bool(coordinator.backend.plans),samples=samples)
finally:
server.shutdown();server.server_close();thread.join();coordinator.close()
def main():
parser=argparse.ArgumentParser()
parser.add_argument('--clients',type=int,default=10)
parser.add_argument('--rounds',type=int,default=3)
args=parser.parse_args()
result=probe(args.clients,args.rounds)
print(json.dumps(result,indent=2))
if not result['all_accepted'] or result['planner_started']:raise SystemExit(1)
if __name__=='__main__':main()
+1
View File
@@ -3,6 +3,7 @@ set -euo pipefail
project_root="$(cd "$(dirname "$0")/.." && pwd)"
cd "$project_root"
bash tools/build_portable.sh
bash tools/test_recovery_policy.sh
python3 -m unittest discover -s tests -p 'test_*.py' -v
python3 ros2/bt_executor/tools/check_static.py
python3 ros2/bt_executor/tools/test_ros_backend.py
+9
View File
@@ -0,0 +1,9 @@
#!/usr/bin/env bash
set -euo pipefail
project_root="$(cd "$(dirname "$0")/.." && pwd)"
cd "$project_root"
mkdir -p build
"${CXX:-g++}" -std=c++17 -O2 -Wall -Wextra -Werror -UNDEBUG \
-Iros2/bt_executor/include ros2/bt_executor/tools/test_recovery_policy.cpp \
-o build/recovery_policy_test
./build/recovery_policy_test