IQ.Pilot Release Commit @ e46d557

This commit is contained in:
IQ.Lvbs CI [bot]
2026-08-07 00:13:39 -05:00
parent 03c3158b81
commit 824bb9ddfd
216 changed files with 7457 additions and 3151 deletions

View File

@@ -17,6 +17,10 @@ from openpilot.common.basedir import BASEDIR
from openpilot.common.params import Params
from openpilot.common.swaglog import cloudlog
MAX_CRASH_BACKOFF = 300.0
CRASH_RESET_TIME = 60.0
CRASH_LOOP_THRESHOLD = 6
try:
from openpilot.system.proprietary_runtime.runtime_paths import preferred_runner_path
except ModuleNotFoundError:
@@ -82,6 +86,10 @@ class ManagerProcess(ABC):
name = ""
shutting_down = False
restart_if_crash = False
crash_count = 0
last_restart_time = 0.0
last_alive_time = 0.0
crash_loop_logged = False
@abstractmethod
def prepare(self) -> None:
@@ -318,13 +326,33 @@ def ensure_running(procs: ValuesView[ManagerProcess], started: bool, params=None
not_run = []
running = []
now = time.monotonic()
for p in procs:
if p.enabled and p.name not in not_run and p.should_run(started, params, CP):
if p.restart_if_crash and p.proc is not None and not p.proc.is_alive():
cloudlog.error(f'Restarting {p.name} (exitcode {p.proc.exitcode})')
p.restart()
if p.restart_if_crash and p.proc is not None and p.proc.is_alive():
p.last_alive_time = now
elif p.restart_if_crash and p.proc is not None:
# uptime, not time-since-restart: the latter also counts the backoff wait,
# which would reset the counter as soon as backoff exceeds CRASH_RESET_TIME
if p.last_alive_time - p.last_restart_time > CRASH_RESET_TIME:
p.crash_count = 0
p.crash_loop_logged = False
backoff = 0.0 if not p.crash_count else min(MAX_CRASH_BACKOFF, 2.0 ** (p.crash_count - 1))
if now - p.last_restart_time >= backoff:
p.crash_count += 1
p.last_restart_time = now
cloudlog.error(f'Restarting {p.name} (exitcode {p.proc.exitcode}) [crash {p.crash_count}]')
if p.crash_count >= CRASH_LOOP_THRESHOLD and not p.crash_loop_logged:
# never stop retrying: giving up on hardwared or ui is worse than restarting slowly
cloudlog.error(f'{p.name} is in a crash loop, backing off to {MAX_CRASH_BACKOFF}s between restarts')
p.crash_loop_logged = True
p.restart()
running.append(p)
else:
p.crash_count = 0
p.crash_loop_logged = False
p.last_alive_time = 0.0
p.stop(block=False)
for p in running:

View File

@@ -98,7 +98,9 @@ def constructiond_onroad(started: bool, params: Params, CP: car.CarParams) -> bo
return started and params.get_bool("ConstructionZoneAssist")
def iqvd_onroad(started: bool, params: Params, CP: car.CarParams) -> bool:
return started and params.get_bool("VisionVehicleTracks")
# held for 1.0d: iqvd runs a detector per frame and the added load is not
# something 1.0c needs to carry. re-enable by restoring the param check.
return False
def only_offroad(started: bool, params: Params, CP: car.CarParams) -> bool:
return not started
@@ -109,6 +111,12 @@ def livestream(started: bool, params: Params, CP: car.CarParams) -> bool:
# down cleanly when the session ends — no subprocess management inside hephaestusd.
return params.get_bool("IsLiveStreaming")
def canlive(started: bool, params: Params, CP: car.CarParams) -> bool:
# Remote live CAN debugging via konn3kt. hephaestusd sets CanLiveStreaming when a viewer
# connects (startCanLive) and clears it when the last one leaves (stopCanLive), so canlived
# runs only during an active debug session — no idle connection or battery cost otherwise.
return params.get_bool("CanLiveStreaming")
def is_tinygrad_model(started, params, CP: car.CarParams) -> bool:
"""Check if the active model runner is tinygrad."""
return bool(get_active_model_runner(params, not started) == custom.IQModelManager.Runner.tinygrad)
@@ -179,6 +187,7 @@ procs = [
# debug procs
NativeProcess("bridge", "cereal/messaging", ["./bridge"], notcar),
PythonProcess("webrtcd", "system.webrtc.webrtcd", or_(iscar, livestream)),
PythonProcess("canlived", "iqpilot.konn3kt.canlive.canlived", canlive),
PythonProcess("webjoystick", "tools.bodyteleop.web", notcar),
]