Repository navigation
for loop swallows exceptions raised by a sys.monitoring callback for StopIteration #160676
Description
Bug report
When a sys.monitoring.events.RAISE callback raises RuntimeError while reporting StopIteration at a for loop, the RuntimeError is swallowed. Execution continues after the loop.
Reproducer
Save as mre_monitoring_stopiteration.py and run python mre_monitoring_stopiteration.py.
import sys
class Empty:
def __iter__(self):
return self
def __next__(self):
raise StopIteration
def run():
for _ in Empty():
pass
def on_raise(code, offset, exc):
if code is run.__code__ and isinstance(exc, StopIteration):
print("callback: raising RuntimeError")
raise RuntimeError("monitor callback failed")
monitoring = sys.monitoring
monitoring.use_tool_id(3, "stopiteration-repro")
try:
monitoring.register_callback(3, monitoring.events.RAISE, on_raise)
monitoring.set_events(3, monitoring.events.RAISE)
run()
finally:
monitoring.set_events(3, 0)
monitoring.register_callback(3, monitoring.events.RAISE, None)
monitoring.free_tool_id(3)
raise AssertionError("RuntimeError from the monitoring callback was swallowed")Expected behavior
RuntimeError: monitor callback failed propagates out of run(). The script does not reach the final AssertionError.
Actual behavior
The callback prints:
callback: raising RuntimeError
However, run() returns normally, and the script reaches the final assertion:
AssertionError: RuntimeError from the monitoring callback was swallowed
Replacing the loop in run() with next(Empty()) makes the same callback's RuntimeError propagate on the tested main build.
CPython versions tested on:
CPython main branch, 3.16, 3.15, 3.14
Operating systems tested on:
Linux, macOS
Activity
pythongh-160676: Add NEWS entry for monitoring callback error fix
pythongh-160676: Cover tier-2 monitoring callback errors
pythongh-160676: Fix tier-2 monitoring regression test