ProcessPoolExecutor swallows falsy Exceptions

[graingert]

Bug report

Bug description:

import sys
import concurrent.futures

class MyException(Exception):

    def __bool__(self):
        return False


def raiser():
    raise MyException("foo")

def main():
    with concurrent.futures.ProcessPoolExecutor() as p:
        print(p.submit(raiser).result())

if __name__ == "__main__":
    sys.exit(main())

This program prints None, but should fail and print a traceback

CPython versions tested on:

CPython main branch, 3.14, 3.13, 3.12, 3.11, 3.10, 3.9

Operating systems tested on:

Linux

[YvesDup]

I think there is something wrong here in the __get_result(self) method of Future class:

cpython/Lib/concurrent/futures/_base.py

Lines 393 to 400 in 6ab4ace

	if self._exception:
	try:
	raise self._exception
	finally:
	# Break a reference cycle with the exception in self._exception
	self = None
	else:
	return self._result

The test should be: if self._exception is not None: because if self._exception calls the __bool__ method of exception.

Edit: same problem here on line 443

cpython/Lib/concurrent/futures/process.py

Lines 441 to 446 in 6ab4ace

	# work_item can be None if another process terminated (see above)
	if work_item is not None:
	if result_item.exception:
	work_item.future.set_exception(result_item.exception)
	else:
	work_item.future.set_result(result_item.result)

[picnixz]

The test should be: if self._exception is not None: because if self._exception calls the bool method of exception.

Well spotted. Indeed, this should be changed to is not None. More generally, are there other places in the stdlib (or in mp-related code) where we have this issue?

[YvesDup]

More generally, are there other places in the stdlib

I saw these tests this morning:

• cpython/Lib/traceback.py

  Line 1123 in 255eb37

  if (e and e.__cause__ is not None

• cpython/Lib/traceback.py

  Line 1144 in 255eb37

  if (e and e.__context__ is not None

• cpython/Lib/traceback.py

  Line 1159 in 255eb37

  if e and isinstance(e, BaseExceptionGroup):

May I open a PR ?

[picnixz]

Unless they can make a silly bug, I think it's fine to keep them. However, unless @graingert wants to work on the issue itself, you may first fix the __get_result method first.

[YvesDup]

It's mandatory to work on both files /Lib/concurrent/futures/process.py and /Lib/concurrent/futures/_base.pyto fix this issue.

This should fix the same issue with a ThreadPoolExcecutor.

[graingert]

Please feel free to work on this, thanks! Thomas Grainger

…

On Fri, 4 Apr 2025, 15:41 Duprat, ***@***.***> wrote: It's mandatory to work on both files /Lib/concurrent/futures/process.py *and* /Lib/concurrent/futures/_base.pyto fix this issue. This should fix the same issue with a ThreadPoolExcecutor. — Reply to this email directly, view it on GitHub <#132063 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AADFATAV3324MMPIVXQ7JYL2X2KZHAVCNFSM6AAAAAB2OBLA4OVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDONZYHEZTGMRQGE> . You are receiving this because you were mentioned.Message ID: ***@***.***> [image: YvesDup]*YvesDup* left a comment (python/cpython#132063) <#132063 (comment)> It's mandatory to work on both files /Lib/concurrent/futures/process.py *and* /Lib/concurrent/futures/_base.pyto fix this issue. This should fix the same issue with a ThreadPoolExcecutor. — Reply to this email directly, view it on GitHub <#132063 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AADFATAV3324MMPIVXQ7JYL2X2KZHAVCNFSM6AAAAAB2OBLA4OVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDONZYHEZTGMRQGE> . You are receiving this because you were mentioned.Message ID: ***@***.***>