Interpreter crash on application shutdown

[tb220]

BPO	8229
Nosy	@birkenfeld, @serwy, @bitdancer, @briancurtin
Files	cabb_appcompat.txt: Error report generated by Windows

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = <Date 2013-10-13.18:32:57.198>
created_at = <Date 2010-03-25.08:50:56.612>
labels = ['interpreter-core', 'OS-windows', 'type-crash']
title = 'Interpreter crash on application shutdown'
updated_at = <Date 2013-10-13.18:32:57.196>
user = 'https://bugs.python.org/tb220'

bugs.python.org fields:

activity = <Date 2013-10-13.18:32:57.196>
actor = 'georg.brandl'
assignee = 'none'
closed = True
closed_date = <Date 2013-10-13.18:32:57.198>
closer = 'georg.brandl'
components = ['Interpreter Core', 'Windows']
creation = <Date 2010-03-25.08:50:56.612>
creator = 'tb220'
dependencies = []
files = ['16642']
hgrepos = []
issue_num = 8229
keywords = []
message_count = 7.0
messages = ['101679', '101684', '101690', '101701', '101731', '165541', '199756']
nosy_count = 5.0
nosy_names = ['georg.brandl', 'roger.serwy', 'r.david.murray', 'brian.curtin', 'tb220']
pr_nums = []
priority = 'normal'
resolution = 'out of date'
stage = 'test needed'
status = 'closed'
superseder = None
type = 'crash'
url = 'https://bugs.python.org/issue8229'
versions = ['Python 2.6']

[tb220]

Attached the error report generated by Windows. The problem occurs in 1 out of 10 shutdowns.

[tb220]

Please let me know how I can provide you with additional information.

[bitdancer]

A recipe for reproducing the problem would be the most useful thing. I suspect the windows error report is pretty much useless in this context.

[briancurtin]

Are you running a service? I can't say I'm too familiar with the file you uploaded, but the file contains a few things that lead me to believe you may be. As David said, it would be helpful to know what you were doing.

[tb220]

The application consists of three processes: Process A provides a web interface to the remote user and watchdog services for process B, which A creates by using the multiprocessing module. Process B does the real work and provides a GUI to the local user. Process A communicates with process B via the xmlrpclib module. Both A and B are python processes. Process B uses the subprocess module to start process C, which is not a python process, and it communicates with it using the xmlrpclib module. Process B is the one that crashes every now and then when sys.exit is invoked.

I will try to produce the smallest possible application I can that still suffers from the problem and I will upload it.

[serwy]

tb220, is this still a bug? Python 2.6 is now in security-fix-only mode.

[birkenfeld]

Closing due to lack of feedback.