Using pickled/unpickled sqlite3.Row results in segfault rather than exception

[Elizacat]

BPO	21975
Nosy	@PCManticore, @serhiy-storchaka
Files	testcase.py: Test case trace.txt issue21975.patch sqlite3_row_new.patch

^{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 = 'https://github.com/serhiy-storchaka'
closed_at = <Date 2014-08-06.14:57:36.397>
created_at = <Date 2014-07-14.00:01:29.138>
labels = ['library', 'type-crash']
title = 'Using pickled/unpickled sqlite3.Row results in segfault rather than exception'
updated_at = <Date 2014-08-06.14:57:36.396>
user = 'https://bugs.python.org/Elizacat'

bugs.python.org fields:

activity = <Date 2014-08-06.14:57:36.396>
actor = 'serhiy.storchaka'
assignee = 'serhiy.storchaka'
closed = True
closed_date = <Date 2014-08-06.14:57:36.397>
closer = 'serhiy.storchaka'
components = ['Library (Lib)']
creation = <Date 2014-07-14.00:01:29.138>
creator = 'Elizacat'
dependencies = []
files = ['35947', '35948', '35952', '35953']
hgrepos = []
issue_num = 21975
keywords = ['patch']
message_count = 11.0
messages = ['222982', '222983', '223001', '223015', '223019', '223020', '223021', '223022', '223024', '223029', '224936']
nosy_count = 5.0
nosy_names = ['ghaering', 'Claudiu.Popa', 'python-dev', 'serhiy.storchaka', 'Elizacat']
pr_nums = []
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'crash'
url = 'https://bugs.python.org/issue21975'
versions = ['Python 2.7', 'Python 3.4', 'Python 3.5']

[Elizacat]

Pickling, unpickling, then using an sqlite3.Row object results in a segfault on at least Python 3.3.5, 3.4.0, and 3.4.1. I have attached a test case and a backtrace below.

I know you're not supposed to pickle sqlite3.Row objects, as the given test case below results in an exception to the effect that sqlite3.Row objects cannot be pickled in Python 2.7. I don't think a segfault is the desired behaviour, however...

[Elizacat]

The backtrace of the crash, if it helps

[PCManticore]

Hi, thanks for the report. Here's a patch which implements __setstate__ and __getstate__ for Row objects.

[serhiy-storchaka]

The issue is not in pickling/unpickling, but in sqlite3.Row.__new__ which creates object in invalid state. Simple example:

>>> import sqlite3
>>> r = sqlite3.Row.__new__(sqlite3.Row)
>>> len(r)
Segmentation fault (core dumped)

[PCManticore]

Using your example, I can't make it to crash using the tip, nor with Python 3.4.

[PCManticore]

Nevermind, I didn't see the len call.

[serhiy-storchaka]

Here is a patch which fixes crash.

[PCManticore]

It doesn't crash anymore with your patch, but the pickle.load fails:

Traceback (most recent call last):
  File "a.py", line 19, in <module>
    load = pickle.loads(dump)
TypeError: function takes exactly 2 arguments (0 given)

[serhiy-storchaka]

For pickling open other issue.

[PCManticore]

Ups, I accidentally removed the patch review stage, sorry for that.

[python-dev]

New changeset c46ad743bcb4 by Serhiy Storchaka in branch '2.7':
Issue bpo-21975: Fixed crash when using uninitialized sqlite3.Row (in particular
http://hg.python.org/cpython/rev/c46ad743bcb4

New changeset c1ca1c4c131b by Serhiy Storchaka in branch '3.4':
Issue bpo-21975: Fixed crash when using uninitialized sqlite3.Row (in particular
http://hg.python.org/cpython/rev/c1ca1c4c131b

New changeset 9244ed41057a by Serhiy Storchaka in branch 'default':
Issue bpo-21975: Fixed crash when using uninitialized sqlite3.Row (in particular
http://hg.python.org/cpython/rev/9244ed41057a