super.__init__ leaks memory if called multiple times

[KevinModzelewski]

BPO	26718
Nosy	@gvanrossum, @brettcannon, @vstinner, @serhiy-storchaka
Files	super_init_leaks.patch super_init_leaks_2.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 2016-04-13.12:38:38.479>
created_at = <Date 2016-04-09.01:02:03.209>
labels = ['interpreter-core', 'performance']
title = 'super.__init__ leaks memory if called multiple times'
updated_at = <Date 2016-04-13.12:38:38.478>
user = 'https://bugs.python.org/KevinModzelewski'

bugs.python.org fields:

activity = <Date 2016-04-13.12:38:38.478>
actor = 'serhiy.storchaka'
assignee = 'serhiy.storchaka'
closed = True
closed_date = <Date 2016-04-13.12:38:38.479>
closer = 'serhiy.storchaka'
components = ['Interpreter Core']
creation = <Date 2016-04-09.01:02:03.209>
creator = 'Kevin Modzelewski'
dependencies = []
files = ['42445', '42447']
hgrepos = []
issue_num = 26718
keywords = ['patch']
message_count = 12.0
messages = ['263053', '263073', '263103', '263195', '263196', '263232', '263249', '263253', '263254', '263270', '263276', '263331']
nosy_count = 6.0
nosy_names = ['gvanrossum', 'brett.cannon', 'vstinner', 'python-dev', 'serhiy.storchaka', 'Kevin Modzelewski']
pr_nums = []
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'resource usage'
url = 'https://bugs.python.org/issue26718'
versions = ['Python 2.7', 'Python 3.5', 'Python 3.6']

[KevinModzelewski]

The super() __init__ function fills in the fields of a super object without checking if they were already set. If someone happens to call __init__ again, the previously-set references will end up getting forgotten and leak memory.

For example:

import sys
print(sys.gettotalrefcount())
sp = super(int, 1)
for i in range(100000):
  super.__init__(sp, float, 1.0)
print(sys.gettotalrefcount())

[serhiy-storchaka]

If super used __new__ instead of __init__, this issue probably wouldn't arise.

I'm surprised that super is subclassable.

[gvanrossum]

Any uses of super() beyond the documented idioms are uninteresting, except they should not be usable as crash or DoS vectors.

[brettcannon]

Based on Guido's feedback and the fact that this isn't documented usage of super() and hence no promises to not re-initialize, I'm closing as "not a bug". Sorry, Kevin.

[gvanrossum]

Actually, the refcount bug is still a bug.

[serhiy-storchaka]

Possible solutions:

1. Correctly decref old values.
2. Raise an exception if super.__init__ is caled multiple times.
3. Remove super.__init__ and add super.__new__.

What is more preferable?

[gvanrossum]

Do #1.

--Guido (mobile)
On Apr 11, 2016 11:31 PM, "Serhiy Storchaka" <report@bugs.python.org> wrote:

Serhiy Storchaka added the comment:

Possible solutions:

1. Correctly decref old values.
2. Raise an exception if super.__init__ is caled multiple times.
3. Remove super.__init__ and add super.__new__.

What is more preferable?

----------

---

Python tracker <report@bugs.python.org>
<http://bugs.python.org/issue26718\>

---

[serhiy-storchaka]

Here is a patch.