slice() leaks memory when part of a cycle

[KevinModzelewski]

BPO	26659
Nosy	@birkenfeld, @pitrou, @vstinner, @benjaminp, @1st1, @serprex

^{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 2016-04-16.21:54:40.705>
created_at = <Date 2016-03-28.20:07:30.598>
labels = ['interpreter-core', 'performance']
title = 'slice() leaks memory when part of a cycle'
updated_at = <Date 2016-04-16.22:52:29.780>
user = 'https://bugs.python.org/KevinModzelewski'

bugs.python.org fields:

activity = <Date 2016-04-16.22:52:29.780>
actor = 'martin.panter'
assignee = 'none'
closed = True
closed_date = <Date 2016-04-16.21:54:40.705>
closer = 'python-dev'
components = ['Interpreter Core']
creation = <Date 2016-03-28.20:07:30.598>
creator = 'Kevin Modzelewski'
dependencies = []
files = []
hgrepos = []
issue_num = 26659
keywords = []
message_count = 3.0
messages = ['262582', '262589', '263590']
nosy_count = 8.0
nosy_names = ['georg.brandl', 'pitrou', 'vstinner', 'benjamin.peterson', 'python-dev', 'yselivanov', 'Kevin Modzelewski', 'Demur Rumed']
pr_nums = []
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'resource usage'
url = 'https://bugs.python.org/issue26659'
versions = ['Python 2.7', 'Python 3.5', 'Python 3.6']

[KevinModzelewski]

The slice type doesn't participate in GC, which means that if you happen to create a cycle involving a slice, that cycle will never get freed. Here's an example:

def f():
    l = []
    l.append(slice(l))
# Will consume memory without bound:
while True:
    f()

This seems pretty hard to trigger accidentally, so it might not be a huge deal -- especially since it seems to have been around for a while. (I only checked 2.7 and trunk though.)

I think this could be solved by either having the slice class participate in GC (ie add tp_traverse and tp_clear methods), or maybe doing some type-filtering during slice_new().

[serprex]

Implementing tp_traverse & tp_clear seems to runs into complications due to slice_cache

[python-dev]

New changeset 879da9400529 by Benjamin Peterson in branch '2.7':
add gc support to slice (closes bpo-26659)
https://hg.python.org/cpython/rev/879da9400529

New changeset 9e2176d18965 by Benjamin Peterson in branch '3.5':
add gc support to slice (closes bpo-26659)
https://hg.python.org/cpython/rev/9e2176d18965

New changeset 870fcc50f1bd by Benjamin Peterson in branch 'default':
merge 3.5 (bpo-26659)
https://hg.python.org/cpython/rev/870fcc50f1bd