Apparent eval() leak for python 2.3.5

[John Marshall]

Hi,

I am reposting this message from python-dev.

Could someone please test the code below
to verify that there is indeed a problem
with eval() under python 2.3.5. I have
rebuilt python 2.3.5 under the latest debian
and under RH 7.3 (in case the problem is
in system libraries).

The following code causes the virtual memory
allocation to constantly increase until there
is no more memory left (the VIRT column of the
'top' utility display).

-----
#! /usr/bin/env python
#
# evalleak.py

def builddict():
d = {}
for i in xrange(10000):
d[i] = { "int": 0, "str": "nada", "dict": {} }
return d

if __name__ == "__main__":
d = builddict()
st = str(d)
print "len(st): %d" % len(st)
while 1:
d = eval(st)
-----

Thanks,
John

[Benjamin Niemann]

John Marshall wrote:

Hi,

I am reposting this message from python-dev.

Could someone please test the code below
to verify that there is indeed a problem
with eval() under python 2.3.5. I have
rebuilt python 2.3.5 under the latest debian
and under RH 7.3 (in case the problem is
in system libraries).

The following code causes the virtual memory
allocation to constantly increase until there
is no more memory left (the VIRT column of the
'top' utility display).

I have this leak, too in python2.3 under debian/sid:
Python 2.3.5 (#2, Nov 20 2005, 16:40:39)
[GCC 4.0.3 20051111 (prerelease) (Debian 4.0.2-4)] on linux2

python2.4 seems to be unaffected.
Python 2.4.2 (#2, Nov 20 2005, 17:04:48)
[GCC 4.0.3 20051111 (prerelease) (Debian 4.0.2-4)] on linux2

--
Benjamin Niemann
Email: pink at odahoda dot de
WWW: http://www.odahoda.de/