Memory leaks and Python 2.2c1 (fwd)

Ben Escoto
Mon, 17 Dec 2001 14:36:24 -0800

Content-Type: text/plain; charset=us-ascii

Ok, I've had a talk with some helpful Python developers about memory
leaks.  They say that it shouldn't matter that, for instance, one of
us is running Suse 7.3 and the other Redhat 7.2.  If we are both
running exactly the same python on exactly the same data, both of us
will see a memory leak or neither of us will.

    So, I guess to fix this problem (which may be in python, or may be
in my script) I need to replicate it on my side.  Then I can whittle
the code away until I get something that I can submit to python's bug
tracker.  If you are experiencing a memory leak, try running this:

md foo
dd if=/dev/zero of=foo/zeroes bs=16k count=8192
rdiff-backup foo foo2

and checking the memory with top.  (The dd command just creates a
128MB file full of zeroes - something I can recreate on my side and
guarantee the same data.  (Actually the data won't exactly be the
same, because rdiff-backup will look at the file creation times of
foo, but hopefully this won't make a difference as far as the leak is

    If there is a leak with the above dataset, tell me exactly which
version of python you are running.  So if it is 2.2c1, tell me where
you got it.  If you are running debian, perhaps you could install the
official 2.2c1 tarball, as I might have trouble installing a .deb on
my RH system.

    Memory leaks are really annoying (especially in garbage collected
languages?) but fortunately most of the work will be on my end (or by
the python developers).  I just need to be able to see the problem on
my end.

Ben Escoto

Content-Type: application/pgp-signature

Version: GnuPG v1.0.6 (GNU/Linux)
Comment: Exmh version 2.5 01/15/2001