[Python-checkins] Python Regression Test Failures refleak (2)

Neal Norwitz nnorwitz at gmail.com
Mon Sep 14 10:14:39 CEST 2009


On Fri, Sep 4, 2009 at 3:35 PM, Tarek Ziadé <ziade.tarek at gmail.com> wrote:
> On Fri, Sep 4, 2009 at 11:42 PM, Neal Norwitz<nnorwitz at gmail.com> wrote:
>> More important issues:
>> ----------------------
>> test_ssl leaked [0, 0, -339] references, sum=-339
>> test_urllib2_localnet leaked [-280, 0, 0] references, sum=-280
>
> Hi,
>
> I can see test_distutils in this list from time to time.
>
> But I can't reproduce the problem on my computer to try to solve it.
>
> I thaught it was because of the tests order, but build.sh doesn't seem
> to do a randomize test execution order.

Correct.

> If anyone hase any useful hint or explanation/doc pointer to track those leaks..

Typically the problem is something like threads that are
non-deterministic.  If threads are not shutdown cleanly, the number of
references can vary during each test run.

Notice that most of the tests with inconsistent ref counts are
networking tests which generally use threads.

n


More information about the Python-checkins mailing list