[Python-checkins] Python Regression Test Failures refleak (1)

Thomas Heller theller at python.net
Sun Apr 16 22:22:28 CEST 2006


Neal Norwitz wrote:
> On 4/13/06, Tim Peters <tim.peters at gmail.com> wrote:
>> [Thomas Heller]
>>> Neal, does this mean that test_ctypes does not leak anymore, or is it
>>> somehow excluded from this list?
>> I'm guessing, but I believe Neal uses Misc/build.sh for this, and
>> ctypes is now in its LEAKY_TESTS pattern of test names to skip:
>>
>> LEAKY_TESTS="test_(cmd_line|ctypes|filecmp|socket|threadedtempfile|threading|threading_local|urllib2)"
> 
> Tim's right, the leak report comes from Misc/build.sh (same with
> building the docs and other emails sent to python-checkins).  ctypes
> leaks inconsistently which is why it's in the LEAKY_TESTS.  In this
> case it didn't show any leaks.  I don't know if the leaks will return.
>  When the leaks aren't consistent and report the same number, it
> usually means there are references elsewhere that may or may not be
> cleaned up.
> 
> However, ctypes still leaks when reloaded.  We probably ought to have
> a test for each module that just does reload(module) and see if that
> leaks.  I wouldn't be surprised if several C modules leak.
> 
I've added a tracker item for that.
https://sourceforge.net/tracker/index.php?func=detail&aid=1471423&group_id=5470&atid=105470

Thomas



More information about the Python-checkins mailing list