test_asyncio doesn't give usable results when looking for refleaks:

$ ./python -m test -R 2:4 test_asyncio
[1/1] test_asyncio
beginning 6 repetitions
123456
......
test_asyncio leaked [212, -106, 265, -6360] references, sum=-5989
test_asyncio leaked [59, -29, 76, -1799] memory blocks, sum=-1693
1 test failed:
    test_asyncio

Was this always so or did it recently start? Victor has made a ton of changes.

Anyway, I imagine there may be some objects stuck in cycles and the collection may not happen until a random later time, and the tests do timing-specific stuff so the number of objects created and deleted varies per run.

Perhaps adding some well-aimed gc.collect() calls to some tearDown() methods would make this go away?

I think I'm to blame for exposing this in 4f9f7e0fe1fd. I have a theory on why that exposed it; I think regrtest is holding an extra reference to the TestSuite in runtest_inner since it is using a different branch now that test_asyncio doesn't have a test_main function.

It may be related to the issue bpo-17911.

It may be related to the issue bpo-17911.

I checked: it is. The strange reference count can be seen with a single test. Example:

$ ./python -m test -R 3:3: -m test_default_exc_handler_coro test_asyncio 
[1/1] test_asyncio
beginning 6 repetitions
123456
......
test_asyncio leaked [53, 53, -106] references, sum=0
test_asyncio leaked [15, 15, -30] memory blocks, sum=0
1 test failed:
    test_asyncio

This test uses a coroutine which raises an exception. The exception is stored in a Task object. But the exception contains also a traceback which indirectly creates a reference cycle. For example, the zero_error_coro() coroutine of the test uses the free variable "self".

It's very difficult to find all objects of a reference cycle. We can try to break some cycles, it's already done Task._step() which sets self to None, but it's a waste of time. IMO the correct fix is to not store frame objects in an exception: see the issue bpo-17911.

Le 30/07/2014 06:08, STINNER Victor a écrit :

This test uses a coroutine which raises an exception. The exception
is
stored in a Task object. But the exception contains also a traceback
which indirectly creates a reference cycle.

regrtest calls gc.collect().

I checked on my theory, and removing the extra reference to 'tests' from the runtest_inner scope fixes it for me:

$ python -m test -R 3:3: test_asyncio
Running Debug|Win32 interpreter...
[1/1] test_asyncio
beginning 6 repetitions
123456
......
1 test OK.

Here's the patch.

New changeset 9bca86812857 by Zachary Ware in branch '3.4':
Issue bpo-22104: Don't hold a reference to the loaded TestSuite in runtest_inner
http://hg.python.org/cpython/rev/9bca86812857

New changeset 7bc53cf8b2df by Zachary Ware in branch 'default':
Closes bpo-22104: Merge with 3.4
http://hg.python.org/cpython/rev/7bc53cf8b2df