Garbage Collection Control

When having problems that seem to be caused by memory-management errors, it can be helpful to adjust Python’s cyclic garbage collector or to get garbage colection statistics. The --gc option can be used for this purpose.

If you think you are getting a test failure due to a garbage collection problem, you can try disabling garbage collection by using the --gc option with a value of zero.

>>> import os.path, sys
>>> directory_with_tests = os.path.join(this_directory, 'testrunner-ex')
>>> defaults = ['--path', directory_with_tests]
>>> from zope import testrunner
>>> sys.argv = 'test --tests-pattern ^gc0$ --gc 0 -vv'.split()
>>> _ = testrunner.run_internal(defaults)
Cyclic garbage collection is disabled.
Running tests at level 1
Running zope.testrunner.layer.UnitTests tests:
  Set up zope.testrunner.layer.UnitTests in N.NNN seconds.
  Running:
 make_sure_gc_is_disabled (gc0)
  Ran 1 tests with 0 failures, 0 errors and 0 skipped in N.NNN seconds.
Tearing down left over layers:
  Tear down zope.testrunner.layer.UnitTests in N.NNN seconds.

Alternatively, if you think you are having a garbage collection related problem, you can cause garbage collection to happen more often by providing a low threshold:

>>> sys.argv = 'test --tests-pattern ^gc1$ --gc 1 -vv'.split()
>>> _ = testrunner.run_internal(defaults)
Cyclic garbage collection threshold set to: (1,)
Running tests at level 1
Running zope.testrunner.layer.UnitTests tests:
  Set up zope.testrunner.layer.UnitTests in N.NNN seconds.
  Running:
 make_sure_gc_threshold_is_one (gc1)
  Ran 1 tests with 0 failures, 0 errors and 0 skipped in N.NNN seconds.
Tearing down left over layers:
  Tear down zope.testrunner.layer.UnitTests in N.NNN seconds.

You can specify up to 3 --gc options to set each of the 3 gc threshold values:

>>> sys.argv = ('test --tests-pattern ^gcset$ --gc 701 --gc 11 --gc 9 -vv'
...             .split())
>>> _ = testrunner.run_internal(defaults)
Cyclic garbage collection threshold set to: (701, 11, 9)
Running tests at level 1
Running zope.testrunner.layer.UnitTests tests:
  Set up zope.testrunner.layer.UnitTests in N.NNN seconds.
  Running:
 make_sure_gc_threshold_is_701_11_9 (gcset)
  Ran 1 tests with 0 failures, 0 errors and 0 skipped in N.NNN seconds.
Tearing down left over layers:
  Tear down zope.testrunner.layer.UnitTests in N.NNN seconds.

Specifying more than 3 --gc options is not allowed:

>>> from io import StringIO
>>> out = StringIO()
>>> stdout = sys.stdout
>>> sys.stdout = out
>>> sys.argv = ('test --tests-pattern ^gcset$ --gc 701 --gc 42 --gc 11 --gc 9 -vv'
...             .split())
>>> _ = testrunner.run_internal(defaults)
Traceback (most recent call last):
...
SystemExit: 1
>>> sys.stdout = stdout
>>> print(out.getvalue())
Too many --gc options

Garbage Collection Statistics

You can enable gc debugging statistics using the --gc-options (-G) option. You should provide names of one or more of the flags described in the library documentation for the gc module.

The output statistics are written to standard error.

>>> from io import StringIO
>>> err = StringIO()
>>> stderr = sys.stderr
>>> sys.stderr = err
>>> sys.argv = ('test --tests-pattern ^gcstats$ -G DEBUG_STATS'
...             ' -G DEBUG_COLLECTABLE -vv'
...             .split())
>>> _ = testrunner.run_internal(defaults)
Running tests at level 1
Running zope.testrunner.layer.UnitTests tests:
  Set up zope.testrunner.layer.UnitTests in N.NNN seconds.
  Running:
 generate_some_gc_statistics (gcstats)
  Ran 1 tests with 0 failures, 0 errors and 0 skipped in N.NNN seconds.
Tearing down left over layers:
  Tear down zope.testrunner.layer.UnitTests in N.NNN seconds.
>>> sys.stderr = stderr
>>> print(err.getvalue())        
gc: collecting generation ...