[Python-Dev] test_minidom non-failure failure? (take 2)

[Jeremy Hylton]

>>>>> "NS" == Neil Schemenauer <nas@arctrix.com> writes:

  NS> The -l option tries to find any cyclic garbage produced by the
  NS> tests.  I don't think that that option should be enabled
  NS> default.  The output means that the GC is working and is finding
  NS> stuff that would not be freed by reference counting alone.

  NS> I can't tell if the GC would free this garbage.  The -l option
  NS> sets the DEBUG_SAVEALL option which causes all garbage found to
  NS> end up in gc.garbage, not just garbage the can't be cleaned up.

  NS> I don't have pyexpat installed here so I can't test it.  If you
  NS> want to find out if test_minidom is creating garbage the
  NS> collector can't free you should comment out the:

  NS>     gc.set_debug(gc.DEBUG_SAVEALL)

  NS> line in regrtest.py and run:

  NS>     regrtest.py -l test_minidom

  NS> If that does what I think it does and you still get the
  NS> "garbage: " line then the test is creating evil things.  :)

The test is not creating evil things.  I commented out the
DEBUG_SAVEALL line and got no error report.

The question, then, is what to do about the -l option.  I assume we
should remove the -l option from the Makefile, so that "make test"
doesn't turn on DEBUG_SAVEALL.  Or do we need to change regrtest in
some way so that it still reports on tests that create evil things?

Jeremy