I' ve checked the memory leak with these scripts and no leaks found

./python.exe -m test test_dbm_gnu -R 3:3

def test_gdbm(self):
    code = textwrap.dedent(r"""
        import glob

        import test.support
        dbm = test.support.import_module('_gdbm')
        # or dbm = test.support.import_module('dbm')
        _fname = test.support.TESTFN

        def delete_files():
            for f in glob.glob(_fname + "*"):
                test.support.unlink(f)

        f = dbm.open(_fname, 'n')
        f[b'g'] = b"indented"
        f.close()
        delete_files()
    """)
    ret = test.support.run_in_subinterp(code)
    self.assertEqual(ret, 0)

Note that dbm = test.support.import_module('dbm') is leaked at the master branch :)

Note that dbm = test.support.import_module('dbm') is leaked at the master branch :)

Maybe open an issue to track this bug.

Maybe open an issue to track this bug.

Well, I added a comment to: https://bugs.python.org/issue40987#msg371679

@vstinner

Maybe open an issue to track this bug

Oh I mean the subinterpreter test with dbm = test.support.import_module('dbm')is leaked
With this change, there is no leak.

@vstinner Thanks for the review!

With this change, there is no leak.

Oh ok, that's great! Sorry, I misunderstood your comment.