JMeter结果大小限制(JMeter results size limit)

 问题在于你在测试用例中调用真正的boto，而不是你的模拟。 
 
 您需要使用它在测试模块中具有的相同符号名称来模拟对象。 而不是试图修补'boto.s3.connection.S3Connection'，你想补丁'test_s3.S3Connection'。 它以'test_s3'开头，因为这是您要加载测试代码的模块的名称。 然后，在点之后，使用与该模块内部相同的符号可以访问S3Connection（即，没有boto.s3.connection）。 
 
 另外，补丁为你自动创建一个模拟，所以你不需要创建第一个模拟对象。 由于您已经修补了S3Connection，而不是修补其中的内容，您只需将创建的mock作为正确方法的返回值插入即可： 
 
    mock_bucket = mock.Mock()
    mock_key = mock.Mock()

    with mock.patch('test_s3.S3Connection') as mock_conn:
        mock_conn.get_bucket.return_value = mock_bucket
        mock_bucket.lookup.return_value = mock_key

        mock_key.size.return_value = 50000
        self.assertEquals(fake.check_size(), 50)

The problem is that you are calling the real boto in your test case, not your mock.
 
You need to mock the object using the same symbol name it has inside your test module. Instead of trying to patch 'boto.s3.connection.S3Connection', you want to patch 'test_s3.S3Connection'. It starts with 'test_s3' because that is the name of the module from where you are loading the code under test. Then, after the dot, use the same symbol that inside that module gives you access to S3Connection (that is, without boto.s3.connection).
 
Also, patch creates a mock automatically for you, so you do not need to create the first mock object. And as you already patched S3Connection, instead of patching things inside it, you just need to plug your created mocks as return values for the proper methods:
 
    mock_bucket = mock.Mock()
    mock_key = mock.Mock()

    with mock.patch('test_s3.S3Connection') as mock_conn:
        mock_conn.get_bucket.return_value = mock_bucket
        mock_bucket.lookup.return_value = mock_key

        mock_key.size.return_value = 50000
        self.assertEquals(fake.check_size(), 50)