OverflowError when decode an event data with empty fields

[alebaffa]

OverflowError when decode an event data with empty fields

[reedsa]

@alebaffa In your examples, the data includes offsets which declare the position of the values when decoding the bytes values. In the second case, your positions are finding the values relative to the beginning of the data. When you add the address topics, the data positions change and thus must be reflected in the bytes headers.

I've added a few tests to illustrate this in https://github.com/ethereum/eth-abi/pull/228/files

[alebaffa]

oh I see thank you guys!

[alebaffa]

I keep testing and I found that, from the test above, if I keep only the Data fields (i.e. amount, data, operatorData):

 ["uint256", "bytes", "bytes"],
            (
                # Data
                "0000000000000000000000000000000000000000000004a8f8176590b4e9237f"
                "000000000000000000000000000000000000000000000000000000000000006"
                "0000000000000000000000000000000000000000000000000000000000000008"
                "000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000"
            ),

the decode correctly returns empty bytes where the fields are empty:

(22006395804779840873343, b'', b'')

In this case, the value returns an empty byte as expected:

class HeadTailDecoder(BaseDecoder):
....

    def decode(self, stream):
        start_pos = decode_uint_256(stream)

        stream.push_frame(start_pos)
        value = self.tail_decoder(stream) # <--- correctly returns an empty bytes. While in the previous test - with all the other fields too - this fails.
        stream.pop_frame()

        return value

So the problem seems to be in ByteStringDecoder:

    def read_data_from_stream(self, stream):
        data_length = decode_uint_256(stream)
        padded_length = ceil32(data_length)

        data = stream.read(padded_length) # <--- this fails in the previous test, but it is OK now.

specifically, in the previous failing test the data_length and padded_length were different by one digit:

data_length: 22006395804779840873343 # <--- this is actually the value of the`amount` field in the data
data_length: 22006395804779840873344

while in this second successful test they are both 0 and everything works.

So it seems that decode cannot handle properly topics + data when some of the data fields are empty bytes.

[fselmo]

@reedsa good summary. The topics and the log data are separate parts of the transaction receipt log and so the data field contains position information if you were to decode it separately, not along with the topics. To illustrate this exact transaction using web3.py:

>>> from web3 import Web3, HTTPProvider
>>> from eth_abi import abi

>>> w3 = Web3(HTTPProvider(provider_url))

>>> tx_receipt = w3.eth.get_transaction_receipt("0x128bae16288948bdeead21cbd662bdebae0c25ebb0817741e4577ee83b6339a2")
>>> log_in_question = tx_receipt["logs"][2]

>>> log_in_question["topics"]  # returns topics [0, 1: operator, 2: from, 3: to]
[HexBytes('0x06b541ddaa720db2b10a4d0cdac39b8d360425fc073085fac19bc82614677987'),
 HexBytes('0x000000000000000000000000000000000022d473030f116ddee9f6b43ac78ba3'),
 HexBytes('0x00000000000000000000000050650a86840e5c4c46641a9435418b7458458747'),
 HexBytes('0x00000000000000000000000098409d8ca9629fbe01ab1b914ebf304175e384c8')]

>>> abi.decode(['uint256', 'bytes', 'bytes'], log_in_question["data"])
(22006395804779840873343, b'', b'')