encodePacked appears to have different behavior from solidity for complex types

[charles-cooper]

For arrays, encodePacked appears to fall back to standard encoding.

eth-abi/eth_abi/encoding.py

Line 666 in decaadc

encoded_elements = self.encode_elements(value)

->

eth-abi/eth_abi/encoding.py

Lines 615 to 631 in decaadc

	def encode_elements(self, value):
	self.validate_value(value)
	item_encoder = self.item_encoder
	tail_chunks = tuple(item_encoder(i) for i in value)
	items_are_dynamic = getattr(item_encoder, 'is_dynamic', False)
	if not items_are_dynamic:
	return b''.join(tail_chunks)
	head_length = 32 * len(value)
	tail_offsets = (0,) + tuple(accumulate(map(len, tail_chunks[:-1])))
	head_chunks = tuple(
	encode_uint_256(head_length + offset)
	for offset in tail_offsets
	)
	return b''.join(head_chunks + tail_chunks)

However, solidity skips the length and pads dynamic array elements with zeros:
https://docs.soliditylang.org/en/v0.8.6/abi-spec.html#non-standard-packed-mode

The encoding of an array is the concatenation of the encoding of its elements with padding.

cf. ethereum/solidity#8441 for an example of the current padding scheme.

Additionally, and please correct me if I'm wrong here, eth-abi appears to encode bytestrings in tuples without padding as in the tests here

eth-abi/tests/common/unit.py

Lines 166 to 171 in decaadc

	decode_hex(
	'646176696420617474656e626f726f756768' # encoding of b'david attenborough'
	'00' # encoding for `False`
	'626f617479206d63626f617466616365' # encoding of b'boaty mcboatface'
	'01' # encoding for `True`
	),

However my cursory reading of solc output says that they are in fact padded. (solidity disallows structs in abi.encodePacked, but it uses the packed encoding code path to calculate event indexes).
event.ir.txt
event.sol.txt

[kclowes]

Thanks for the report @charles-cooper. On first glance, this does look like a bug. It will probably be a few weeks before I can get to really digging in though. If you want to open a PR, that will be the fastest way to get it fixed, otherwise I can take a deeper look in a few weeks.

[charles-cooper]

I'm wondering if it's worth fixing. ethereum/solidity#11593 says encodePacked is going to be removed, it's unclear to me what the effect will be on event indexes though.