Investigate and resolve discrepancies with solidity's abi.encode() #158
Conversation
fselmo
force-pushed
the
eth-abi-vs-solidity-abi-discrepancies
branch
6 times, most recently
from
212f4c8 to
072486a
Compare
fselmo
changed the title
fselmo
force-pushed
the
eth-abi-vs-solidity-abi-discrepancies
branch
from
072486a to
85fc4f6
Compare
fselmo
force-pushed
the
eth-abi-vs-solidity-abi-discrepancies
branch
from
85fc4f6 to
b4069f5
Compare
- Empty values for dynamic encoders had extra 32-byte zero padding when compared to Solidity's abi.encode() method. This led to correct instructions for the EVM but also required more gas than necessary and ultimately led to differing hashes, etc. - Add more tests. Decode tuple test values not just as separate list of types but as a tuple type with nested values for encode_abi() tests for tuples. - Eliminate padding for empty dynamic arrays as well. If an array is static or if it is dynamic and empty do not pad with 32-byte zero padding.
fselmo
force-pushed
the
eth-abi-vs-solidity-abi-discrepancies
branch
from
b4069f5 to
75ead60
Compare
|
Thx! Any idea when this gets officially released (and even used by web3py)? |
|
We'll probably release |
|
No chance of a patch release for web3py 5.28.*? Really struggling issue right now... |
|
I'm hesitant to pull it into web3py 5.28 because this is technically a breaking change (though arguably it's a bugfix which is a little less black and white). Most of the breaking changes in web3 v6 right now are dependency-related, so I hope it won't be too bad to update. The other big breaking change is dropping support for python 3.6. Any chance you can try updating to web3 v6 and let me know how it goes? If you run into issues, let me know and I'll consider releasing a patch. |
|
Hi, ran into this again on |
What was wrong?
Related issues:
bytes#151encode_abi()does not yield the same result as Solidity'sabi.encode()#157How was it fixed?
It seems that, for
encode_abi(), extra 32-byte zero padding was happening in a few places, compared to what Solidity'sabi.encode()returns. This seems to be the case when dealing with empty dynamic types, including empty dynamic arrays.i.e.
encode_abi(['string'], [''])vs Solidityabi.encode('')encode_abi(['string[]'], [[]])vs Soliditystring[] sl; abi.encode(sl);These discrepancies still yielded correct instructions for the EVM but ultimately led to different hashes and higher gas values for transactions encoded this way.
To-Do
Cute Animal Picture