Contract Function Signatures: Look at all blocks

[kunalmodi]

Issue

Example contract https://etherscan.io/address/0x1c6178ae715ce8e91812779c6486ee53243abd41 is not marked as ERC20, even though it's function signatures should match:

SELECT *
FROM `bigquery-public-data.crypto_ethereum.contracts`
WHERE address = '0x1c6178ae715ce8e91812779c6486ee53243abd41'

The bytecode, however, does look like it has all of the correct signatures: https://etherscan.io/bytecode-decompiler?a=0x1c6178ae715ce8e91812779c6486ee53243abd41

I wasn't sure why we were only looking at the first block of instructions, but in the cases I tested, it seems like there can be file signatures in subsequent blocks. Might be missing something though!

Testing

echo "0x1c6178ae715ce8e91812779c6486ee53243abd41" >> contracts_addresses_test.csv
ethereumetl export_contracts --contract-addresses contracts_addresses_test.csv -o contracts.csv -w 3 -p file:///mnt/disks/ssds/geth/geth.ipc
cat contracts.csv

0x1c6178ae715ce8e91812779c6486ee53243abd41,<removed for brevity>,"0x061c82d0,0x06fdde03,0x095ea7b3,0x13114a9d,0x15ab88c9,0x178ef307,0x18160ddd,0x20b9588c,0x23b872dd,0x2cde6081,0x2d838119,0x313ce567,0x3685d419,0x39509351,0x3b124fe7,0x3bd5d173,0x437823ec,0x4549b039,0x4a74bb02,0x4dfefc4b,0x52390c02,0x5342acb4,0x6b455243,0x6bc87c3a,0x70a08231,0x715018a6,0x76e2b7ab,0x77e5006f,0x791ac947,0x7d1db4a5,0x8187f516,0x862a4bf2,0x88f82020,0x8da5cb5b,0x8ee88c53,0x95d89b41,0xa457c2d7,0xa4d66daf,0xa69df4b5,0xa9059cbb,0xad5c4648,0xb6c52324,0xb8c9d25c,0xc0d78655,0xc21ebd07,0xc49b9a80,0xd543dbeb,0xdd467064,0xdd62ed3e,0xea2f0b37,0xf2fde38b,0xf305d719,0xffffffff",True,False,

(of note: ERC20 column is True now, and we get a lot more function signatures than before)

[rahulmadduluri]

I've found many contracts that are missing from the database as well (eg Lazy Lions https://etherscan.io/address/0x8943C7bAC1914C9A7ABa750Bf2B6B09Fd21037E0). Is there any blocker to merging in this change? Thanks in advance! @evgeniuz @medvedev1088

[rahulmadduluri]

@evgeniuz @medvedev1088 gentle bump

[medvedev1088]

Thanks for the PR!

Looking only at the 1st block could produce false negatives (some function sighashes will be missing), while looking at all blocks could product false positives (some function sighashes will be added mistakenly). The PR needs some test cases and comparison of results before and after to understand the impact

I kindly request to add test cases that could surface false positives.