addmultisigaddress does not fallback to legacy when there are uncompressed public keys in the parameters #16011

Following a discussion that I started firstly on the Bitcoin dev's mailing list, and then on StackExchange and Reddit, regarding some weird inconsistencies I noticed when I was working on a function to derive the p2sh from a multisig script, I'm here now to report what seems to be a bug in addmultisigaddress method.

Reference of the discussions: https://bitcoin.stackexchange.com/questions/87579/multisig-scripts-hashing-p2sh-difference-between-legacy-and-segwit https://www.reddit.com/r/Bitcoin/comments/bmiuk5/multisig_scripts_hashing_p2sh_difference_between/

Example:

bitcoin-cli addmultisigaddress 1 '["045897fee25bd7c5692510b2f50fcae9aa20fbc4d49d59814f4c7fdb5c4bc6eb1c0ce382458f9588e922e0d509ed8d34856787380075b00418b02e0bf7c652ef9d","02ac46c6d74d15e60f4f1035ff07ef740aca1d68d55ba0b8d336a73d7a35858831","0224a4dc5620714a9ecf67a09583d1e4c04f5bedb8ecea99028da05bb15a2a7e07"]'
{
  "address": "36ULucjWUTrDvaJzCyhFoVbDoNS6Zum2Du",
  "redeemScript": "5141045897fee25bd7c5692510b2f50fcae9aa20fbc4d49d59814f4c7fdb5c4bc6eb1c0ce382458f9588e922e0d509ed8d34856787380075b00418b02e0bf7c652ef9d2102ac46c6d74d15e60f4f1035ff07ef740aca1d68d55ba0b8d336a73d7a35858831210224a4dc5620714a9ecf67a09583d1e4c04f5bedb8ecea99028da05bb15a2a7e0753ae"
}

bitcoin-cli createmultisig 1 '["045897fee25bd7c5692510b2f50fcae9aa20fbc4d49d59814f4c7fdb5c4bc6eb1c0ce382458f9588e922e0d509ed8d34856787380075b00418b02e0bf7c652ef9d","02ac46c6d74d15e60f4f1035ff07ef740aca1d68d55ba0b8d336a73d7a35858831","0224a4dc5620714a9ecf67a09583d1e4c04f5bedb8ecea99028da05bb15a2a7e07"]'
{
  "address": "3GiimyxF1R5VixfBFAbQZbuy9EesD2r6n1",
  "redeemScript": "5141045897fee25bd7c5692510b2f50fcae9aa20fbc4d49d59814f4c7fdb5c4bc6eb1c0ce382458f9588e922e0d509ed8d34856787380075b00418b02e0bf7c652ef9d2102ac46c6d74d15e60f4f1035ff07ef740aca1d68d55ba0b8d336a73d7a35858831210224a4dc5620714a9ecf67a09583d1e4c04f5bedb8ecea99028da05bb15a2a7e0753ae"
}

Expected behaviour:

Both the two calls should return the same legacy address, because addmultisigaddress uses p2sh-segwit which is the default address type for the wallet, but segwit does not support uncompressed keys, so the first call should fallback to legacy.

Actual behaviour:

The first call is returning an invalid segwit address. @achow101 answered to my email to the dev list and also to my question on StackExchange and he believes that this is a bug and it could be related to the fact that the uncompressed pubkey is not part of my wallet.

Another interesting thing is that, as user @ugamkamat pointed out on StackExchange, changing the position of the uncompressed public key from the first to the last, the first call returns a legacy address:

./bitcoin-cli addmultisigaddress 1 '["02ac46c6d74d15e60f4f1035ff07ef740aca1d68d55ba0b8d336a73d7a35858831","0224a4dc5620714a9ecf67a09583d1e4c04f5bedb8ecea99028da05bb15a2a7e07","045897fee25bd7c5692510b2f50fcae9aa20fbc4d49d59814f4c7fdb5c4bc6eb1c0ce382458f9588e922e0d509ed8d34856787380075b00418b02e0bf7c652ef9d"]'
{
  "address": "35RZ8AHKWGGbNcRDaCUtznSVnPRWGqVPJu",
  "redeemScript": "512102ac46c6d74d15e60f4f1035ff07ef740aca1d68d55ba0b8d336a73d7a35858831210224a4dc5620714a9ecf67a09583d1e4c04f5bedb8ecea99028da05bb15a2a7e0741045897fee25bd7c5692510b2f50fcae9aa20fbc4d49d59814f4c7fdb5c4bc6eb1c0ce382458f9588e922e0d509ed8d34856787380075b00418b02e0bf7c652ef9d53ae"
}

I'd really like to try to contribute to the project, so I'm trying to figure out where the problem could be in the code. Any suggestion is welcome.

More than a bug, this seems to be an unmanaged condition. I see that in addmultisigaddress function's code there's no specific checks on the provided public keys. The output type seems to be just given by pwallet->m_default_address_type. (https://github.com/bitcoin/bitcoin/blob/master/src/wallet/rpcwallet.cpp#L1018) Maybe this part can be improved, at least to check if there are some uncompressed public keys and adapting the output type accordingly