Multisig scripts hashing (p2sh): difference between legacy and segwit

Question

Some time ago I noticed that, given the same inputs, addmultisigaddress and createmultisig Bitcoin core's methods were generating different p2sh (addresses).

I asked around and a user from the bitcoin dev list told me that this was because:

"addmultisigaddress uses the default address type of the wallet, which is p2sh-segwit and createmultisig uses a default address type of legacy"

An example of this behaviour is:

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

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

The point is that, later, I noticed that with certain public keys the two methods generate the same hash:

bitcoin-cli addmultisigaddress 1 '["03fc5e16d0ece343a94735ca467d5812922fcc30e3ce43ceaf3cff7d7617631146","02ea92a0cd1738cef7502e42fe119a322845f8f1a2dd9b7216635e17dd2fffd101","04633794a75bfbd9fb2bc6bd54836831c0916dc27f9ac528045cc9352bb2cc97a003b3a6ae397101d801f3d95e6a153368b939aaf7b27bd3a5bb8a30ed92aac204"]'
{
  "address": "32AVnYUUSvm4bGgEXoQ2zuFAJBPH4P31Gi",
  "redeemScript": "512103fc5e16d0ece343a94735ca467d5812922fcc30e3ce43ceaf3cff7d76176311462102ea92a0cd1738cef7502e42fe119a322845f8f1a2dd9b7216635e17dd2fffd1014104633794a75bfbd9fb2bc6bd54836831c0916dc27f9ac528045cc9352bb2cc97a003b3a6ae397101d801f3d95e6a153368b939aaf7b27bd3a5bb8a30ed92aac20453ae"
}

bitcoin-cli createmultisig 1 '["03fc5e16d0ece343a94735ca467d5812922fcc30e3ce43ceaf3cff7d7617631146","02ea92a0cd1738cef7502e42fe119a322845f8f1a2dd9b7216635e17dd2fffd101","04633794a75bfbd9fb2bc6bd54836831c0916dc27f9ac528045cc9352bb2cc97a003b3a6ae397101d801f3d95e6a153368b939aaf7b27bd3a5bb8a30ed92aac204"]'
{
  "address": "32AVnYUUSvm4bGgEXoQ2zuFAJBPH4P31Gi",
  "redeemScript": "512103fc5e16d0ece343a94735ca467d5812922fcc30e3ce43ceaf3cff7d76176311462102ea92a0cd1738cef7502e42fe119a322845f8f1a2dd9b7216635e17dd2fffd1014104633794a75bfbd9fb2bc6bd54836831c0916dc27f9ac528045cc9352bb2cc97a003b3a6ae397101d801f3d95e6a153368b939aaf7b27bd3a5bb8a30ed92aac20453ae"
}

I don't get how this can be possible if the inputs and the redeem scripts are the same.

What's the difference between the two cases and in general between a legacy multisig transaction and a segwit one?

Reading raw transactions data from the blockchain, how is possible to know whether a script needs to be hashed using legacy or segwit?

Answer 1

I believe the behavior you are seeing is actually a bug.

By default, addmultisigaddress uses p2sh-segwit which is the default address type for the wallet. By default createmultisig uses legacy. These are two different address types, and normally, you get two different addresses.

However segwit does not allow uncompressed public keys. So when the redeemScript is being created, if an uncompressed public key is detected, both RPCs will return legacy addresses. This is why you see the same address in the second example.

But I believe there is a bug. Your first example should also give the same legacy addresses, but it does not. I think that this is because the uncompressed pubkey is not part of the wallet, so it does not correctly detect that it is an uncompressed pubkey. Thus addmultisigaddress follows its default to give you the p2sh-segwit address and createmultisig follows its default to give you the legacy address.

Reading raw transactions data from the blockchain, how is possible to know whether a script needs to be hashed using legacy or segwit?

For p2sh-segwit, the redeemScript is not actually the redeemScript you get from the RPCs. Rather it is the script for a P2WSH output. The "redeemScript" is actually the witnessScript for the P2WSH script that is the actual redeemScript. The script type (either in the scriptPubKey or in the redeemScript) determine whether that output is segwit or not.