In v0.17.1, importing a p2wsh-in-p2sh multisig address as watchonly+solvable requires importing it as two separate addresses, once with each of the two scripts. See my question on StackExchange.

In the current bitcoin master (d8794a78a887a920276c7124f1c46d69592c6c4e), this breaks, I suspect due to #14454. While that PR is a nice improvement, isn't it bad to break backward compatibility?

Problem

I want to import a p2wsh-in-p2sh multisig address as watchonly+solvable. I run:

bitcoin-cli -testnet importmulti '[{"redeemscript": "5221035a0cf2b8ad46945154d80b339f730ac0cdbc39a95550a95821adf6df6e6e3c9421038f339e9149fda8496360d689b5d6b4d66f8e64e28b1c89846efd0831512eab882103442945263f31819baf5799dc9595eba49b8f6674dadf21189f717abd630ab15053ae", "scriptPubKey": {"address": "2MvQKEWTdtH7uM5C72quuMukWNFkFYjm34N"}, "timestamp": "now"}, {"redeemscript": "0020d017965b0cae52a8b6f6e43b00118be27dae84b825959dcb77ca6b5310e53ea3", "scriptPubKey": {"address": "2MvQKEWTdtH7uM5C72quuMukWNFkFYjm34N"}, "timestamp": "now"}]'

The result comes back as:

[
  {
    "success": true,
    "warnings": [
      "Importing as non-solvable: redeemScript does not match the scriptPubKey. If this is intentional, don't provide any keys, pubkeys, witnessscript, or redeemscript."
    ]
  },
  {
    "success": true,
    "warnings": [
      "Importing as non-solvable: missing witnessscript. If this is intentional, don't provide any keys, pubkeys, witnessscript, or redeemscript."
    ]
  }
]

And then getaddressinfo shows that it is not solvable:

{
  "address": "2MvQKEWTdtH7uM5C72quuMukWNFkFYjm34N",
  "scriptPubKey": "a91422a07fe0ea8b8293eb336b9423f7e3958917924387",
  "ismine": false,
  "solvable": false,
  "iswatchonly": true,
  "isscript": true,
  "iswitness": false,
  "label": "",
  "ischange": false,
  "timestamp": 1296688602,
  "labels": [
    {
      "name": "",
      "purpose": "receive"
    }
  ]
}

Firstly, is it okay for the importmulti to report success:true when some of the provided information was ignored? That seems...not good.

Secondly, how can I import this address in a way that will work on both v0.17.1 and (future) v0.18?

v0.17.1 behavior

Address is imported successfully:

[
  {
    "success": true
  },
  {
    "success": true
  }
]

And the getaddressinfo shows everything I expect:

{
  "address": "2MvQKEWTdtH7uM5C72quuMukWNFkFYjm34N",
  "scriptPubKey": "a91422a07fe0ea8b8293eb336b9423f7e3958917924387",
  "ismine": false,
  "iswatchonly": true,
  "isscript": true,
  "iswitness": false,
  "script": "witness_v0_scripthash",
  "hex": "0020d017965b0cae52a8b6f6e43b00118be27dae84b825959dcb77ca6b5310e53ea3",
  "embedded": {
    "isscript": true,
    "iswitness": true,
    "witness_version": 0,
    "witness_program": "d017965b0cae52a8b6f6e43b00118be27dae84b825959dcb77ca6b5310e53ea3",
    "script": "multisig",
    "hex": "5221035a0cf2b8ad46945154d80b339f730ac0cdbc39a95550a95821adf6df6e6e3c9421038f339e9149fda8496360d689b5d6b4d66f8e64e28b1c89846efd0831512eab882103442945263f31819baf5799dc9595eba49b8f6674dadf21189f717abd630ab15053ae",
    "sigsrequired": 2,
    "pubkeys": [
      "035a0cf2b8ad46945154d80b339f730ac0cdbc39a95550a95821adf6df6e6e3c94",
      "038f339e9149fda8496360d689b5d6b4d66f8e64e28b1c89846efd0831512eab88",
      "03442945263f31819baf5799dc9595eba49b8f6674dadf21189f717abd630ab150"
    ],
    "address": "tb1q6qtevkcv4ef23dhkusasqyvtuf76ap9cyk2emjmhef44xy89863s6yyfks",
    "scriptPubKey": "0020d017965b0cae52a8b6f6e43b00118be27dae84b825959dcb77ca6b5310e53ea3"
  },
  "label": "",
  "timestamp": 1296688602,
  "labels": [
    {
      "name": "",
      "purpose": "receive"
    }
  ]
}

@MeshCollider

While we do try to keep backwards compatibility, what you were able to do in 0.17.1 was (IMO) a bug so we shouldn't keep compatibility with it.

Is there any other way to import this into v0.17.1?

Is there any other way to import this into v0.17.1?

There is not, but that is also a bug.

OK...I guess I will wait for v0.18.

There is still the question of importmulti reporting success:true despite ignoring some of the information provided. Is that acceptable?

The reason for making a number of things acceptable with a warning instead of failing was to improve backward compatibility.

I'm actually a bit confused why just providing a redeemScript for multisig used to be enough. But it doesn't seem unreasonable to make sure it keeps working in 0.18, depending on the complexity of doing so.

TL:DR: I think breaking backward compatibility is a bug, and to the extent possible it should keep working, but with a warning.

Providing only the redeemScript was insufficient; I had to import the same address twice, once with the redeemScript and once with the witnessScript (in the redeemScript field). This ugly double-importmulti was similar to #12418 which was fixed by allowing the multisig script to sign either p2sh or p2wsh-in-p2sh.

Perhaps I should have opened a new issue at the time, instead of assuming this double-import was the only/best way.

In my specific case, this double-importmulti code is not yet deployed, so it's not a big deal to me if it breaks. I'll just hold off on deploying it until I can redo it in v0.18.

Another concern here is that addmultisigaddress with p2sh-segwit does not give me the redeemScript, only the witnessScript. (Or do I have that backwards? Anyway it's one or the other.)

So to use that address with importmulti I first have to run decodescript to get the matching redeemScript, then pass that back in to importmulti. That feels rather uncoordinated.

Do you think it would make sense to have importmulti work similar to signrawtransactionwithkeys, where I can sign a p2sh-segwit by providing only the witnessScript? (See #12418) That would also cleanly solve this backward compatibility issue.

@bitcoinhodler In 0.18 the ideal way for doing these things will be using descriptors; so you can just import "descriptor":"sh(wsh(multi(2,<pubkey>,<pubkey>,<pubkey>)))" for example to get P2SH-P2WSH-multisig.

Agree with the above comments, it was definitely a bug + temporary workaround not intended behaviour, so I don't think it should be preserved. But regarding the addmultisigaddress comment, I plan on opening a PR to add the redeemScript to that output anyway, as a follow-up to #14481

I'll close this as not-a-bug, or wont-fix, depending on interpretation ;)