Issue867 Add CLI command to add signatures to a partially signed transaction

[windycrypto]

Fix for #867

Note by @abitmore: the final implementation of this PR is online-signing but not offline-signing.

--- below is original message by @cifer-lee ---

Roughly implemented multi-sig approach.

Some explanation:

The new added multisig_trx_data struct was in consideration of user experience, thus when import the transaction we will know what keys are needed to sign it. But I'm not sure if this was a appropriate practice, if it was, I feel like to show which keys were already signed and which were not --- so make it more user friendly.

Would like to hear from more devs.

[pmconrad]

IMO this API is much too complicated. Specifically, it introduces a new type of state into the wallet API, which can lead to surprising behaviour.

I think the desired functionality can be implemented with a simple idempotent call - tx with no/few sigs in, tx with more/all sigs out.

[windycrypto]

@pmconrad thanks very much~ I have added some explanation on this topic. The new introduced type was intend to easily use..
Maybe you are right, I should implemented it simpler..

[ryanRfox]

@cifer-lee Based on your comments above, it seems you are now refactoring Issue #867. Therefore, I will Label this PR as 2 Developing. Once you complete development on the Issue #867 , notify me and I will mark it 2 Approved (by you the developer, because it passes your testes). Next, I will mark this PR as 2 Final Review which will prompt @pmconrad to review your updates. Ultimately, he will mark merge the PR and Close the Issue.

This whole flow/organization is still a work in progress, so appreciate your comments for improvement.

[jmjatlanta]

I have not tested this code (will do so soon), but did find a few areas where you may want to fix the log messages...

[windycrypto]

Thanks Peter and John~ @pmconrad I have made a lot of simplification according to your suggestion, may you check it again~

So far, self test ok:

1. got a signed_transaction output using any existing api
2. stripped spaces of the signed transaction and pasted to call with multisig_sign_transaction
3. got a multi-signed transaction output

[abitmore]

I wrote "rename to sign_transaction2 to add some confusion" because IMHO sign_transaction2 would confuse people. Hoped that someone else would come up with a better idea, but still waiting.. @ryanRfox @pmconrad @jmjatlanta @oxarbitrage please advise. Thanks.

Anyway, the comments/docs should clearly describe the differences between the 2 commands.

BTW @cifer-lee please resolve conflicts.

[jmjatlanta]

As for the method name, add_transaction_signature

Pro: kind of indicates that others can add their signature as well.
Con: functionality is very similar to sign_transaction but the name does not make that obvious.

[windycrypto]

I think sign_transaction2 is ok in here, the difference with sign_transaction is just whether or not keep the origin signatures.

[cogutvalera]

I do not think sign_transaction2 is very meaningful. It doesn't make sense for me.
add_transaction_signature as @jmjatlanta suggests makes more sense for me.
IMHO add_partially_signed_transaction_signature is more clear.

[ryanRfox]

@cifer-lee please review the Reviewer comments above and update the PR accordingly. I added this to the 201904 Feature Release.

[windycrypto]

@cifer-lee please review the Reviewer comments above and update the PR accordingly. I added this to the 201904 Feature Release.

@ryanRfox Okay, sorry for the delay of work, I will pick it up this afternoon

[windycrypto]

As for the method name, add_transaction_signature

Pro: kind of indicates that others can add their signature as well.
Con: functionality is very similar to sign_transaction but the name does not make that obvious.

I think we need a name to convey the meaning of "sign this transaction to add new signatures while keeping the existing signatures unchanged". So I think a name like sign_transaction_keep_existing_signatures is more meaningful.. But it's really long so actually I still think sign_transaction2 is a appropriate name, using number suffix in function name is not a bad practice, even well-known projects adopt this naming convention too, e.g. open2 dup2 in linux source code.

IMHO, sign_transaction2 is a nice name in two aspects:

1. It's short, thus easy to read
2. It clearly indicates that it's another version of sign_transaction, just with minor difference

We just need clearly describe the difference in document.

@jmjatlanta @abitmore @cogutvalera please reconsider it?

---

Update:

I just come up with a few new names: sign_transaction_incrementally / sign_transaction_by_append. I'm not sure if users can catch the point. How do you think about this name?

[pmconrad]

IMO add_transaction_signature is sufficiently clear in that it only adds something, which implies that it does not remove existing signatures.
Much better than sign_transaction2, which gives no indication in what way it is different than sign_transaction. (Note that in your dup2 example, 2 is the number of arguments which is different from dup which has only 1 argument.)

[abitmore]

Sorry, #1626 introduced a conflict, please fix it. Thanks.

other developers have different opinions

[windycrypto]

Oh, have resolved the conflicts