Conversation
|
I'd rather not have a separate key file for this feature and try to protect it with a password and all that. This More importantly the use of bip322 as I understand it is to be able to sign a message using what ever bitcoin descriptor script your bitcoin wallet has. If you have a private key descriptor you should be able to sign the invalid input message transaction, and if not you should still be able to create a PSBT that your hardware signers can sign. Does that make sense? Also have you seen the bdk-reserves crate? it does something similar. |
Still actively working on this. Will push updates once I'm done (likely within this week). |
|
Hi @aagbotemi, can you rebase and fix the CI failures? |
tvpeter
left a comment
There was a problem hiding this comment.
This PR is similar to the reserves feature that was taken out in the v1.0 update.
@aagbotemi, please consider changing bip322 feature flag to reserves.
Also, after rebasing, check whether there is need to update the workflows. I would prefer the workflows to go in a separate PR if necessary.
@notmandatory I don't know whether it is ideal that the crate he is referencing should be published for security reasons.
Alright @tvpeter, will do that.
Yes, I'm revamping the crate to use descriptor instead of private key, but I will continue the PR with the reserves crate. |
|
Thanks for the review and suggestion to rename the A PR (bitcoindevkit/bdk-reserves#39) updates |
9afe400 to
7b94ecc
Compare
7b94ecc to
007f71b
Compare
|
The update is a follow-up to the BIP322 refactor that migrated signing from raw keys to descriptor-based |
Pull Request Test Coverage Report for Build 21173945658Details
💛 - Coveralls |
Description
This PR added BIP322 feature into BDK CLI.
It also has a usage file for testing purposes
Notes to the reviewers
Changelog notice
Checklists
All Submissions:
cargo fmtandcargo clippybefore committingNew Features:
CHANGELOG.md