Party ID logged during TSS signing is no the same as on-chain index of a member
Opened this issue · 0 comments
#616 calculates the order in which signers should submit a signature by determining signer's position in address[] public members
field of AbstractBondedECDSAKeep
and then calculating the delay based on signer's index:
Lines 465 to 478 in 7062199
Line 454 in 7062199
Looking into logs, it seems that the signer that was identified as tss-lib party {2,}
during the signature generation has index 0
in the array array returned from GetMembers()
. Signer that was identified as tss-lib party {1,}
has index 1
in that array and the signer that was identified as tss-lib party {0,}
has index 2
in that array:
2020-11-23T12:08:12.948+0100 INFO tss-lib party {2,}: signing finished!
2020-11-23T12:08:12.956+0100 INFO keep-ecdsa waiting [0s] before publishing signature for keep [0xe1C664a63BD1D017270aaE9bda45A776bA86a078]
020-11-23T11:22:10.222+0100 INFO tss-lib party {1,}: signing finished!
2020-11-23T11:22:10.226+0100 INFO keep-ecdsa waiting [5m0s] before publishing signature for keep [0xe1C664a63BD1D017270aaE9bda45A776bA86a078]
2020-11-23T11:22:10.226+0100 INFO tss-lib party {0,}: signing finished!
2020-11-23T11:22:10.229+0100 INFO keep-ecdsa waiting [10m0s] before publishing signature for keep [0xe1C664a63BD1D017270aaE9bda45A776bA86a078]
This is confusing and can potentially mislead during problem debugging or when recovering from an issue.
We should make sure party ID logged during tss signing is the same as the index under which the given member is stored on-chain.
Important: This may not be just a logging problem. When working on this issue make sure the change is backward-compatible and two clients in different versions can still generate a signature.