In enterCommit(height int, commitRound int),

  // If we don't have the block being committed, set up to get it.
  if !cs.ProposalBlock.HashesTo(hash) {
    if !cs.ProposalBlockParts.HasHeader(partsHeader) {
      // We're getting the wrong block.
      // Set up ProposalBlockParts and keep waiting.
      cs.ProposalBlock = nil
      cs.ProposalBlockParts = types.NewPartSetFromHeader(partsHeader)
    } else {
      // We just need to keep waiting.
    } 
  }

We just trample the ProposalBlock & ProposalBlockParts, regardless of what the proposal was for that round (in this case the current round was 1, but commitRound was 0, so ProposalBlock* was getting reset to round 0's proposal). Which is fine, but it causes an unexpected error message in the logs.

There are two "solutions" to this issue... one is, we stop trampling Proposal* upon discovery of a Commit (+2/3 precommits), but rather use a new field in RoundState... The other is to not log an error on certain conditions... Probably best to do the first solution but to do it later when we have time.

Proposal getting out of sync with ProposalBlockPart? #304

Description

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions