Skip to content

Unsettled RSKj parity: which writes survive when a Bridge call throws #380

Open

Description

@SDL-A1

From the external review of #374. An open question, not a known defect.

When a Bridge call throws, RSKj keeps writes made directly to the repository
before the throw — the BTC block store, and the transfer — and loses only the
provider-cached writes.

Nobody has established whether Rustock drops all of them.

If Rustock discards writes RSKj keeps, that is a consensus divergence on any
mainnet block containing a throwing Bridge call. If it keeps them the same way,
there is nothing to do but write the test that says so.

Why it needs settling rather than reasoning

Per CLAUDE.md: a consensus rule stricter than the chain it must accept is a
fork, and argument is not evidence. The resolution is a replay over mainnet
blocks that contain a throwing Bridge call, not a reading of both codebases.

Related: #295 covers the pre-RSKIP197 case where a throwing precompile must kill
the transaction rather than just burn forwarded gas. These may share a root.

What would settle it

Identify mainnet blocks with a throwing Bridge call, then replay them with
mainnet_block_rules and compare the resulting state root against the committed
one.

Reported by the external reviewer.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions