blockchaintest: blocks are executed from the decoded JSON fields, so the block RLP body is never validated against its header

未关闭
#1,726 0 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

维护者通常 1 天内回复

还没有人认领这个 Issue。

评估

难度
4/5
预计耗时
3-5 天
新手友好度
38/100
Issue 类型
缺陷
描述清晰度
描述清楚
活跃度
活跃
技术栈
cpp

调研方向

The runner currently applies blockchain_test blocks from the JSON decoded fields (blockHeader, transactions, uncleHeaders, withdrawals) instead of decoding the rlp bytes and checking that body against the header. Find where fixtures are loaded and blocks executed, then decode rlp as [header, txs, ommers, withdrawals] and fail when that body does not match the committed transactionsTrie (and related header fields). Done when a fixture whose rlp body was edited (last tx dropped) is rejected while an unedited multi-tx blockchain_test still passes.

由索引模型根据 Issue 内容生成。

描述

Context

While running the Merkle Patricia Trie structural test suite of https://github.com/ethereum/execution-specs/pull/3600 against several clients, I corrupted fixtures on purpose to check that each runner rejects them. That surfaced which part of a blockchain_test fixture evmone consumes.

A fixture block carries the block twice: rlp, the canonical bytes, and the sibling fields blockHeader, transactions, uncleHeaders, withdrawals, documented as "Decoded block header fields included in the block RLP" and "List of decoded transactions included in the block RLP". The consumption procedure says to decode rlp (step 7.2) and then apply "the current decoded block" (step 7.3).

So a fixture whose RLP body disagrees with the header contained in that same RLP is consistent on the JSON side and passes.

Evidence

One fixture, one block, two transactions. Two edits, each self-consistent on its own side:

A. Remove the last transaction from the block RLP body only, leaving blockHeader (including transactionsTrie) and the JSON transactions untouched.

B. Remove the last transaction from the JSON transactions array only, leaving rlp untouched.

A runner that follows 7.2 and 7.3 fails A and passes B. evmone passes A and fails B. Of eight runners I tried the same pair on, six (go-ethereum, erigon, reth, Nethermind, nimbus-eth1, Besu) fail A and pass B; evmone and ethrex invert it.

Reproduction

Take any blockchain_test fixture with a block holding at least two transactions, decode blocks[0].rlp as [header, txs, ommers, withdrawals], drop the last element of txs, re-encode, and write it back to rlp without touching the other fields. Then

evmone test <dir>

It passes; per steps 7.2 and 7.3 it should fail, because the decoded block's transactions do not hash to the transactionsTrie committed in its own header.

Fixtures to reproduce with: uv run fill tests/mpt/ --clean on the branch of ethereum/execution-specs#3600, or any existing multi-transaction blockchain_test.

主要语言
C++
星标
979
派生
340
平均合并
12 小时 16 分钟
30 天内合并 PR
25

环境准备

  • 提供 Dockerfile 或 Docker Compose 文件
  • 没有 Pull Request 模板
  • 没有贡献指南

从这里开始

  1. 先读完整个 Issue,再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库,在一个分支上完成修改。
  4. 提交 Pull Request,并在描述里引用这个 Issue 编号。

ipsilon/evmone 的其他 Issue

查看 ipsilon/evmone 的全部 Issue

相似的 Issue

更多 C++ Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。