fix(ethrpc): prevent uint64→int64 overflow in AssembleChain33Tx - #3
Closed
bysomeone wants to merge 627 commits into
Closed
fix(ethrpc): prevent uint64→int64 overflow in AssembleChain33Tx#3bysomeone wants to merge 627 commits into
bysomeone wants to merge 627 commits into
Conversation
Bumps [golang.org/x/net](https://github.com/golang/net) from 0.8.0 to 0.17.0. - [Commits](golang/net@v0.8.0...v0.17.0) --- updated-dependencies: - dependency-name: golang.org/x/net dependency-type: direct:production ... Signed-off-by: dependabot[bot] <support@github.com>
Proxy exec update
update checkProxyExecTx
平行链可以开启收取手续费
Recomputed from latest coverage_tn.out, excluding *.pb.go files (go tool cover -func reports 65.5% statement coverage). Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Establish docs/superpowers/ as the standard location for design specs and implementation plans produced via the Claude + superpowers workflow. - README.md explains the brainstorm → spec → plan → execute → verify flow - specs/ and plans/ subdirectories with .gitkeep placeholders - Include the spec and plan for this Claude collaboration setup itself Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Document chain33-specific conventions Claude (and humans new to the codebase) need to know: Makefile-wrapped git workflow, GOPATH path constraint, GOPROXY for CN users, layered module map, testing discipline, and pitfall list. The conventions cover the chain33 specifics that aren't visible from code alone — for instance, `make branch` / `make push` wrap git, `types/*.pb.go` shouldn't be hand-edited, and `.travis.yml` / `.gitlab-ci.yml` should not be touched without confirmation. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Two corrections from code review of the initial CLAUDE.md commit: - 模块计数:目录速查列了 18 个代码模块,文首误标为 19;修正 并说明 build/ 与 docs/ 不算代码模块 - gg18 TSS 实际位于 system/crypto/tss/gg18/,不在 p2p/ 下; 移除 p2p 段的错误括注,改在 system/ 段补充指引 Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Add team-shared Claude Code permissions for chain33. Allowlist covers: - make targets that are read-only or safely scoped (test, race, lint, coverage, fmt, depends, build, cli) — excludes push/sync/docker - go subcommands that don't mutate remote state - git read-only commands (status, diff, log, show, blame) - search/navigation utilities (ls, find, grep, rg, wc, file) - gh read commands (pr/issue/run view & list) Commands that mutate shared state (git push, make push, gh pr create, git reset --hard) stay outside the allowlist and prompt every time. Adjust .gitignore so .claude/settings.json (team config) is tracked while .claude/settings.local.json (personal overrides) stays ignored. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Three small corrections from the final code review:
- Add coverage_*.out and build/coverage/ to .gitignore so the claim in
CLAUDE.md ("已在 .gitignore") becomes factually true; also stops these
test artifacts from cluttering `git status`.
- Align workflow stage naming: CLAUDE.md said "TDD" where the rest of
the docs say "execute"; change to "execute(TDD)" to keep TDD visible
while matching docs/superpowers/README.md.
- Add a back-reference from docs/superpowers/README.md to CLAUDE.md so
contributors landing on the README know where the entry point is.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The old badge used a stale camo-proxied URL whose hex payload pointed to golang/gddo's status.svg on godoc.org (now shut down). The image no longer loads. Replace with the current pkg.go.dev badge pointing at this repo. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Establish docs/superpowers/ as the standard location for design specs and implementation plans produced via the Claude + superpowers workflow. - README.md explains the brainstorm → spec → plan → execute → verify flow - specs/ and plans/ subdirectories with .gitkeep placeholders - Include the spec and plan for this Claude collaboration setup itself Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Document chain33-specific conventions Claude (and humans new to the codebase) need to know: Makefile-wrapped git workflow, GOPATH path constraint, GOPROXY for CN users, layered module map, testing discipline, and pitfall list. The conventions cover the chain33 specifics that aren't visible from code alone — for instance, `make branch` / `make push` wrap git, `types/*.pb.go` shouldn't be hand-edited, and `.travis.yml` / `.gitlab-ci.yml` should not be touched without confirmation. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Two corrections from code review of the initial CLAUDE.md commit: - 模块计数:目录速查列了 18 个代码模块,文首误标为 19;修正 并说明 build/ 与 docs/ 不算代码模块 - gg18 TSS 实际位于 system/crypto/tss/gg18/,不在 p2p/ 下; 移除 p2p 段的错误括注,改在 system/ 段补充指引 Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Add team-shared Claude Code permissions for chain33. Allowlist covers: - make targets that are read-only or safely scoped (test, race, lint, coverage, fmt, depends, build, cli) — excludes push/sync/docker - go subcommands that don't mutate remote state - git read-only commands (status, diff, log, show, blame) - search/navigation utilities (ls, find, grep, rg, wc, file) - gh read commands (pr/issue/run view & list) Commands that mutate shared state (git push, make push, gh pr create, git reset --hard) stay outside the allowlist and prompt every time. Adjust .gitignore so .claude/settings.json (team config) is tracked while .claude/settings.local.json (personal overrides) stays ignored. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Three small corrections from the final code review:
- Add coverage_*.out and build/coverage/ to .gitignore so the claim in
CLAUDE.md ("已在 .gitignore") becomes factually true; also stops these
test artifacts from cluttering `git status`.
- Align workflow stage naming: CLAUDE.md said "TDD" where the rest of
the docs say "execute"; change to "execute(TDD)" to keep TDD visible
while matching docs/superpowers/README.md.
- Add a back-reference from docs/superpowers/README.md to CLAUDE.md so
contributors landing on the README know where the entry point is.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
…bility) Fix critical bugs found during systematic code review including: - Remote crash vectors (DER parsing panic, MAVL proof nil deref, download nil deref) - Security issues (webhook RCE, AES nonce/IV reuse, ethrpc auth bypass) - Consensus risks (map iteration non-determinism, executor cache logic error) - Resource leaks (ethrpc Close() not stopping HTTP server) - Concurrency bugs (queue double-close, WaitGroup race, tasks slice race) - Correctness (account overflow, error swallowing, CORS method typo) Each fix includes a regression test demonstrating the bug pattern. See docs/bugfix.md for detailed descriptions of all findings. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
为 chain33-cli account 子命令增加 k2addr：直接根据 hex 私钥推导 BTC/ETH 地址， 无需启动 RPC 节点。 ./chain33-cli account k2addr -k <hex> [-t 0|1|2] addressType: btc(0), btcMultiSign(1), eth(2) Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Go 1.23 锁定了 linker 对 //go:linkname 的使用 (cmd/link 默认开启
-checklinkname=1)，fjl/memsize 通过 //go:linkname 拿到 runtime.stopTheWorld，
触发 "invalid reference to runtime.stopTheWorld" 链接错误，导致 Go 1.23+
工具链下 chain33 编译失败。
go-ethereum 在 PR #30253 (2024-08-02 合并) 移除了 memsize 依赖，首个稳定
发行版本是 v1.14.8。本提交把 ethereum/go-ethereum 从 v1.12.0 升到 v1.14.8，
连同其传递依赖一起升级 (protobuf v1.31→v1.34, golang.org/x/net v0.17→v0.24
等)，从根上去掉 -checklinkname=0 这种绕过 workaround 的需求。
rpc/jrpchandler_test.go: protobuf v1.34 在某些路径上会写 sizeCache 内部字段，
导致 testify mock 用 `On("SendTx", tx)` 做按值比较时与运行时的 tx 实例不等。
改用 mock.MatchedBy + proto.Equal 做语义比较。
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
7198f44 改 CBCEncrypterPrivkey 使用随机 IV (修 F-WAL-002 IV 复用)， 随之产生两处后续问题，本提交一起修掉： 1. wallet/wallet_proc.go: ProcImportPrivKey 通过比较 ciphertext hex 字符串判断私钥是否已存在；随机 IV 后同一私钥每次加密结果不同， 导致 `Account.Privkey == Encrypteredstr` 永远不等，把"已存在"误判 成 ErrPrivkey。改成解密 Account.Privkey 后与原始 privkey bytes 比较。 这也是 TestAll → testProcImportPrivkeysFile2 在 PR 分支上回归 (master 通过) 的根因。 2. wallet/common/common_test.go: 测试原本用 16 字节私钥；新格式 `IV(16) + ciphertext(len(privkey))` 下 16 字节输入会让"new format" 解密分支返回非 32 字节结果而回退到 legacy，断言失败。改成 32 字节 私钥 (匹配 secp256k1 生产实际)，并把固定长度断言从 `len(privkey)` 改为 `aes.BlockSize + len(privkey)`。 Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
- common/db/go_pegasus.go: wrap deferred Delete bench write in a closure so time.Since(start) measures elapsed time, not 0 (defer evaluates args at registration time) - queue/exec_map_order_bug_test.go: gofmt key-value alignment - system/mempool/mempool_test.go: drop pointless append() over single slice value (vet "append with no values" warning) - system/p2p/dht/protocol/download/wgdone_bug_test.go: remove unreachable wg.Done() line; comment explains why the test deliberately reproduces the buggy pattern via panic Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
go-ethereum v1.14.8 (upgraded in 3ae8ca5) requires Go 1.21+ — the toolchain directive in go.mod becomes "unknown directive: toolchain" when parsed by 1.20, which broke the build-with-latest-stable-go matrix job and the AppVeyor build. - .github/workflows/build.yml: bump the only hardcoded version (`go-version: '1.20'`) to 1.21. All other jobs already read from `go-version-file: 'go.mod'`. - appveyor.yml: switch the hardcoded C:\go120 GOROOT/PATH to AppVeyor's `stack: go 1.21` directive, which auto-installs the requested Go toolchain regardless of image preset. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The check_fmt job runs ineffassign/gosec via golangci-lint v1.55.2,
whose bundled go/types cannot parse Go 1.24's stdlib export data
("unsupported version: 2"). The toolchain go1.24.5 directive (auto-
added by go mod tidy on a 1.24 host) forces CI to download 1.24 even
though setup-go installed 1.21 — breaking the lint pass with
spurious typecheck failures across packages.
Drop it. go 1.21 alone is enough for go-ethereum v1.14.8; the
toolchain directive was incidental, not required.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
- queue/rpc_close_leak_bug_test.go: add ReadHeaderTimeout to the test http.Server to satisfy gosec G112 (Slowloris). - queue/webhook_rce_bug_test.go: annotate the env-var-sourcing placeholder string with #nosec G101; the value is not a real credential. - types/evm_event.pb.go: regroup imports (stdlib first) per goimports v0.24.0 — pre-existing generated-file drift that newer goimports flags as needing 'make fmt'. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
queue/download_tasks_race_bug_test.go declared `index` and `name` fields on the local task struct but never read them, which golangci-lint structcheck flags as unused. Access them in the read loops where the original code already touches the slice — the test remains a race-detection demonstration, just with the field accesses made explicit. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The concurrent-sign section judged pass/fail by a 60s wall-clock budget (select + time.After). Under -race plus per-sign Paillier-2048 keygen on a shared CI runner it intermittently overran the budget. The same commit passed gg18 on the push-event run and failed on the pull_request-event run of PR 33cn#1361 -- a classic flaky signature. Replace the wall-clock budget with a structural guarantee: - drop the select{ <-c / time.After(60s) } gate; use wg.Wait(), so the test body no longer has any path that fails due to elapsed time - correctness is decided solely by verifySignatureWithDKG (cryptographic verify), which is time-independent - reduce concurrent signs 10 -> 4 to shrink the -race CPU long tail - widen the remaining timeouts into deadlock-only backstops, centralized as named constants (peerWait/dkg/sign/reshare/barrier/childExit) Verified locally: -race x10 all PASS, no data race, ~15.4s; gofmt/go vet ok. Design: docs/superpowers/specs/2026-06-08-gg18-flaky-fix-design.md Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The test deliberately reproduced an "Add after goroutine launch" race
(go func(){ wg.Done() }(); wg.Add(1)) to demonstrate F-BCSYNC-001, but its
recover() sat in the OUTER goroutine while the panic happened in the INNER
child goroutine it spawned (push_wg_bug_test.go:30). Go's recover does not
cross goroutines, so the negative-WaitGroup panic escaped and crashed the
whole queue test binary whenever scheduling ran Done() before Add() --
intermittently failing the unit-test CI job (same commit passed on the
pull_request run, failed on the push run).
It also had no real value: blockchain/push.go already uses the correct
Add-before-go pattern (Add at :576, goroutine at :577), so the demonstrated
bug does not exist in the source; the test exercised only Go's sync.WaitGroup,
never chain33 code, and its sole assertion was assert.True(t, true).
Verified: queue package passes 20/20 runs after removal.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…rgence TestConnManager called FetchNearestPeers(1) immediately after RoutingTable().TryAddPeer and then indexed peers[0]. The routing table is eventually consistent, so FetchNearestPeers intermittently returned an empty slice; require.NotNil passed (an empty slice is non-nil) but peers[0] panicked with "index out of range [0] with length 0", crashing the unit-test CI job (same commit passed on the pull_request run, failed on the push run). Replace the immediate read with require.Eventually that polls until the routing table reflects h2, then asserts. Pre-existing flaky test, not introduced by this PR; conns.go was not the cause. Verified: TestConnManager passes 10/10 locally. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
First fix (poll FetchNearestPeers for 5s) still failed intermittently (--- FAIL: TestConnManager 5.10s). Root cause: h2 is a plain libp2p host that does NOT run the DHT protocol, so h1's DHT background maintenance evicts h2 from the routing table after the one-shot TryAddPeer; a read-only poll then never sees it again. Re-add h2 inside the poll loop (each tick TryAddPeer then read immediately) and widen the timeout to 30s, so a transient eviction is recovered on the next tick. Pre-existing flaky test, not introduced by this PR. Verified 5/5 locally (the flaky does not reproduce locally; CI is the real check). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
fix: address 20 bugs from codebase review
Same overflow pattern as the EVM plugin fix — bigAmount.Int64() and uint64(amount) conversions without range check. Add IsInt64() guard and nil protection in SendRawTransaction caller. Co-Authored-By: Claude <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
6 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
同步修复 eth_sendRawTransaction 入口中同类型的 uint64→int64 溢出隐患。
Changes
🤖 Generated with Claude Code
Co-Authored-By: Claude noreply@anthropic.com