Conversation
…s are passing now
…ructors. Edited discv4 client to not use only posix
- RlpxSession: inherit enable_shared_from_this; connect() returns shared_ptr; I/O coroutines capture shared_ptr to prevent use-after-free crash on session teardown - ETH/68+69: StatusMessage as std::variant<StatusMessage68, StatusMessage69> with get_common_fields() helper and 3-param validate_status() - Fix Snappy compression: msg_id uncompressed, payload compressed separately - Fix test files: eth_handshake_test, eth_messages_test, eth_watch_integration_test updated for variant API - 512/512 tests pass
… per chain or fork ID
…to match with what we want to watch
✅ ENRRequest wire encode (type 0x05) ✅ ENRResponse wire decode (type 0x06) ✅ request_enr() in discv4_client with reply matching ✅ ParseEthForkId() extracts ForkId from ENR record ✅ Bond → ENR → eth_fork_id enrichment in handle_neighbours coroutine ✅ make_fork_id_filter() + DialScheduler::filter_fn for pre-dial chain filtering
…ode registry, crawler, and client Co-authored-by: Super-Genius <1320023+Super-Genius@users.noreply.github.com>
Co-authored-by: Super-Genius <1320023+Super-Genius@users.noreply.github.com>
…+17 compatibility Co-authored-by: Super-Genius <1320023+Super-Genius@users.noreply.github.com>
…S.md Co-authored-by: Super-Genius <1320023+Super-Genius@users.noreply.github.com>
…ISTAKES.md M015 Co-authored-by: Super-Genius <1320023+Super-Genius@users.noreply.github.com>
- clean corrupted patch artifacts in discv5_client.cpp - fix rlp::Bytes/std::vector conversion boundaries - harden noexcept endpoint/address lookups with error_code overloads - link discv5 against rlpx to resolve Ecdh symbol - update CHECKPOINT.md with build-blocker/repair context
- expect unsolicited WHOAREYOU packets to remain uncounted - add regression assertions that unrelated protocol counters stay zero for generic valid UDP packets
… flow - initialize discv5_crawl with generated secp256k1 keypair - add handshake/message diagnostics to client + crawl summary - gate deep diagnostics to --log-level trace - reduce per-peer discovery logs at info level - align WHOAREYOU/handshake details closer to geth behavior - fix packet construction bug by preserving header/AAD and appending ciphertext
- fix compiler warnings across rlp, rlpx, discv4, and tests - normalize signed/unsigned size comparisons and remove tautological limit checks - harden frame and ECIES parsing with explicit wire-size validation - fully initialize discv4 pending reply state to avoid missing-field issues - restore selected warnings as errors (-Werror=unused-lambda-capture, -Werror=sign-compare, -Werror=type-limits) - replace protocol magic numbers with named constexpr constants in rlpx, discv4, and discv5 - clean up discovery tests by using condition-variable synchronization instead of sleep_for - remove unused lambda captures, dead test fields, and skip the non-real timeout test
- register discv4 pending reply matchers before sending PING, FINDNODE, and ENRRequest packets - erase pending discv4 reply entries on send failure to avoid stale state - prevent fast PONG/NEIGHBOURS/ENR responses from being dropped as unsolicited - clear stale discv5 pending FINDNODE requests that never received WHOAREYOU - allow discv5 silent peers to be retried instead of becoming permanently blocked by pending_requests_ - keep discovery retry state bounded and aligned with the intended request/response flow
…tdown ping() and request_enr() both treat operation_aborted as a successful reply, but stop() cancels every pending timer unconditionally, which caused the same code path to fire during shutdown races. - ping(): require pong_slot->expiration != 0 before declaring bond success; prevents default-constructed discv4_pong from marking unreachable peers as bonded and driving follow-up discovery on bad state. - request_enr(): require !enr_slot->record_rlp.empty() before returning the response; prevents an empty/i
Discv5 implementation
Warning cleanup
Removed triggers for push, pull_request, and schedule events.
Removed triggers for push, pull request, and schedule events.
Removed triggers for push, pull_request, and schedule events.
Removed triggers for push, pull request, and schedule events.
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 473efa351b
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
…king inactive ones
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: ca19c92c0b
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: ad037727b2
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| // Cancel the waiting find_node() coroutine — we got the reply. | ||
| it->second.timer->cancel(); |
There was a problem hiding this comment.
Cancelling the pending FIND_NODE timer on the first NEIGHBOURS packet means find_node() immediately wakes and erases the matcher, so additional NEIGHBOURS packets from the same reply window are dropped by the pending_replies_ guard. In discv4, peers are often split across multiple NEIGHBOURS packets due UDP size limits, so this truncates discovery results and can significantly reduce crawl coverage.
Useful? React with 👍 / 👎.
| const std::string key = reply_key(ip, port, kPacketTypeNeighbours); | ||
| auto timer = std::make_shared<asio::steady_timer>(io_context_); | ||
| pending_replies_[key] = PendingReply{ timer, nullptr, nullptr, {} }; |
There was a problem hiding this comment.
This matcher key is only ip:port:ptype, so concurrent requests of the same type to one endpoint overwrite each other in pending_replies_. When that happens, one coroutine can wait on a timer that will never be canceled, and later erase(key) calls can remove another in-flight request’s matcher, producing spurious timeouts and lost replies during concurrent crawl activity.
Useful? React with 👍 / 👎.
No description provided.