fix(rtps): same-network locator selection uses the interface's prefix length, not a fixed /24 by finger563 · Pull Request #867 · esp-cpp/espp

finger563 · 2026-11-01T03:54:53Z

Summary

Follow-up to #866 / #865. The same-subnet locator preference compared the first three octets, i.e. it assumed every network is a /24. The issue's own host had its Wi-Fi on 10.5.15.154/17: a board on that network with a different third octet would have fallen back to the first announced locator, the original bug in a narrower form.

rtps core: isSameSubnetAddress() / isSameSubnet() take a prefix length (default 24, the previous behaviour; clamped to 32, 0 matches everything); prefixLengthFromNetmask() converts a dotted-quad mask (non-contiguous masks are rejected). The participant carries m_localPrefixLength, set through DomainConfig::local_prefix_length and Participant::reuse(), and every selection site fix(rtps): prefer the peer's unicast locator on our subnet; correct the GAP length #866 added (SEDP topic data, the CycloneDDS default-locator fallback, the SPDP locator lists) plus the existing metatraffic selection use it. The SPDP multicast lists no longer go through the on-network replacement, which never applied to them.
Facade: RtpsParticipant::Config::interface_netmask ("255.255.128.0" or "17"). Empty means: on the host, the netmask of the interface that owns interface_address (read from getifaddrs / GetAdaptersAddresses, for an auto-detected and for an explicitly configured address alike); on ESP targets, a /24 with a debug log. An invalid value fails start().
Ethernet::netmask() and ethernet_netmask() on the four Ethernet boards, so an application can pass the netif's mask next to its IP; the rtps and esp32-p4-function-ev-board examples do.
Python: RtpsParticipant.Config(interface_netmask=...) (keyword, appended last so positional callers are unaffected).
Test: pc/tests/rtps_locator_subnet gains the prefix-length checks (explicit /16, /17, /32, /0, clamp, netmask conversion) and /17 SEDP + SPDP selection cases, alongside the existing /24 ones.

No wire-format change.

Test plan

Host: rtps_locator_subnet (new checks), rtps_golden, rtps_sedp_dedicated_locator, rtps_pubsub pass against the installed library
components/rtps/example builds for esp32 (IDF 6.1, component manager off)
components/esp32-p4-function-ev-board/example builds for esp32p4 (IDF 6.1)
Host library with ESPP_BUILD_PYTHON=ON builds (binding compiles)
CI (interop matrix, static analysis, examples)

🤖 Generated with Claude Code

https://claude.ai/code/session_01DFjjnq3XCTRAXENSxsCxJU

… length, not a fixed /24 The locator preference from #866 compared the first three octets, so a peer on a network wider than a /24 (the issue's host had its Wi-Fi on 10.5.15.154/17) with a different third octet fell back to the first announced locator, the original #865 bug in a narrower form. - rtps core: isSameSubnetAddress() / isSameSubnet() take a prefix length (default 24, the previous behaviour); prefixLengthFromNetmask() converts a dotted-quad mask. The participant carries the local prefix length (DomainConfig::local_prefix_length -> Participant::reuse()), and every selection site uses it: SEDP topic data, the CycloneDDS default-locator fallback, the SPDP locator lists and the metatraffic selection. The SPDP multicast lists skip the on-network replacement, which never applied to them. - Facade: RtpsParticipant::Config::interface_netmask ("255.255.128.0" or "17"). Empty: the netmask of the interface that owns interface_address on the host (getifaddrs / GetAdaptersAddresses, auto-detected or configured address alike), a /24 on ESP targets. - Ethernet::netmask() and ethernet_netmask() on the Ethernet boards so applications can pass the netif's mask; the rtps and esp32-p4-function-ev-board examples do. - Python: RtpsParticipant.Config(interface_netmask=...). - pc/tests/rtps_locator_subnet: prefix-length checks and /17 SEDP + SPDP selection cases next to the existing /24 ones. No wire-format change. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DFjjnq3XCTRAXENSxsCxJU

github-actions · 2026-11-01T03:56:06Z

✅Static analysis result - no issues found! ✅

Copilot

Warning

Copilot couldn't run its full agentic review because it didn't start before the timeout. Make sure your repository has a runner available, or add a copilot-code-review.yml file specifying one with the runs-on attribute. See the docs for more details.

Copilot review overview

3 open issues

parse_netmask() accepts dotted-quad strings with trailing garbage (e.g. \"255.255.255.0foo\")… · New
This adds an unconditional INFO log on every start() even when interface_address is explicitly… · New
The helper name isSameSameSubnetClamp reads like a typo and is hard to parse. Renaming to… · New

What changed in this PR

Updates RTPS “same subnet” locator selection to use the interface’s actual prefix length (CIDR) instead of assuming a fixed /24, improving correctness on networks like /17 and other non-/24 subnets.

Changes:

Add prefix-length aware subnet checks (isSameSubnetAddress, isSameSubnet) and netmask→prefix conversion.
Thread the local prefix length through participant/domain/discovery selection paths (SEDP/SPDP/metatraffic).
Expose interface_netmask in the facade (C++/Python) and surface Ethernet netmask on ESP boards + update examples/tests.

File	Description
pc/tests/rtps_locator_subnet.cpp	Expands tests to cover prefix-length behavior (/16, /17, /32, /0) and netmask conversion.
lib/python_bindings/rtps_bindings.cpp	Adds `interface_netmask` to Python config/init and forwards into `Rtps::Config`.
components/rtps/src/rtps_participant.cpp	Implements interface netmask parsing + host netmask detection and passes computed prefix length into Domain config.
components/rtps/src/entities/Participant.cpp	Stores and reuses local prefix length in Participant instances.
components/rtps/src/entities/Domain.cpp	Passes domain-configured prefix length into participant reuse.
components/rtps/src/discovery/TopicData.cpp	Makes SEDP unicast locator preference prefix-length aware.
components/rtps/src/discovery/SPDPAgent.cpp	Uses participant prefix length when selecting on-subnet metatraffic locator.
components/rtps/src/discovery/SEDPAgent.cpp	Uses participant prefix length when selecting on-subnet endpoint locators and parsing TopicData.
components/rtps/src/discovery/ParticipantProxyData.cpp	Uses prefix length in list replacement logic; prevents multicast lists from using on-network replacement.
components/rtps/include/rtps_participant.hpp	Adds `interface_netmask` to facade config and updates interface resolution API.
components/rtps/include/rtps/entities/Participant.hpp	Adds participant field for local prefix length and updates reuse signature.
components/rtps/include/rtps/entities/Domain.hpp	Adds `local_prefix_length` to `DomainConfig`.
components/rtps/include/rtps/discovery/TopicData.hpp	Updates `readFromBuffer` API to accept optional prefix length.
components/rtps/include/rtps/discovery/ParticipantProxyData.hpp	Documents and updates locator-list reading API to include prefix length.
components/rtps/include/rtps/common/Locator.hpp	Introduces `DEFAULT_SUBNET_PREFIX_LENGTH`, prefix-aware subnet checks, and netmask→prefix conversion.
components/rtps/example/main/rtps_example.cpp	Passes Ethernet netmask into `RtpsParticipant` config on ESP example.
components/rtps/README.md	Documents `interface_netmask` behavior and recommendations for ESP targets.
components/ethernet/src/ethernet.cpp	Tracks netmask alongside IP and resets it on link/IP changes.
components/ethernet/include/ethernet.hpp	Exposes `Ethernet::netmask()` and stores netmask atomically.
components/esp32-p4-module-dev-kit/include/esp32-p4-module-dev-kit.hpp	Adds `ethernet_netmask()` helper.
components/esp32-p4-function-ev-board/include/esp32-p4-function-ev-board.hpp	Adds `ethernet_netmask()` helper.
components/esp32-p4-function-ev-board/example/main/esp32_p4_function_ev_board_example.cpp	Fetches netmask and passes it to RTPS participant config.
components/esp32-p4-eth/include/esp32-p4-eth.hpp	Adds `ethernet_netmask()` helper.
components/esp32-ethernet-kit/include/esp32-ethernet-kit.hpp	Adds `ethernet_netmask()` helper.

_{Review effort: Lite}

finger563 · 2026-11-01T04:31:59Z

+bool parse_netmask(const std::string &text, uint8_t &prefix_length) {
+  unsigned a = 0, b = 0, c = 0, d = 0;
+  if (std::sscanf(text.c_str(), "%u.%u.%u.%u", &a, &b, &c, &d) == 4) {
+    if (a > 255 || b > 255 || c > 255 || d > 255) {
+      return false;
+    }


Done in 9fbb686: the dotted-quad parse now uses a %c sentinel and only accepts exactly four conversions, so "255.255.255.0foo" is rejected (and "17" already used the same sentinel).

finger563 · 2026-11-01T04:32:01Z

+    logger_.debug("No netmask known for {}; assuming a /{} network (set interface_netmask)", addr,
+                  prefix_length);
+  }
+  logger_.info("Interface {}/{}", addr, prefix_length);


Done in 9fbb686: the auto-detected address is logged at INFO only where auto-detection happens (host, as before this PR); the per-start "Interface a.b.c.d/n" line is now DEBUG, so an explicitly configured address adds no INFO output.

finger563 · 2026-11-01T04:32:03Z


-bool check_spdp() {
-  // 3. Fill each list with off-subnet locators, then announce the on-subnet one.
+bool isSameSameSubnetClamp() {


Renamed to isSameSubnetPrefixClamp() in 9fbb686, with a comment on what it checks.

- parse_netmask(): a dotted quad with trailing characters ("255.255.255.0foo") is rejected (a %c sentinel must not match). - resolve_interface(): the auto-detected address is logged at INFO where it is found (host blocks only, as before); the per-start "Interface a.b.c.d/n" line is DEBUG. ESP targets share the empty-address error with the host instead of an early return, which also removes the duplicate / always-false auto_detect conditions cppcheck reported on the ESP configuration. - Windows adapter cursor is a pointer to const. - Test helper renamed isSameSubnetPrefixClamp(). Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DFjjnq3XCTRAXENSxsCxJU

Copilot AI balanced review requested due to automatic review settings November 1, 2026 03:54

Copilot AI reviewed Nov 1, 2026

View reviewed changes

Copilot stopped reviewing on behalf of finger563 due to an error November 1, 2026 04:14
The job was not acquired by a runner even after multiple attempts.

Navigation Menu

Repository navigation

fix(rtps): same-network locator selection uses the interface's prefix length, not a fixed /24 - #867

fix(rtps): same-network locator selection uses the interface's prefix length, not a fixed /24#867
finger563 wants to merge 2 commits into
mainfrom
fix/rtps-subnet-netmask

finger563 commented Nov 1, 2026

Uh oh!

github-actions Bot commented Nov 1, 2026 •

edited

Loading

Uh oh!

Copilot AI left a comment

Uh oh!

finger563 Nov 1, 2026

Uh oh!

finger563 Nov 1, 2026

Uh oh!

finger563 Nov 1, 2026

Uh oh!

This branch has not been deployed

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

finger563 commented Nov 1, 2026

Summary

Test plan

Uh oh!

github-actions Bot commented Nov 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

✅Static analysis result - no issues found! ✅

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Copilot review overview

Uh oh!

finger563 Nov 1, 2026

Choose a reason for hiding this comment

Uh oh!

finger563 Nov 1, 2026

Choose a reason for hiding this comment

Uh oh!

finger563 Nov 1, 2026

Choose a reason for hiding this comment

Uh oh!

This branch has not been deployed

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

github-actions Bot commented Nov 1, 2026 •

edited

Loading