fix(client): per-transport SD session tracking via dual discovery sockets (+ Windows CI) by JustinKovacich · Pull Request #130 · luminartech/simple_someip

JustinKovacich · 2026-06-16T17:10:20Z

What

Per-transport SD session tracking via dual discovery sockets, plus a windows-latest CI job to confirm the bind behavior is cross-platform.

Previously the client tracked a single SD session-counter domain across both multicast and unicast Service Discovery, so the two domains' counters collided and produced spurious SenderRebooted churn against a steadily-running sensor.

This change:

keeps the multicast discovery socket unchanged (INADDR_ANY:30490 + group join, still sends),
adds a receive-only unicast socket bound to the interface IP — "most-specific bind wins" diverts the sensor's unicast SD datagrams to it,
threads the real TransportKind into SessionTracker::check() via a shared process_discovery(), so the two session domains are tracked independently and stop colliding.

No bind-to-group (Windows-safe), no change to the send path.

Verification

Live, on real hardware (Iris 192.168.11.151, FW 0.11.0.3, via the dft/EnVision consumer with this branch patched in):

tcpdump on the SD port captured both 96 multicast offers and 31 unicast SD datagrams; the unicast SD landed on <iface>:30490 (the new socket) as designed.
0 false SenderRebooted over a 60s steady-state window; stable scan profile.
Survived a real power-cycle (clean evict → auto re-adopt) with 0 false reboots.

CI: Windows portability

The unicast divert ("most-specific bind wins") is the OS-dependent part of this design. simple_someip CI was ubuntu-only, so this PR adds a test-windows (windows-latest) job running cargo test --all-features — which builds the client socket code and runs socket_manager::dual_socket_splits_multicast_from_unicast (real binds + send/drain asserting the divert).

⚠️ Caveat: that test has SKIP guards (no multicast route / no egress IPv4 → early vacuous pass). Check the Windows runner log to confirm it actually exercised the divert rather than skipping. If it skips on the hosted runner, we'll need a loopback-only variant of the assertion.

Follow-ups (not in this PR)

Cut a release (0.7.1) so the dft consumer can drop its local [patch] and pick this up.
Pairs with iris_someip_client's service_instances() override (dft side) so SD reboot detection actually runs.

🤖 Generated with Claude Code

Pins the SessionTracker contract the SD transport-attribution fix relies on: interleaved multicast/unicast session domains for the same (service,instance) must not look like a reboot, while a real counter reset within a domain must. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Validates the per-transport SD fix's core assumption: a socket bound to the MULTICAST GROUP address (joined) receives only multicast, and a socket bound to the SPECIFIC host interface IP (not joined) receives only unicast — no cross-delivery. NOTE the spike also found the naive variant fails: binding the unicast socket to INADDR_ANY double-delivers (it receives the multicast too via SO_REUSEPORT), so the unicast socket MUST bind the specific host IP. Skips on hosts with no multicast route (e.g. lo-only CI). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…ANY mc + interface-IP unicast) recvmsg/PKTINFO is per-platform cmsg (no Windows/bare-metal story), so dual-socket is the portable mechanism. Bind-to-group (my first spike) is rejected on Windows, so validate the portable variant instead: keep the multicast socket on INADDR_ANY + join (unchanged, still sends) and add a more-specific interface-IP socket for unicast. 'Most-specific bind wins' diverts the sensor's unicast SD to it, so the wildcard socket sees only multicast — no bind-to-group, no send-path change. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…kets Splits the sensor's multicast and unicast SD session/reboot domains onto separate SessionTracker keys, fixing the false-reboot churn (interleaved counters looked like perpetual reboots). Mechanism (validated by dual_socket_splits test, portable to Windows + bare-metal Transport trait): - multicast discovery socket: unchanged (INADDR_ANY:30490 + group join, still sends outgoing SD); its receives are now tagged TransportKind::Multicast. - NEW receive-only unicast discovery socket bound to the interface IP — by most-specific-bind-wins the kernel diverts the sensor's unicast SD here; tagged TransportKind::Unicast. The shared process_discovery() feeds both arms, so each transport's session counter is tracked independently. No send-path change. SessionTracker was already correct (keys by (sender, transport, svc, inst)); this stops feeding it a hardcoded Multicast for every datagram. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

The per-transport SD fix binds a wildcard multicast socket alongside an interface-IP unicast socket and relies on "most-specific bind wins" to divert the sensor's unicast SD to the unicast socket. That divert is OS-dependent, so add a `windows-latest` job running the full `--all-features` suite (incl. the `client` socket code + `dual_socket_splits_multicast_from_unicast`) to confirm the behavior holds on Windows. Note: the divert test SKIPs (vacuous pass) on hosts with no multicast route / no egress IPv4 — watch the runner log to confirm it actually ran. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Copilot

Pull request overview

This PR fixes false SenderRebooted churn by tracking SOME/IP-SD session state separately for multicast vs unicast discovery traffic. It does so by introducing a dedicated receive-only unicast SD socket (bound to the interface IP) and threading the true TransportKind into session tracking, plus adds a Windows CI job to validate cross-platform bind behavior.

Changes:

Add SocketManager::bind_discovery_unicast() to create an interface-IP-bound SD receive socket.
Refactor SD receive handling in the client loop into a shared process_discovery() that keys SessionTracker by real transport (multicast vs unicast).
Add a windows-latest GitHub Actions job running cargo test --all-features.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 3 comments.

File	Description
`src/client/socket_manager.rs`	Adds unicast SD bind helper and a kernel-behavior regression test for multicast/unicast split.
`src/client/inner.rs`	Introduces a second discovery socket and routes SD datagrams through shared processing with correct `TransportKind`.
`src/client/session.rs`	Adds unit tests documenting/guarding the per-transport session tracking behavior.
`.github/workflows/ci.yml`	Adds a Windows CI job to exercise the socket binding behavior on Windows runners.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

…cket test `socket2::Socket::set_reuse_port` is Unix-only; the test helper called it unconditionally, breaking the Windows build (the just-added windows-latest CI job caught it). Production `bind_discovery*` already gates it `#[cfg(unix)]` and compiled fine on Windows — only the test helper was missing the gate. `set_reuse_address` (cross-platform) stays unconditional. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…lity `combined_sd_subscribe_uses_its_own_options_run` sent the test datagram to the SD socket's `local_addr()`, which is `0.0.0.0:<port>` (the SD socket binds the wildcard to receive multicast). Sending unicast to `0.0.0.0` loops back on Linux but is dropped on Windows, so the server's `recv_from` timed out — the new windows-latest CI job surfaced this pre-existing Linux-only assumption. Target `127.0.0.1:<port>` explicitly so the test is cross-platform; behavior on Linux is unchanged. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

codecov-commenter · 2026-06-16T18:04:12Z

Codecov Report

❌ Patch coverage is 80.76923% with 40 lines in your changes missing coverage. Please review.
✅ All tests successful. No failed tests found.

Files with missing lines	Patch %	Lines
src/client/inner.rs	71.42%	24 Missing ⚠️
src/client/socket_manager.rs	84.15%	16 Missing ⚠️

@@            Coverage Diff             @@
##             main     #130      +/-   ##
==========================================
- Coverage   96.55%   96.50%   -0.06%     
==========================================
  Files          30       30              
  Lines        7993     8144     +151     
==========================================
+ Hits         7718     7859     +141     
- Misses        275      285      +10

Files with missing lines	Coverage Δ
src/client/session.rs	`100.00% <100.00%> (ø)`
src/server/mod.rs	`93.20% <100.00%> (+0.01%)`	⬆️
src/client/socket_manager.rs	`87.94% <84.15%> (-1.11%)`	⬇️
src/client/inner.rs	`92.75% <71.42%> (+0.75%)`	⬆️

Copilot

Pull request overview

Copilot reviewed 5 out of 5 changed files in this pull request and generated 2 comments.

…ndows The previous commit's loopback retarget was a no-op: the SD socket binds `config.interface:30490` and the test config's interface is already `127.0.0.1`, so `local_addr()` was never the wildcard. The real cause is that every test server binds the same fixed `127.0.0.1:30490` with `SO_REUSEADDR`; under parallel execution Windows delivers the test's unicast to a different bound socket, so `recv_from` timed out. Drop the socket round-trip entirely: `message` is already in memory and the sender addr isn't asserted (the subscriber endpoint must come from the SubscribeEventGroup's `options[1]`), so parse it directly and call `handle_sd_message` with a synthetic sender. Hermetic, no shared-port dependency, identical assertions. Verified on Linux; behavior unchanged. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

The Windows job's purpose is to confirm the dual-socket SD bind divert is portable; that assertion lives in the library unit tests (`dual_socket_splits_multicast_from_unicast`). Running the full suite also ran `tests/client_server.rs`, whose integration tests bind fixed SD ports and are flaky under parallel execution on every platform — they fail on `main` on the same host and flake on the ubuntu coverage job too (tracked separately, #84). Split into `--no-run` (compile everything incl. integration tests = build portability) + `--lib` (run unit tests incl. the divert = behavior portability). The integration suite still compiles on Windows but isn't run here, so the job reflects this fix's portability without unrelated cross-platform flakiness. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Copilot

Pull request overview

Copilot reviewed 5 out of 5 changed files in this pull request and generated 3 comments.

- Bind the multicast probe socket to an ephemeral port (:0) and reuse the OS-assigned port for the unicast socket + sender target, instead of a hardcoded 51900 that could collide on a shared CI runner. - Treat probe send_to failures as an explicit, logged SKIP (env issue, not a logic regression) rather than relying on an empty drain to pass. - Fix inaccurate assertion message: the mc socket is bound to INADDR_ANY, not the group address. - Clarify bind_discovery_unicast doc: it is receive-only by usage, not by type (the returned SocketManager still carries an unused send half). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…-port self-collision The client's receive-only unicast discovery socket binds interface:30490 with SO_REUSEPORT. When the integration-test server bound its SD socket on the same IP:30490, the client's own SubscribeEventgroup could be SO_REUSEPORT load-balanced to the client's own discovery socket, so the server never registered the subscriber and wait_for_subscribers failed (client_server.rs:172). This is a test-only artifact: in production a full SD-announcing server is a remote sensor on its own IP, and the co-located server in iris_someip_client is new_passive (binds an ephemeral SD port, never 30490). So the collision only arises when a full Server::new is co-located on the same IP as a client, which only the integration test did. Move the test server to 127.0.0.2 (client stays 127.0.0.1), mirroring the remote-sensor topology. The serial-sd-port nextest group is retained — multiple client_server tests each bind their server's :30490 and must stay serial vs each other. Full suite passes via nextest (423/423). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

JustinKovacich · 2026-06-17T13:48:07Z

Pushed 9a2f3b2, which turns the Build, Test & Coverage job green. All five checks now pass.

Root cause of the coverage-job failure (client_server::test_add_endpoint_and_send_to_service and friends failing at client_server.rs:172, "server should have registered the subscriber"): a test-only artifact of this PR's new receive-only unicast discovery socket. That socket binds interface:30490 with SO_REUSEPORT. The integration-test server (Server::new) also binds its SD socket on the same IP :30490, so the client's own SubscribeEventgroup (sent unicast to 127.0.0.1:30490) could be SO_REUSEPORT load-balanced back to the client's discovery socket — the server never saw the subscribe, and wait_for_subscribers timed out.

Why it's production-safe (not a real regression): the collision only occurs when a full SD-announcing Server::new (binds :30490) is co-located on the same IP as a client. In production a full SD-announcing server is a remote sensor on its own IP; the co-located server in iris_someip_client is Server::new_passive, which binds an ephemeral SD port and never :30490. Only the integration test put both on 127.0.0.1.

Fix: move the test server to 127.0.0.2 (client stays 127.0.0.1), mirroring the remote-sensor topology. The serial-sd-port nextest group is retained — multiple client_server tests each bind their server's :30490, so they must stay serial vs. each other (that's the inter-test sharing, distinct from the intra-test collision this fixes).

Local verification: cargo nextest run --all-features → 423/423 (the CI-equivalent run, honors the serial group); fmt clean; clippy -D warnings -D clippy::pedantic clean on both --all-features and --no-default-features.

This supersedes the earlier "pre-existing flake / serial-group / PR #84" diagnosis in the threads above — it was neither pre-existing nor a parallelism/serial-group issue; it's an intra-test self-collision introduced by bind_discovery_unicast, now resolved test-side.

Copilot

Pull request overview

Copilot reviewed 6 out of 6 changed files in this pull request and generated 1 comment.

…g fix Add `most_specific_bind_diverts_unicast_on_loopback`, a deterministic companion to `dual_socket_splits_multicast_from_unicast`. The multicast variant needs a live multicast route and SKIPs on a lo-only host (CI), leaving the "most-specific bind wins" divert — the load-bearing mechanism of the per-transport SD fix — unexercised in coverage and prone to the vacuous-skip trap. The new test proves the same divert with pure unicast on loopback (ephemeral port, never the fixed SD port), so it never skips: a datagram to the interface IP lands on the interface-bound socket, not the wildcard socket sharing the port. Unix-gated (leans on SO_REUSEPORT, the discovery socket's own #[cfg(unix)] reuse path); cross-platform compilation stays covered by the build-all CI job. Also fix the lone `SubscribeEventgroup` -> `SubscribeEventGroup` (capital G) typo in the client_server doc comment to match the spec spelling used everywhere else in the crate (Copilot review). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

The companion test bound `0.0.0.0:<ephemeral>` with `SO_REUSEPORT`, which can collide with a parallel integration-test client that the OS hands the same ephemeral port: that client's `bind` (REUSEADDR only, no REUSEPORT) then fails with `EADDRINUSE` and its test fails. This surfaced as a `test_multiple_ subscribers_receive_events` failure on the coverage job. The production `bind_discovery_unicast` is already covered via the `inner.rs` discovery tests, so the marginal coverage isn't worth the cross-test hazard. The `SubscribeEventGroup` spelling fix (prior commit) stays. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Copilot

Pull request overview

Copilot reviewed 6 out of 6 changed files in this pull request and generated 1 comment.

The SD-port-collision comment said the client's unicast discovery socket uses `SO_REUSEPORT`, but `bind_discovery_unicast` only sets it under `#[cfg(unix)]` (Windows gets `SO_REUSEADDR` only). Reword to "address/port reuse — SO_REUSEPORT on Unix" so the comment is accurate cross-platform (Copilot review). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

… update The coverage job intermittently failed `test_multiple_subscribers_receive_events` with `client1 expected Unicast, got DiscoveryUpdated(SubscribeAckEventGroup)`. Root cause: this PR's unicast SD socket now delivers the SubscribeAck to the client as a `DiscoveryUpdated` on the same update channel as events. The tests did a single 250ms pre-drain then asserted the *next* update was the `Unicast` event — racy, because under the slower coverage build the ack can arrive after the pre-drain and get consumed where the event is expected. Add a `recv_unicast` helper that drains the update stream until the `Unicast` event arrives, skipping interleaved discovery acks / reboot notices (bounded by a 5s deadline). Apply it to every test that asserts event receipt. The opportunistic pre-drains are kept (harmless) and the helper makes ordering no longer load-bearing. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Copilot

Pull request overview

Copilot reviewed 6 out of 6 changed files in this pull request and generated no new comments.

Forward-ports #130 (released on the 0.7.x line, never ported to the 0.8.0 spine) onto the post-#125 client architecture. The spine's discovery-loop rewrite dropped #130's dual-socket split, so 0.8.0 regressed the false-reboot fix: a sensor's interleaved multicast (~1468) and unicast (~739) SD session counters collided on one SessionTracker key and looked like perpetual reboots. Mechanism (re-applied to the generic TransportFactory / BindDispatch / Spawner surface, not the old raw-socket2 path): - `SocketManager::bind_discovery_unicast_with_transport[_local]`: a receive-only unicast SD socket bound to the *specific* interface IP on the SD port (vs the multicast socket's INADDR_ANY), no group join. "Most-specific bind wins" diverts the sensor's unicast SD here. - `BindDispatch::bind_discovery_unicast`: claims a buffer lease and binds via the Send / Local dispatch (one extra socket-loop buffer per client). - `Inner`: new `discovery_unicast_socket`, best-effort bound in `bind_discovery` (failure is non-fatal — multicast still works), shut down in `unbind_discovery`. The discovery-handling block is extracted to `Self::handle_discovery_datagram` and driven from both the multicast arm (TransportKind::Multicast) and a new unicast arm (TransportKind::Unicast). The spine's SessionTracker already keys by transport, so no engine change there. Tests: port #130's two per-transport SessionTracker regression tests, the dual_socket_splits kernel-behaviour spike, the hermetic combined-SD server test, and the client_server hardening this fix requires (server on 127.0.0.2 so the client's unicast SD socket can't steal its own SubscribeAck; `recv_unicast` drains discovery acks before asserting events). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

`SO_REUSEPORT` / socket2's `set_reuse_port` is Unix-only and does not compile on Windows. The Windows CI job compiles the integration suite (`--no-run`); this spine-era test (added post-#130, which had gated its own `set_reuse_port` calls) was the lone unguarded call site — every `set_reuse_port` in `src/` is already `#[cfg(unix)]`-guarded. The test is `#[ignore]`'d and only runs on Linux CI, so the Windows build just needs it to compile. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Forward-ports #130 (released on the 0.7.x line, never ported to the 0.8.0 spine) onto the post-#125 client architecture. The spine's discovery-loop rewrite dropped #130's dual-socket split, so 0.8.0 regressed the false-reboot fix: a sensor's interleaved multicast (~1468) and unicast (~739) SD session counters collided on one SessionTracker key and looked like perpetual reboots. Mechanism (re-applied to the generic TransportFactory / BindDispatch / Spawner surface, not the old raw-socket2 path): - `SocketManager::bind_discovery_unicast_with_transport[_local]`: a receive-only unicast SD socket bound to the *specific* interface IP on the SD port (vs the multicast socket's INADDR_ANY), no group join. "Most-specific bind wins" diverts the sensor's unicast SD here. - `BindDispatch::bind_discovery_unicast`: claims a buffer lease and binds via the Send / Local dispatch (one extra socket-loop buffer per client). - `Inner`: new `discovery_unicast_socket`, best-effort bound in `bind_discovery` (failure is non-fatal — multicast still works), shut down in `unbind_discovery`. The discovery-handling block is extracted to `Self::handle_discovery_datagram` and driven from both the multicast arm (TransportKind::Multicast) and a new unicast arm (TransportKind::Unicast). The spine's SessionTracker already keys by transport, so no engine change there. Tests: port #130's two per-transport SessionTracker regression tests, the dual_socket_splits kernel-behaviour spike, the hermetic combined-SD server test, and the client_server hardening this fix requires (server on 127.0.0.2 so the client's unicast SD socket can't steal its own SubscribeAck; `recv_unicast` drains discovery acks before asserting events). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

`SO_REUSEPORT` / socket2's `set_reuse_port` is Unix-only and does not compile on Windows. The Windows CI job compiles the integration suite (`--no-run`); this spine-era test (added post-#130, which had gated its own `set_reuse_port` calls) was the lone unguarded call site — every `set_reuse_port` in `src/` is already `#[cfg(unix)]`-guarded. The test is `#[ignore]`'d and only runs on Linux CI, so the Windows build just needs it to compile. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…criber test after #134 rebase bf21b2c made the SOME/IP server caps env-configurable but lowered the *defaults* (SUBSCRIBERS_PER_GROUP 16->1, EVENT_GROUPS_CAP 32->4, ServerConfig EVENT_GROUP_IDS_CAP 32->1, ACCEPTED_OFFERS_CAP 16->4) for every build. Those consts are shared with the std/tokio server, so plain std consumers that never inject SIMPLE_SOMEIP_MAX_* were silently capped (e.g. one subscriber per event group), and 11 lib unit tests + the multi-subscriber integration test went red. - Gate the fallback defaults by feature: bare-metal keeps the tight values (the host build injects exact sizes via the env vars — the memory win), std builds keep the historical generous defaults. Env override still applies to both via from_env_or(option_env!(...)). - Expose ServerConfig::SUBSCRIBERS_PER_GROUP_CAP so callers/tests can adapt to the build-time capacity. - from_env_or: fix clippy::pedantic (doc backticks, manual_assert, i32->usize sign-loss via is_ascii_digit) and gate on `server` (its only callers) to drop the dead_code warning on client-only / bare_metal-only builds. Test fixes (rebase artifacts from porting onto #134): - test_multiple_subscribers_receive_events / test_client_send_sd_auto_binds_discovery sent to Ipv4Addr::LOCALHOST while create_server binds SERVER_IP (127.0.0.2, the #130 dual-socket split) -> subscribes hit a dead address, 0 registered. Restore SERVER_IP. Also gate the second-subscriber assertion on SUBSCRIBERS_PER_GROUP_CAP so it passes at the bare-metal default of 1 and exercises the real fan-out under SIMPLE_SOMEIP_MAX_SUBS>=2. - inner.rs: restore the discovery_unicast_socket teardown in unbind_discovery dropped during the rebase. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

JustinKovacich and others added 5 commits June 15, 2026 15:56

Copilot AI review requested due to automatic review settings June 16, 2026 17:10

Copilot started reviewing on behalf of JustinKovacich June 16, 2026 17:11 View session

Copilot AI reviewed Jun 16, 2026

View reviewed changes

Comment thread src/client/socket_manager.rs

Comment thread src/client/socket_manager.rs

Comment thread src/client/socket_manager.rs

JustinKovacich and others added 2 commits June 16, 2026 13:25

Copilot AI review requested due to automatic review settings June 16, 2026 18:01

Copilot started reviewing on behalf of JustinKovacich June 16, 2026 18:02 View session

Copilot AI reviewed Jun 16, 2026

View reviewed changes

Comment thread src/client/socket_manager.rs

Comment thread src/client/socket_manager.rs

JustinKovacich and others added 2 commits June 16, 2026 14:22

Copilot AI review requested due to automatic review settings June 16, 2026 18:54

Copilot started reviewing on behalf of JustinKovacich June 16, 2026 18:55 View session

Copilot AI reviewed Jun 16, 2026

View reviewed changes

Comment thread src/client/socket_manager.rs

Comment thread src/client/socket_manager.rs

Comment thread src/client/socket_manager.rs Outdated

JustinKovacich and others added 2 commits June 16, 2026 16:34

Copilot AI review requested due to automatic review settings June 17, 2026 13:44

Copilot started reviewing on behalf of JustinKovacich June 17, 2026 13:45 View session

Copilot AI reviewed Jun 17, 2026

View reviewed changes

Comment thread tests/client_server.rs Outdated

JustinKovacich and others added 2 commits June 17, 2026 10:25

Copilot AI review requested due to automatic review settings June 17, 2026 14:41

Copilot started reviewing on behalf of JustinKovacich June 17, 2026 14:41 View session

Copilot AI reviewed Jun 17, 2026

View reviewed changes

Comment thread tests/client_server.rs Outdated

JustinKovacich and others added 2 commits June 17, 2026 10:51

Copilot AI review requested due to automatic review settings June 17, 2026 15:11

Copilot started reviewing on behalf of JustinKovacich June 17, 2026 15:11 View session

Copilot AI reviewed Jun 17, 2026

View reviewed changes

zheylmun approved these changes Jun 17, 2026

View reviewed changes

JustinKovacich added this pull request to the merge queue Jun 17, 2026

Merged via the queue into main with commit c17941b Jun 17, 2026
6 checks passed

JustinKovacich deleted the fix/sd-per-transport-session branch June 17, 2026 15:29

JustinKovacich mentioned this pull request Jun 26, 2026

feat: consolidated 0.8.0 stack (forward-ports + #130) → main #138

Merged

This was referenced Jun 26, 2026

feat: forward-port per-source RX demux (v0.7.3) onto 0.8.0 #137

Closed

feat(server): forward-port publish_raw_event_to + subscriber_addresses onto 0.8.0 #135

Closed

JustinKovacich mentioned this pull request Jun 26, 2026

test/ci: restore code coverage (regressed ~96% → ~87% after 0.8.0 consolidation) #139

Open

4 tasks

Uh oh!

Conversation

JustinKovacich commented Jun 16, 2026

What

Verification

CI: Windows portability

Follow-ups (not in this PR)

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Uh oh!

Uh oh!

codecov-commenter commented Jun 16, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

Uh oh!

JustinKovacich commented Jun 17, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

codecov-commenter commented Jun 16, 2026 •

edited

Loading