Repository navigation
test(net): pin the late better path on lite-07 and lite-06 meshes - #7762
Conversation
A relay that hears a broadcast over a longer path first moves to the shorter one when it arrives. On a lite-07 mesh the move keeps the instance, so lite-06 and lite-07 watchers see an update in place. On a lite-06 mesh each relay mints its own instance, so the watcher sees the live broadcast retracted and announced again; pinned as a residual for routes and announces. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
Warning Review limit reachedYou've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository. Next included review available in 8 minutes. View limit details
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. CodeRabbit Verify
|
kixelated
left a comment
kixelated
left a comment
There was a problem hiding this comment.
Automated review by review (OpenAI)
Reviewed head ca536a1 against base 51b1ff2.
No actionable issue found in the new late-better-path fixture. It observes the publisher and watcher over real mock sessions, drains intermediate announcements from a separate task, checks both lite-06 and lite-07 clients on the lite-07 mesh, and retains the lite-06 mesh restart as an explicit residual case. The quest now points to both outcomes.
Static GitHub-only review; I did not run the simulated-clock tests or the pre-fix probes.
![pullfrog[bot]](https://avatars.githubusercontent.com/in/1768019?s=60&v=4){kind=small}
There was a problem hiding this comment.
Important
The branch conflicts with main: #7676 split the quest file this PR edits. Rebase and move the doc change to quest/m1/cluster-routing/routes/acceptance.md before merging. The test file auto-merges, and both new tests pass at this head.
Reviewed changes
- Late better path fixture —
LATE_BETTER_PATHandlate_better_pathbuild a five-relay mesh where relay 3 first hears the broadcast over the 3-hop0-2-4-3path and then over the 2-hop0-1-3path. - In-place update —
late_better_path_updates_in_placechecks that a lite-06 and a lite-07 client on a lite-07 mesh both seeStart, Updatewhile live and a singleEndon close. - Residual restart —
residual_late_better_path_restarts_on_lite06pinsStart, End, StartthenEndon a lite-06 mesh. - Quest doc — the "Publishing toggles" paragraph now names both tests.
Both new tests pass at ca536a1, and cargo fmt --all -- --check is clean.
⚠️ The quest doc edit targets a file #7676 removed
quest/m1/cluster-routing/routes.md no longer exists on main: #7676 (85b2bef78) renamed it to routes/README.md and moved the "Publishing toggles" paragraph into routes/acceptance.md. The branch forked at 571c956a (before that split), so GitHub reports CONFLICTING, and the merge conflicts in routes/README.md via rename detection.
Technical details
# Retarget the quest doc hunk to routes/acceptance.md
## Affected sites
- `quest/m1/cluster-routing/routes.md:601-606` — deleted on main; the paragraph now lives at `quest/m1/cluster-routing/routes/acceptance.md:44-48`
- `rs/moq-net/tests/integration/mesh_withdraw.rs` — auto-merges; only the residual comment's link changed on main
## Required outcome
- Rebase onto `main`, drop the `routes.md` hunk, and apply the updated paragraph to `routes/acceptance.md`.
- Keep the two new tests; main already points the residual comment at `routes/README.md`.
- Re-run `late_better_path_updates_in_place` and `residual_late_better_path_restarts_on_lite06` after the rebase.deepseek-v4.1-flash-1017 (free via Pullfrog for OSS) ｜ 𝕏
# Conflicts: # quest/m1/cluster-routing/routes/README.md
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
There was a problem hiding this comment.
✅ No new issues found.
Reviewed changes
- Merged
main, retargeting the quest edit toquest/m1/cluster-routing/routes/acceptance.mdand dropping the removedroutes.mdhunk, which clears the prior conflict. - Clarified the
LATE_BETTER_PATHdoc comment with the 3-hop/30 ms and 2-hop/310 ms path timings, applying the earlier suggestion. - Updated the residual comment's link to
routes/README.md.
deepseek-v4.1-flash-1017 (free via Pullfrog for OSS) ｜ 𝕏
Problem
A relay that hears a broadcast first over a longer path moves to the shorter one when it arrives. Before routes carried an instance (moq-dev/moq#7489), that move was an epochless restart, which a lite-06 client receives as a retraction and a fresh announcement while the broadcast stays live. A deployment that meshes its relays on lite-07 and serves lite-06 clients saw this at every relay two hops from a publisher: on a five-relay mesh under steady churn, each such relay toggled about a third of the broadcasts it learned, always as a back-to-back retraction and announcement.
routes asked for this fixture ("Publishing toggles too", decided 2026-10-13), and nothing covered a lite-06 client behind a lite-07 mesh.
Change
Test-only.
mesh_withdraw.rsgains a five-relay fixture where relay 3 hears a broadcast first through two relays (30 ms) and 280 ms later through one (310 ms):late_better_path_updates_in_place: on a lite-07 mesh, lite-06 and lite-07 watchers both seeStart, Updatewhile the broadcast is live and oneEndon close. At80cafb3af(before feat(net)!: routes carry an instance and a per-instance epoch #7489) the lite-06 watcher sawStart, End, Start, and the lite-07 watcherStart, Restart.residual_late_better_path_restarts_on_lite06: on a lite-06 mesh each relay mints its own instance, so the watcher still seesStart, End, Start. Pinned beside the otherresidual_*tests for routes to turn into an update.The routes quest's "Publishing toggles" paragraph, now in
routes/acceptance.mdafter moq-dev/moq#7676 split the quest, names both tests.Impact
No public API or wire change.
Testing
just check origin/mainafter mergingmain: passed (nextest 9701 run, 9701 passed).cargo test -p moq-net --test integration late_better_path: both tests pass after the merge.80cafb3afas probes, giving the sequences quoted above.Review
quest review --base origin/main,gpt-6.1-sol, medium effort, session01a1a35c-2db7-7001-aed3-28c7622b33f2.ca536a18a5c11d0c56456c64e5b756a6d59ef35f: 0 findings.main(moving the quest edit toroutes/acceptance.md) and applying Pullfrog's doc-comment suggestion on the path latencies: new session01a1a37e-e41d-7303-8100-78b10d3edcfb, round 1, head48544b34fb86c4c18cd0450812c5aded4b35fc69: 0 findings.(Written by Claude Opus 5.5)
🤖 Generated with Claude Code