[test][node] Cover join stage-1 with no known right neighbor

[TheP2P (thep2p)]

Summary

• Add test_join_stage1_link_level0_no_right_neighbor_sends_single_link_request: when s's neighbor query resolves z as None, join_stage1_link_level0 never sends a second GetLinkOp and resolves once the single s-side request is applied, leaving the right-side table entry unset.
• Required acceptance-criteria coverage for [Node] Implement: level-0 join linking (Phase 1) #90 (issue [Node] Implement: level-0 join linking (Phase 1) #90's third required test scenario). No production code change: the None-handling branch this exercises already shipped in the orchestration PR earlier in this stack.

Test plan

• cargo fmt --check
• cargo clippy --all-targets -- -D warnings
• cargo clippy --all-targets --all-features -- -D warnings
• cargo test

[Copilot]

Copilot review overview

🟢 Approval recommended

The reviewed test changes are complete and introduce no unresolved issues.

Review effort: Lite
Findings: None

What changed in this PR

Adds regression coverage for stage-1 joining when the predecessor has no right neighbor.

Changes:

• Verifies only one link request is sent.
• Confirms the join completes with the right table entry unset.

File	Description
src/​node/​base_node.rs	Adds the no-right-neighbor stage-1 join test.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[TheP2P (thep2p)]

Claude (@claude) review

[github-actions]

Claude finished TheP2P (@thep2p)'s task —— View job

---

PR Review: Join stage-1, no right neighbor

• Read repo guidelines (no CLAUDE.md present)
• Read the changed test in src/node/base_node.rs
• Cross-check against production join_stage1_link_level0
• Review test correctness and quality

---

Verdict: Approve ✅

This is a clean, well-scoped test-only PR. The new test faithfully exercises the production None-branch and the refactor to poll_until_some is a nice readability win.

What I verified

1. The test mirrors real behavior. In join_stage1_link_level0 (src/node/base_node.rs), when get_neighbor returns None the code takes the branch that only awaits search_link.await? and sends no second link request. The test drives exactly this: RetNeighborOp { neighbor: None }, then asserts link_request_count == 1 and that the right (query_direction) entry stays None. Correct.

2. Direction logic is right. introducer < node_id ⇒ search_direction = Right, query_direction = Left. The test builds introducer = random_identifier_less_than(&node_id) and asserts the search goes Direction::Right and the neighbor query reuses Direction::Right. Matches production (get_neighbor(search_result, search_direction, …)). Good.

3. The poll_until_some refactor is sound. Replacing the hand-rolled loop { … yield_now().await } blocks with poll_until_some(|| …) (and s.zip(m) / .then_some(())) preserves the same bounded-poll semantics while cutting repetition across four existing tests. The helper (src/node/testutil.rs) yields between probes, so it keeps the project's "no wall-clock sleep" discipline.

Minor observations (non-blocking)

• Left-entry assertion via the reply, not a distinct identity. The SetLinkOp reply carries dir: Direction::Left, linked: Some(s_identity), and the test then asserts the left slot holds s_id. This confirms the reply is routed to the correct slot, which is the point of the test — just noting the left-side value comes from the injected reply rather than from production choosing it, so it's a routing assertion, not a value-selection one. That's appropriate for this unit.
• Doc comment polish (commit 35a14e9) reads clearly now and drops the introducer.id() < u.id() jargon in favor of plain language — good improvement.
• Couldn't run cargo fmt/clippy/test locally (command approval not granted in this environment), so I'm relying on CI for the mechanical checks. The code reads as fmt/clippy-clean and the test logic is correct by inspection.

Nice incremental coverage for #90.
• Branch: thep2p/90-join-stage1-no-right-neighbor-test

[github-actions]

Clean, well-scoped test-only PR. The new test faithfully exercises the production None-branch of join_stage1_link_level0 (only search_link is awaited, no second link request), the direction logic matches production (introducer < node_id => search Right, query Left), and the poll_until_some refactor preserves bounded-poll semantics while cutting repetition. Approving. Note: relied on CI for fmt/clippy/test as I couldn't run cargo locally.