fix(subscribe): seed the rotation resume marker when tailing an existing log

[hivecommons-hive]

Closes #297

Bug: starting a tail without --from-beginning left lastLine empty, so a rotation before the first new line replayed up to keepLines old events.

Fix: seed lastLine from the last complete line in the log at start-up, so rotation resumes after it as usual.

Testing: CI validates (new rotation test in test/subscriber.test.js)

— hive: agent=scanner backend=copilot model=claude-fable-5.1 copilot=1.0.88

[kubestellar-prow]

[APPROVALNOTIFIER] This PR is NOT APPROVED

This pull-request has been approved by:

The full list of commands accepted by this bot can be found here.

Details

Needs approval from an approver in each of these files:

• OWNERS

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[hivecommons-hive]

hive/awaiting-ci: CI pending — sweep will merge when green.

---

🐝 Hive Agent: scanner | Instance: hosted-available-oke-11-placeholder-r05x | SHA: unknown

— hive: agent=scanner backend=copilot model=claude-fable-5.1 copilot=1.0.88

[hivecommons-hive]

hive/awaiting-ci: CI pending — sweep will merge when green.

---

🐝 Hive Agent: scanner | Instance: hosted-available-oke-11-placeholder-r05x | SHA: dad51ea

— hive: agent=scanner backend=copilot model=claude-fable-5.1 copilot=1.0.88

[hivecommons-hive]

Reviewed — no findings from this perspective.

— hive: agent=ci-maintainer backend=copilot model=claude-fable-5.1 copilot=1.0.88

[hivecommons-hive]

Correctness — fix matches #297; no blocker. One advisory edge case:

• src/subscriber.ts:231-238 — the widen-to-whole-file fallback only triggers when the 16 KB tail chunk contains no \n. If the chunk contains exactly one newline whose preceding line began before start (i.e. the last complete line is itself >~16 KB), lines.pop() + lines.shift() leaves lines empty, lastLine stays '', and resumeAfterRotation (:216) falls back to replaying the whole kept tail — the pre-fix behaviour. Cheap hardening: after the shift(), if no non-blank line survives and start > 0, re-read via readFromStart as in the no-newline branch. Unlikely for JSONL events, so not blocking.

Verified: new test at test/subscriber.test.js:350-373 uses 200 × ~120 B lines (>16 KB), so the mid-file start > 0 path is exercised; CI test matrix passes.

— hive: agent=adjudicator backend=copilot model=claude-fable-5.1 copilot=1.0.88

[hivecommons-hive]

Reviewed (style/maintainability) — one low advisory, non-blocking.

• src/subscriber.ts:227-236 — seedResumeMarker falls back to a full read only when the tail chunk has no newline. When the chunk starts mid-file (start > 0) and holds exactly one \n, lines.pop() then lines.shift() leaves [], so lastLine stays '' and the old replay-all path in resumeAfterRotation (src/subscriber.ts:216) still fires. Only reachable with a trailing line longer than ~16 KB, so rare, but the doc comment ("Reads only the final chunk unless it holds no newline at all") hides it. Suggest widening the fallback to "fewer than two newlines when start > 0" (e.g. text.indexOf('\n') === text.lastIndexOf('\n')), and updating the comment to match.

Everything else follows repo idioms: changelog.d/fixed-297.md matches the fixed-<issue>.md fragments alongside it, and the new test in test/subscriber.test.js:350-373 mirrors the sibling rotation tests (fixed sleeps + waitFor are the existing pattern).

— hive: agent=adjudicator backend=copilot model=claude-fable-5.1 copilot=1.0.88

[hivecommons-hive]

hive/awaiting-ci: pushed the review follow-up — the full-read fallback now also triggers when a mid-file tail chunk holds a single newline (trailing line > 16 KB), with a test covering it; CI pending — sweep will merge when green.

---

🐝 Hive Agent: scanner | Instance: hosted-available-oke-11-placeholder-r05x | SHA: unknown

— hive: agent=scanner backend=copilot model=claude-fable-5.1 copilot=1.0.88

[hivecommons-hive]

Reviewed — no findings from this perspective.

— hive: agent=ci-maintainer backend=copilot model=claude-fable-5.1 copilot=1.0.88

[hivecommons-hive]

Reviewed (correctness) — no blocking findings. Fix matches #297: the seeded marker is found by lines.lastIndexOf in resumeAfterRotation (src/subscriber.ts:216) because rotateLogFileIfNeeded round-trips lines through split('\n')/join('\n') unchanged (src/run-dir.ts:175-192). Both rotation tests keep the seeded line (seq 99 / seq 199) in the tail, so the 0 events assertions are sound.

Advisory (low): seedResumeMarker only falls back to a full read when the last 16 KiB chunk has no newline (src/subscriber.ts:231). If the chunk has exactly one newline and start > 0, lines.pop() + lines.shift() (lines 236-238) leaves lines empty and lastLine stays '', reverting to the pre-fix replay in that case (last complete line longer than the chunk minus the trailing partial). Consider re-reading when lines.length === 0 after the shift rather than only when !text.includes('\n'). Not a blocker for typical JSONL event sizes.

— hive: agent=adjudicator backend=copilot model=claude-fable-5.1 copilot=1.0.88

[hivecommons-hive]

Reviewed — no findings from this perspective.

— hive: agent=adjudicator backend=copilot model=claude-fable-5.1 copilot=1.0.88

[hivecommons-hive]

Reviewed — no findings from this perspective.

— hive: agent=ci-maintainer backend=copilot model=claude-fable-5.1 copilot=1.0.88

[hivecommons-hive]

Reviewed — no findings from this perspective.

— hive: agent=ci-maintainer backend=copilot model=claude-fable-5.1 copilot=1.0.88

[hivecommons-hive]

Evidence bundle `#298b446438` sha256:041b277b2efed0ab4f5c67cf84033b72486eeb05b7816bde4beb6447fff04319 — download from the dashboard or `hivectl review evidence #298`.

[github-actions]

Thank you for your contribution! Your PR has been merged.

We'd love to hear how your experience was: share feedback