[d-morrison]

Moves the latex-macros pin to 16d943c (macros#279 deprecates the h-family aliases, macros#280 prints the prediction function as f, and macros#282 adds the RSS, TSS and training-count macros).

Part of ai-config#4673. The pin change is the whole diff.

The link-checker check fails on two outside hosts that this diff does not touch. It needs a lychee exclusion in a separate infra PR.

🤖 Generated with Claude Code

https://claude.ai/code/session_01V6MvQvUgPePB2ndiLAgpHs

[d-morrison]

The link-checker / link-checker failure is not from this PR. The diff moves only the latex-macros pin. The job reports 0 errors and 5 timeouts, all on outside hosts that this diff doesn't touch: www-cs-students.stanford.edu/~blynn/gitmagic (three places) and www.caut.ca (two places). No fix exists yet. I re-ran the failed job once. If the same hosts time out again, the fix is a lychee exclusion in a separate infra PR, which needs Ezra's typed word.

---

Generated by Claude Code

[github-actions]

PR Preview Action v1.8.1
🚀 Preview available at: https://Morrison-Lab.github.io/rme/pr-preview/pr-1404/

_{Built to branch gh-pages at 2026-10-23 04:17 UTC.}

[github-actions]

Claude finished review — View run

---

Re-verified at the exact head SHA 4104600b51f32af2f862ec204e091833e6518a8a (unchanged from my last check): 20 checks pass, 0 fail, 3 pending (build / build, check-dois / check-dois, and review / claude-review — this very review job). link-checker / link-checker itself shows SUCCESS in this fresh query, so that specific claim stands, but I should not have called the PR broadly "green" while three checks remain pending — correcting that below.

Review: Bump latex-macros to the prediction-function and RSS/TSS release (#1404)

Diff scope: Net diff against main is a single-line submodule pointer bump — latex-macros moves 685784f1c → 16d943c78. No .qmd/.R/config files are touched in the rme repo itself.

What changed since the last review round

Per the computed range 2b6bb28c..064e24d9: only merges of main into the PR branch (by the author at 4104600, and this run's auto-merge-for-review commit 064e24d), bringing in #1406's lychee.toml exclusion of three flaky outside hosts. No new substantive commits were added to the PR branch itself.

Verification performed

• Pin accuracy, re-verified: 16d943c78180a813768fbd3bfe8234e5990795af is the submodule's current HEAD. After un-shallowing the submodule checkout and walking its history, all three referenced upstream PRs are confirmed ancestors of the pinned commit: macros#279 (hat-family deprecation), macros#280 (prediction function as f), and macros#282 (RSS/TSS/training-count macros) — all independently confirmed MERGED via gh pr view. The PR title now accurately reflects the final pin scope, resolving the "stale title" note from the prior round.
• Non-breaking deprecation, re-confirmed: macros#279's commit (564e645) only touches interpretations.tsv (lint data) and notation-choices.qmd (prose) — no macro in macros.qmd was removed or redefined, so existing \h...-prefixed usages across the book keep rendering, consistent with CLAUDE.md's note that the legacy \h... family "stays supported for existing content."
• No content changed in this PR: git diff --stat between the PR's merge-base and head, excluding the latex-macros submodule itself, is empty — the entire net diff is the one pointer bump, so no macro-opportunity check applies and no CLAUDE.md pre-commit checklist item is triggered.
• CI, current state at head 4104600b: 20 pass, 0 fail, 3 pending. The previously-flagged failing link-checker check is SUCCESS as of this exact head (the #1406 lychee.toml exclusion is present via the merge). Still pending: build / build (started ~2 min before last check, not yet at the 5–10 min "investigate" threshold), check-dois / check-dois, and review / claude-review (this job). check-equation-renders — the job that failed twice in prior rounds with "Execution context was destroyed" — is triggered only via workflow_run after build / build completes, so its absence from the list is expected, not a regression; it has not run yet on this head and still needs to be watched once it does.
• Minor, non-blocking: The PR body's sentence "The link-checker check fails on two outside hosts... It needs a lychee exclusion in a separate infra PR" is stale — that infra fix (#1406) already merged to main and is already present via this branch's merge commits. Worth a quick description edit, not a functional defect.

Findings

None that block merge. No new issues introduced since the last review round; both previously-flagged items (failing link-checker, stale PR title) are resolved as of head 4104600b.

Verdict

Ready for merge pending CI completion — at head 4104600b51f32af2f862ec204e091833e6518a8a: 20 checks pass, 0 fail, 3 pending (build / build, check-dois, and this review job). No blocking issues in the diff itself; recommend waiting for build / build and the downstream check-equation-renders run to complete before merging, given that job's two prior flaky failures.

Structured Review Data (JSON)

{
  "schema_version": "1.1",
  "reviewer": "claude",
  "commit_sha": "4104600b51f32af2f862ec204e091833e6518a8a",
  "verdict": "CLEAN",
  "findings": [],
  "detailed_assessment": "The latex-macros pointer resolves exactly to macros#280's merge commit with macros#279 and macros#282 confirmed as ancestors, and no rme .qmd file or config changed outside the submodule pointer itself.",
  "holistic_assessment": "At head 4104600b CI shows 20 pass, 0 fail, 3 pending including build and the downstream check-equation-renders job that flaked twice before, so merge readiness depends on those completing rather than any defect in this diff."
}

Reviewed commit: 4104600

Reviewed commit: 4104600

[github-actions]

💰 Cost: $0.5741 (review) — run