ci(reborn): add cargo-llvm-cov integration-tier coverage job (T0-COV)

[henrypark133]

What

Adds a per-PR, informational coverage signal for the Reborn backend (roadmap task T0-COV, docs/reborn/reborn-backend-coverage-roadmap.md). A new Reborn Coverage workflow runs cargo-llvm-cov over the in-process integration-tier test binaries and surfaces a Reborn-crate-scoped line-coverage % plus a per-crate hole list in the PR job summary.

This is what makes the roadmap's "100% int-tier coverage" goal measurable and yields the real hole list to drive the rest of the coverage workstreams.

Measured baseline

Running the job's exact scope locally produced:

Line coverage (Reborn crates): 13.99% — 8351 / 59713 lines

Lowest-covered crates (the hole list): ironclaw_reborn_config, ironclaw_reborn_identity, ironclaw_reborn_traces, ironclaw_webui_v2(_static) at 0%; ironclaw_reborn_composition 18.9%; ironclaw_reborn 25.8%.

How the % surfaces per-PR

The Render Reborn coverage summary step writes the markdown table to $GITHUB_STEP_SUMMARY, so the % and per-crate breakdown appear on the workflow run's summary page for every PR (the repo's existing job-summary convention; the full per-file JSON export is uploaded as the reborn-coverage artifact for offline inspection).

Scope decisions

• Test scope = the in-process int-tier suites per the roadmap's definition: tests/reborn_integration_*.rs + tests/reborn_group_*/. Discovered dynamically (scripts/ci/reborn-coverage-int-tier-tests.sh) so new suites auto-enroll.
• Features = default (postgres+libsql+html-to-markdown+tui), matching how scripts/ci/run-reborn-root-partition.sh runs these same suites. No live Postgres needed (suites skip Postgres backends when DATABASE_URL is unset; backend_matrix stays libsql-only).
• Combined cargo llvm-cov --workspace --test … --json one-shot is deliberate: the standalone report subcommand cannot take --workspace, so a split --no-report/report would scope the report to only the root package's src/ and report 0% for the linked Reborn crates. The --workspace combined form surfaces the member-crate files (verified: 0% → 13.99%).
• Non-gating: no coverage threshold, no PR fail — matches "do not gate on coverage yet".
• Reuses coverage.yml's pinned action SHAs / install action; LLM-neutralizing env mirrors reborn-tests.yml; crate-scope regex mirrors the reborn-tests.yml package allowlist (prefix-match for reborn_*/product_*/webui_v2_*, exact-match for the four single crates).

Verification

• shellcheck + actionlint clean; YAML validated.
• Local run: all 13 int-tier binaries pass; baseline % reproduced via scripts/ci/reborn-coverage-summary.sh.
• No .rs / Cargo.toml / Cargo.lock changes — cargo build/clippy state is identical to the green main base.
• Reviewed via thermo-nuclear + code-review (approach/local-patterns/maintainability) loops; findings addressed.

Added since open

• Sticky PR comment (scripts/ci/reborn-coverage-comment.sh): upserts one marker'd comment so the %/hole-list lives in the conversation, not just the job summary. PR-only, continue-on-error: true — a fork's read-only token 403 can't red the check. Per-crate table folded into <details> to keep it compact.
• Cache hygiene (save-if gated to main): PRs restore a warm main-seeded instrumented cache instead of each PR writing a branch-scoped copy that churns the ~10 GB LRU.
• Review fixes: iterate all llvm-cov data[] datasets; GH_TOKEN fast-fail guard; timeout-minutes: 180 runaway backstop.
• Regression tests (scripts/ci/test-reborn-coverage.sh, 43 cases): summary report/--zero-crates, comment POST/PATCH upsert via fake gh, int-tier discovery — mirrors the test-classify-test-scope.sh precedent.

Change Type

CI / tooling (new workflow + shell helpers + tests). No production .rs, Cargo.toml, or Cargo.lock changes.

Linked Issue

Roadmap task T0-COV (docs/reborn/reborn-backend-coverage-roadmap.md). No separate tracking issue.

Validation

• shellcheck + actionlint clean; YAML validated.
• scripts/ci/test-reborn-coverage.sh: 43/43 pass (mutation-checked non-vacuous).
• Live: the Reborn Coverage job ran on this PR and posted the sticky comment (real ~13.9% run) — confirms the workflow, permissions, and upsert path end-to-end.
• No .rs/Cargo.* changes — cargo build/clippy state identical to the green main base.

Security / Blast Radius

• Trigger is pull_request (not pull_request_target) — no privileged-context checkout of untrusted code. Minimal token scope: contents: read + pull-requests: write (issues-comment upsert).
• Fork PRs get a read-only token; the comment step is continue-on-error so its 403 is non-fatal. No secrets consumed (LLM env is neutralized).
• Purely informational: no gate, cannot block or alter any other check.

Rollback Plan

Self-contained and additive. Revert the workflow + four scripts/ci/reborn-coverage-*.sh files (or disable via .github/workflows/reborn-coverage.yml) — nothing else depends on them, no state/migrations, no effect on other jobs. If the sticky comment misbehaves, deleting the comment step alone leaves the job-summary signal intact.

Review track

Highest review-risk track (CI workflow change). Reviewed via thermo-nuclear + multi-agent code-review loops; valid findings fixed, non-issues documented in the triage comment below.

🤖 Generated with Claude Code

[coderabbitai]

Note

Reviews paused

It looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the reviews.auto_review.auto_pause_after_reviewed_commits setting.

Use the following commands to manage reviews:

• @coderabbitai resume to resume automatic reviews.
• @coderabbitai review to trigger a single review.

Use the checkboxes below for quick actions:

• ▶️ Resume reviews
• 🔍 Trigger review

📝 Walkthrough

Walkthrough

A new coverage workflow runs Reborn integration-tier tests with cargo llvm-cov, exports JSON, renders a coverage summary, and posts a sticky PR comment. Two Bash scripts discover test targets and format Reborn crate coverage output, with regression tests for the helper scripts.

Changes

Reborn Coverage CI Workflow

Layer / File(s)	Summary
Workflow wiring and execution .github/workflows/reborn-coverage.yml	Adds the workflow triggers, permissions, concurrency, runner setup, test execution, summary rendering, PR comment step, and unconditional JSON artifact upload.
Integration-tier test discovery scripts/ci/reborn-coverage-int-tier-tests.sh	Discovers reborn_integration_*.rs files and reborn_group_* directories under tests/, deduplicates suite names, and emits --test <name> arguments.
Coverage summary and sticky comment scripts/ci/reborn-coverage-summary.sh, scripts/ci/reborn-coverage-comment.sh	Computes overall and per-crate Reborn coverage from llvm-cov JSON, renders Markdown output, and upserts a sticky PR comment with the summary plus zero-coverage crate callout.
Coverage helper regression tests scripts/ci/test-reborn-coverage.sh	Exercises summary, zero-crate, sticky-comment, and discovery helpers against fixture data and temporary test layouts.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~25 minutes

Possibly related issues

• ci(reborn): unify Reborn crate-family allowlist across reborn-tests.yml and reborn-coverage-summary.sh #5477 — Adds the Reborn coverage summary script and the same crate-family allowlist/regex it describes.

Poem

🐰 I hop through tests with coverage bright,
JSON blooms in morning light.
A sticky note on the PR wall,
Keeps little crate gaps in view for all.
The carrot of CI gleams so neat,
And bash drums softly: thump-thump-beat.

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 22.22% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly names the new Reborn cargo-llvm-cov integration-tier coverage workflow and matches the main change.
Description check	✅ Passed	The description covers the workflow, scope, validation, security, rollback, and linked issue, though it does not mirror every template heading exactly.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

---

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.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[gemini-code-assist]

Code Review

This pull request introduces two new CI helper scripts: reborn-coverage-int-tier-tests.sh to dynamically discover and list Reborn integration-tier test binaries, and reborn-coverage-summary.sh to filter and format coverage data from a cargo-llvm-cov JSON export into a Markdown summary. Feedback was provided regarding a potential crash in the jq parsing logic of reborn-coverage-summary.sh when the data array is empty, suggesting a more robust and idiomatic indexing pattern.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[railway-app]

🚅 Deployed to the ironclaw-pr-5430 environment in ironclaw-ci-preview

Service	Status	Web	Updated (UTC)
ironclaw	✅ Success (View Logs)	Web	Jul 1, 2026 at 6:37 am

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In @.github/workflows/reborn-coverage.yml:
- Around line 21-46: The workflow block is missing an explicit job timeout in
the Reborn Coverage CI, so add a bounded timeout to the job that runs this
workflow to prevent runaway or hung runs from consuming a runner for the default
360 minutes. Update the job definition associated with the Reborn Coverage
workflow so the coverage job has a generous but finite limit (for example,
90–120 minutes) while keeping the existing trigger, permissions, and concurrency
settings unchanged.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Pro Plus

Run ID: db2f1147-4a1f-4c48-9282-c4830de53e51

📥 Commits

Reviewing files that changed from the base of the PR and between 0abb631 and ca39cda.

📒 Files selected for processing (4)

• .github/workflows/reborn-coverage.yml
• scripts/ci/reborn-coverage-comment.sh
• scripts/ci/reborn-coverage-int-tier-tests.sh
• scripts/ci/reborn-coverage-summary.sh

💤 Files with no reviewable changes (3)

• scripts/ci/reborn-coverage-comment.sh
• scripts/ci/reborn-coverage-int-tier-tests.sh
• scripts/ci/reborn-coverage-summary.sh

[Copilot]

Pull request overview

Adds a new CI workflow that runs cargo llvm-cov against the Reborn integration-tier (in-process) test binaries and publishes a Reborn-crate-scoped line coverage percentage plus a per-crate “hole list” (job summary + optional sticky PR comment) to make the backend coverage roadmap measurable.

Changes:

• Introduces Reborn Coverage GitHub Actions workflow to generate and upload an llvm-cov JSON export and render a PR-visible markdown summary.
• Adds scripts to (a) dynamically discover integration-tier test binaries and (b) compute Reborn-crate-scoped aggregate + per-crate coverage from the JSON export.
• Adds a script to upsert a sticky PR comment containing the same coverage summary for better visibility.

Reviewed changes

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

File	Description
scripts/ci/reborn-coverage-summary.sh	Renders a Reborn-scoped markdown summary + per-crate breakdown from cargo llvm-cov --json output.
scripts/ci/reborn-coverage-int-tier-tests.sh	Discovers integration-tier test binaries and emits --test <name> args for coverage execution.
scripts/ci/reborn-coverage-comment.sh	Upserts a sticky PR comment with the rendered summary (visibility-only).
.github/workflows/reborn-coverage.yml	New workflow wiring toolchain setup, coverage generation, summary rendering, sticky comment posting, and artifact upload.

---

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

[Copilot]

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

[github-actions]

⚠️ 5 Reborn crate(s) have 0 int-tier coverage (target: 0) — ironclaw_reborn_config, ironclaw_reborn_identity, ironclaw_reborn_traces, ironclaw_webui_v2, ironclaw_webui_v2_static

Reborn integration-tier coverage

Line coverage (Reborn crates): 14.2% — 8649 / 60891 lines

Per-crate breakdown (12 crates, lowest-covered first)

Crate	Line %	Covered / Total
ironclaw_reborn_config	0%	0 / 1142
ironclaw_reborn_identity	0%	0 / 237
ironclaw_reborn_traces	0%	0 / 6662
ironclaw_webui_v2	0%	0 / 2653
ironclaw_webui_v2_static	0%	0 / 110
ironclaw_reborn_event_store	0.73%	6 / 825
ironclaw_product_adapter_registry	5.62%	25 / 445
ironclaw_product_workflow	6.22%	562 / 9036
ironclaw_product_adapters	12.74%	283 / 2221
ironclaw_reborn_composition	19.59%	5954 / 30395
ironclaw_reborn	25.3%	1809 / 7151
ironclaw_product_context	71.43%	10 / 14

This signal is informational: coverage never gates the PR — not the percentage, not the per-crate holes, not the 0-coverage callout.

[henrypark133]

Code Review (multi-agent)

Intent: Add an informational per-PR Reborn integration-tier coverage workflow that reports cargo-llvm-cov line coverage and per-crate coverage holes.

Stats: 7 findings (from 9 raw, 7 after duplicate/live-thread suppression) across 4 files. Reviewers run: security, bugs, performance, tests, conventions, local-patterns, maintainability, approach. Reviewers failed: none. Body-only: 1.

Existing unresolved threads already cover the issues: write permission, jq .data[0]/grouping correctness, GH token validation, and job timeout surfaces, so this review does not repost those.

performance

1. Medium Coverage clean deletes the restored instrumented cache (.github/workflows/reborn-coverage.yml:103-109, confidence 88) — anchor: .github/workflows/reborn-coverage.yml:103
   The job restores a main-seeded instrumented cache, but then runs cargo llvm-cov clean --workspace immediately before coverage. cargo llvm-cov clean removes prior coverage build artifacts; for this workflow that means PRs can discard the expensive restored instrumented target and pay the cold rebuild cost the cache comment says it is avoiding.

tests

1. Medium Coverage aggregation has no committed edge-case test (scripts/ci/reborn-coverage-summary.sh:49-103, confidence 75) — anchor: scripts/ci/reborn-coverage-summary.sh:49
   The new summary CLI owns the Reborn crate filter, aggregate percentage, per-crate sorting, no-data fallback, and --zero-crates output, but nothing exercises mixed Reborn/non-Reborn llvm-cov JSON, empty/no-match JSON, or zero-covered crates. A jq or regex change can silently make the PR-visible coverage signal report the wrong scope or omit breadth holes.
2. Medium Sticky comment upsert path is untested (scripts/ci/reborn-coverage-comment.sh:69-78, confidence 75) — anchor: scripts/ci/reborn-coverage-comment.sh:69
   The new PR-visible comment path branches between PATCHing an existing marker comment and POSTing a new one via gh api --paginate, but no test stubs gh to verify either branch. A marker, pagination, or endpoint regression would duplicate comments or stop updating the sticky coverage summary without a local signal.
3. Low Dynamic test discovery lacks empty and multi-suite coverage (scripts/ci/reborn-coverage-int-tier-tests.sh:19-35, confidence 75) — anchor: scripts/ci/reborn-coverage-int-tier-tests.sh:19
   The workflow relies on this script to discover every Reborn integration-tier binary and to fail when none exist, but there is no test covering an empty tests tree, single-file suites, group-directory suites, and sorted/deduplicated output. A pattern or maxdepth regression could silently drop test binaries from coverage.

maintainability

1. Medium Coverage scope duplicates the Reborn crate family allowlist (scripts/ci/reborn-coverage-summary.sh:47-47, confidence 80) — anchor: scripts/ci/reborn-coverage-summary.sh:47
   The summary script hardcodes a Reborn crate-family regex that is intended to mirror the package allowlist in reborn-tests.yml. Keeping those scopes as separate hand-maintained filters can let coverage percentage and hole-list reporting drift away from the test surface it claims to measure.
2. Low Zero-coverage callout is split across two scripts (scripts/ci/reborn-coverage-summary.sh:27-31, confidence 75) — anchor: scripts/ci/reborn-coverage-summary.sh:27
   reborn-coverage-summary.sh has a second --zero-crates mode only so the comment script can call it again and stitch one extra callout around the normal report. That makes the PR comment and job summary shapes differ and spreads report composition across two files.

conventions

1. Low CI workflow change is missing a documented rollback plan (.github/workflows/reborn-coverage.yml:1-1, confidence 75) — anchor: CONTRIBUTING.md:121 (no diff position — body only)
   The PR adds a new CI workflow, and the repo contribution rules classify CI workflow changes as needing a documented rollback plan. The PR body documents scope and verification, but not what to do if the workflow overloads CI or emits bad reports.

[henrypark133]

Code Review (multi-agent)

Intent: Add non-gating Reborn integration-tier coverage reporting to CI with workflow, shell helpers, and PR comment summaries.

Stats: 2 findings (from 10 raw, 2 after dedup/filter) across 2 files. Reviewers run: security, bugs, performance, tests, conventions, local-patterns, maintainability, approach. Reviewers failed: none. Body-only: 0

Performance / Concurrency

1. Medium Concurrency key can cancel unrelated fork PR coverage runs (.github/workflows/reborn-coverage.yml:36, confidence 88) — anchor: .github/workflows/pr-label-classify.yml:13
   For pull_request events, github.head_ref is only the source branch name, so two different fork PRs with the same branch name share one coverage concurrency group.

Conventions

1. Low Do not cite a missing coverage roadmap as the scope source (scripts/ci/reborn-coverage-int-tier-tests.sh:6, confidence 75) — anchor: AGENTS.md:78
   The helper and workflow cite docs/reborn/reborn-backend-coverage-roadmap.md, but that file is absent from this PR head.

Suppressed as already addressed or not fresh: the issues: write concern is already resolved in live review threads and contradicted by the posted sticky comment; the allowlist duplication is explicitly deferred in the PR body; the fork-token concern is limited by pull_request read-only fork tokens and first-time-contributor approval; low-value missing-test suggestions were not posted.

[Copilot]

Pull request overview

Copilot reviewed 5 out of 5 changed files in this pull request and generated no new comments.

[henrypark133]

Code Review (multi-agent)

Intent: Add an informational Reborn integration-tier coverage workflow using cargo-llvm-cov with summaries, sticky PR comments, helper scripts, and tests.
Stats: 5 findings (from 7 raw, 5 after dedup/manual validation) across 3 files. Reviewers run: security, bugs, performance, tests, conventions, local-patterns, maintainability, approach. Reviewers failed: none. Body-only: 0

Security

1. High PR code receives a write-scoped GitHub token (.github/workflows/reborn-coverage.yml:125-131, confidence 75) — anchor: .github/workflows/reborn-coverage.yml:129
   This pull_request workflow checks out and executes the PR-controlled version of scripts/ci/reborn-coverage-comment.sh, then passes it github.token while the job grants pull-requests: write. GitHub downgrades fork PR tokens to read-only, but same-repo PRs use the configured write scope, so a modified PR script can use the bot token to mutate PR conversation state instead of only publishing the coverage summary.

Approach

1. Medium Informational coverage still fails PR checks (.github/workflows/reborn-coverage.yml:94-114, confidence 75) — anchor: .github/workflows/reborn-coverage.yml:96
   The workflow and PR describe this as informational/non-gating, but the pull_request job runs discovery and cargo llvm-cov under set -e in a normal check. Any coverage tooling, discovery, or test failure turns Reborn Coverage red on the PR, duplicating the existing Reborn test gate rather than merely surfacing a coverage signal.

Tests

1. Medium Crate allowlist boundary cases are not covered (scripts/ci/reborn-coverage-summary.sh:43-47, confidence 75) — anchor: scripts/ci/reborn-coverage-summary.sh:47
   The regex intentionally prefix-matches the reborn/product/webui_v2 families while exact-matching single crates such as ironclaw_architecture and the v2 adapters, but the regression fixtures only exercise family prefixes and non-Reborn exclusion. A future regex change could include lookalikes like ironclaw_architecture_extra or drop an exact-match adapter without failing the suite. Also flagged by: maintainability/Low.
2. Low Summary missing-file branch is untested (scripts/ci/reborn-coverage-summary.sh:35-37, confidence 100) — anchor: scripts/ci/reborn-coverage-summary.sh:35
   The summary helper has an explicit fast-fail path for a nonexistent coverage JSON file, but the regression suite covers valid, empty, and non-Reborn JSON fixtures only. A regression could drop the clear error or non-zero exit without failing tests.
3. Low Comment helper missing-file branch is untested (scripts/ci/reborn-coverage-comment.sh:28-30, confidence 100) — anchor: scripts/ci/reborn-coverage-comment.sh:28
   The sticky-comment helper rejects a missing coverage JSON file before reading GitHub environment or calling gh, but the tests only cover env guards and POST/PATCH behavior. That pre-mutation failure path can regress independently.

Dropped During Validation

• Dropped the performance JSON-size item because the PR intentionally uploads the full JSON artifact for offline inspection.
• Dropped the maintainability allowlist-duplication item as already covered by a resolved prior thread and tracked separately; retained the narrower test-boundary gap above.

[henrypark133]

Code Review (multi-agent)

Intent: Add a non-gating Reborn coverage workflow that measures integration-tier test coverage and posts per-crate hole lists in PR summaries/comments.
Stats: 1 finding (from 2 raw, 1 after dedup/reply validation) across 1 file. Reviewers run: security, bugs, performance, tests, conventions, local-patterns, maintainability, approach. Reviewers failed: none. Body-only: 0

Tests

1. Medium Malformed coverage JSON lacks caller-level no-mutation coverage (scripts/ci/reborn-coverage-comment.sh:44-44, confidence 75) — anchor: scripts/ci/reborn-coverage-comment.sh:44
   reborn-coverage-comment.sh renders the coverage JSON before it POSTs or PATCHes the sticky PR comment, so malformed existing JSON should fail before any GitHub mutation. The current tests cover missing files and env guards, but not an existing malformed JSON file, so they do not lock in that jq/render failures abort before the gh mutation path.

Prior Reply Notes

• Accepted the write-scoped token reply as sound for this PR: the workflow uses pull_request, fork tokens remain read-only, and same-repo PR authors already have collaborator push rights. I did not repost that item.
• Accepted the non-gating reply as sound: API/ruleset checks did not show Reborn Coverage as a required status check, so a red coverage job is visible but not merge-gating. I did not repost that item.
• Verified the prior test gaps are fixed in 113c5d09: A6 covers allowlist exact-vs-prefix boundaries, A7 covers missing summary JSON, and C7 covers missing comment JSON with no fake-gh mutation.
• Suppressed the cache-clean finding as already covered by a resolved thread with a sound reply: cargo llvm-cov clean --workspace preserves dependency artifacts while avoiding stale workspace coverage; the suggested --profraw-only/--no-clean path risks stale coverage for changed workspace code.

[Copilot]

Pull request overview

Copilot reviewed 5 out of 5 changed files in this pull request and generated 1 comment.