JSONbored · JSONbored · Jul 1, 2026 · Jul 1, 2026
@@ -5,7 +5,7 @@ description: >-
   JSONbored/gittensory repo — picking/validating an issue, implementing a change, writing tests
   that pass Codecov, running the local CI gate, predicting the gittensory gate, and formatting the
   commit + PR. gittensory reviews PRs ONE-SHOT via the gittensory gate (a GitHub App / CI) plus a
-  strict CI suite with Codecov (97% patch coverage, hard); there is no review back-and-forth, so a
+  strict CI suite with Codecov (99% patch coverage, hard); there is no review back-and-forth, so a
   PR must be correct, fully tested, house-style-compliant, and green before it is pushed. Invoke
   for any "contribute to / open a PR against / fix a bug in / add a feature to gittensory" task.
 ---
@@ -49,7 +49,7 @@ opening a **fresh** PR. This is the entire reason to get it right before you pus
 
 ## The non-negotiables (read once, hold throughout)
 
-1. **97%+ patch coverage is a HARD wall.** Codecov `codecov/patch` has `target: 97%, threshold: 0%`
+1. **99% patch coverage is a HARD wall.** Codecov `codecov/patch` has `target: 99%, threshold: 0%`
    — *zero slack* — and it counts **BRANCH** coverage, not just lines. **Aim for 100% on every line
    you change, including invariants and a regression test for any bug you fix.** Only `src/**` counts;
    `apps/**`, `test/**`, `scripts/**`, and `src/env.d.ts` are ignored by Codecov. (So a UI-only change
@@ -183,8 +183,8 @@ This is where most PRs fail Codecov. The bar is **every changed line AND every c
   so a single shard under-reports).
 - **Find the uncovered branch.** In the v8 text report, read the **% Branch** column and the
   **Uncovered Line #s** for your changed file — a line at 100% lines but <100% branch has an un-taken
-  `??`/ternary/`&&` side; add that case. Aim for **≥98% branch on your diff locally** so normal CI
-  variance never drops you under the 97% wall.
+  `??`/ternary/`&&` side; add that case. Aim for **100% branch on your diff locally** so normal CI
+  variance never drops you under the 99% wall.
 
 ---
 

@@ -47,8 +47,8 @@ workflow in this repo. There is **no** root-level Prettier gate — Prettier is
 
 ## 2. Codecov — the real coverage gate (`codecov.yml`)
 
-- **`codecov/patch`: `target: 97%`, `threshold: 0%`, `if_ci_failed: error`.** Every line your PR
-  changes must be ≥97% covered. With 0% threshold and small diffs, **one uncovered branch can fail it.**
+- **`codecov/patch`: `target: 99%`, `threshold: 0%`, `if_ci_failed: error`, `only_pulls: true`.**
+  Every line your PR changes must be ≥99% covered. With 0% threshold and small diffs, **one uncovered branch can fail it.**
 - **It counts BRANCH coverage** (v8 → lcov `BRDA`). A changed line whose branches are only partially
   exercised counts against you. "100% lines" ≠ "100% branches."
 - **`codecov/project`: `informational: true`** — a trend, never blocks.
@@ -168,7 +168,7 @@ Date("2026-05-28T00:00:00Z"))` then `vi.useRealTimers()`.
 or the `test:unit` / `test:integration` scripts. Verify before pushing with the full unsharded
 `npm run test:coverage`. **Find a partial branch:** read the v8 text report's **% Branch** column and
 the **Uncovered Line #s** for your changed file (or open `coverage/lcov-report/…` / `coverage/lcov.info`)
-— a line at 100% lines but <100% branch has an un-taken side. Aim ≥98% branch on the diff locally.
+— a line at 100% lines but <100% branch has an un-taken side. Aim for 100% branch on the diff locally.
 
 **Branch coverage — the rule that fails most PRs.** Each of `if/else`, `? :`, `&&`, `||`, and `??`
 is two branches; exercise **both**.

@@ -14,7 +14,7 @@
 - [ ] `git diff --check`
 - [ ] `npm run actionlint`
 - [ ] `npm run typecheck`
-- [ ] `npm run test:coverage` locally; **`codecov/patch` requires ≥97% coverage of the lines AND branches you changed** (aim for **98%+** on your diff so CI variance does not fail near the threshold). Global coverage is a non-blocking trend with a loose 90% backstop, not the gate.
+- [ ] `npm run test:coverage` locally; **`codecov/patch` requires ≥99% coverage of the lines AND branches you changed** (aim for **100%** on your diff so CI variance does not fail near the threshold). Global coverage is a non-blocking trend with a loose 90% backstop, not the gate.
 - [ ] `npm run test:workers`
 - [ ] `npm run build:mcp`
 - [ ] `npm run test:mcp-pack`

@@ -129,6 +129,13 @@ jobs:
           echo "::error title=Tests::The backend test coverage suite failed."
           echo "Coverage itself is gated by Codecov on changed lines (codecov/patch), computed from the complete lcov generated by this job."
           echo "Reproduce locally with: 'npm run test:coverage'."
+      - name: Verify coverage report exists
+        if: ${{ success() && (github.event_name == 'push' || needs.changes.outputs.backend == 'true') }}
+        run: |
+          if [ ! -s coverage/lcov.info ]; then
+            echo "::error title=Coverage::coverage/lcov.info is missing or empty"
+            exit 1
+          fi
       - name: Upload coverage to Codecov
         if: ${{ success() && (github.event_name == 'push' || needs.changes.outputs.backend == 'true') }}
         uses: codecov/codecov-action@fb8b3582c8e4def4969c97caa2f19720cb33a72f # v7.0.0
@@ -139,7 +146,12 @@ jobs:
           override_branch: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.ref || github.ref_name }}
           override_commit: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
           override_pr: ${{ github.event_name == 'pull_request' && github.event.pull_request.number || '' }}
-          fail_ci_if_error: false
+          # Coverage upload is part of the hard gate: upload or service errors
+          # should fail CI instead of allowing a PR to merge without patch data.
+          fail_ci_if_error: true
+      # Test results are useful Codecov annotations, not the coverage gate. Keep
+      # their upload non-blocking so a JUnit ingestion hiccup does not fail CI
+      # after the tests and hard coverage upload have already passed.
       - name: Upload Vitest results to Codecov
         if: ${{ !cancelled() && (github.event_name == 'push' || needs.changes.outputs.backend == 'true') }}
         uses: codecov/codecov-action@fb8b3582c8e4def4969c97caa2f19720cb33a72f # v7.0.0

@@ -18,7 +18,7 @@ process evolves — edits to those files improve both Claude Code and Codex.
    (including `codecov/patch`), a gate failure, a base conflict, or an ineligible linked issue is
    **auto-CLOSED, one-shot** (recovery = open a fresh PR). Green CI + passing gate + mergeable-clean +
    a valid linked issue is **auto-MERGED**. So make it perfect before you push.
-2. **97% patch coverage, branch-counted.** Aim for **100% of every changed line *and branch*** — test
+2. **99% patch coverage, branch-counted.** Aim for **100% of every changed line *and branch*** — test
    both sides of every `??` / ternary / `&&` (a `SUM()` can return `NULL`, so the nullish arm is real),
    plus invariant tests and a regression test for every fix. Only `src/**` is measured by Codecov.
 3. **The whole local gate must be green:** `npm run test:ci` (+ `npm audit --audit-level=moderate`).

@@ -1,9 +1,9 @@
 # Codecov configuration.
 #
-# The real coverage gate is `patch`: every line a PR changes must be >=97%
-# covered. Patch coverage depends only on the PR's own diff, so merging one PR
-# can never push another below the bar -- this is what kills the cross-PR churn
-# the old global vitest threshold caused.
+# The real coverage gate is `patch`: changed lines and branches in a PR must be
+# >=99% covered. Patch coverage depends only on the PR's own diff, so merging
+# one PR can never push another below the bar -- this is what kills the cross-PR
+# churn the old global vitest threshold caused.
 #
 # `project` (whole-repo total) is informational only: it is reported as a trend
 # but never blocks a merge. vitest keeps a loose 90% local backstop separately.
@@ -15,10 +15,12 @@ coverage:
   status:
     patch:
       default:
-        target: 97%
+        target: 99%
         threshold: 0%
-        # Only fail if the PR actually changes coverable lines.
+        # Keep the patch status in error if the upstream CI run failed.
         if_ci_failed: error
+        # Suppress codecov/patch on main pushes; uploads still maintain history.
+        only_pulls: true
     project:
       default:
         informational: true

@@ -0,0 +1,69 @@
+import { readFileSync } from "node:fs";
+import { parse } from "yaml";
+
+function readYaml(path: string): Record<string, unknown> {
+  return record(parse(readFileSync(path, "utf8")), path);
+}
+
+function record(value: unknown, label: string): Record<string, unknown> {
+  if (!value || typeof value !== "object" || Array.isArray(value)) {
+    throw new Error(`${label} must be an object`);
+  }
+  return value as Record<string, unknown>;
+}
+
+function recordArray(value: unknown, label: string): Array<Record<string, unknown>> {
+  if (!Array.isArray(value)) throw new Error(`${label} must be an array`);
+  return value.map((entry, index) => record(entry, `${label}[${index}]`));
+}
+
+function nestedRecord(source: Record<string, unknown>, path: string[]): Record<string, unknown> {
+  return path.reduce((current, key) => record(current[key], path.join(".")), source);
+}
+
+describe("Codecov policy", () => {
+  it("keeps patch coverage strict and PR-scoped", () => {
+    const config = readYaml("codecov.yml");
+    const patch = nestedRecord(config, ["coverage", "status", "patch", "default"]);
+    const project = nestedRecord(config, ["coverage", "status", "project", "default"]);
+
+    expect(patch.target).toBe("99%");
+    expect(patch.threshold).toBe("0%");
+    expect(patch.if_ci_failed).toBe("error");
+    expect(patch.only_pulls).toBe(true);
+    expect(project.informational).toBe(true);
+  });
+
+  it("fails closed when the backend coverage report is missing or cannot upload", () => {
+    const workflow = readYaml(".github/workflows/ci.yml");
+    const validateCode = nestedRecord(workflow, ["jobs", "validate-code"]);
+    const steps = recordArray(validateCode.steps, "jobs.validate-code.steps");
+
+    const stepNames = steps.map((step) => step.name);
+    const verifyIndex = stepNames.indexOf("Verify coverage report exists");
+    const coverageUploadIndex = stepNames.indexOf("Upload coverage to Codecov");
+    const testResultsUploadIndex = stepNames.indexOf("Upload Vitest results to Codecov");
+
+    expect(verifyIndex).toBeGreaterThan(-1);
+    expect(coverageUploadIndex).toBeGreaterThan(verifyIndex);
+    expect(testResultsUploadIndex).toBeGreaterThan(coverageUploadIndex);
+
+    const verifyStep = steps[verifyIndex]!;
+    const coverageUpload = steps[coverageUploadIndex]!;
+    const testResultsUpload = steps[testResultsUploadIndex]!;
+
+    expect(verifyStep.if).toBe(coverageUpload.if);
+    expect(String(verifyStep.run)).toContain("coverage/lcov.info is missing or empty");
+    expect(String(verifyStep.run)).toContain("exit 1");
+
+    const coverageUploadWith = record(coverageUpload.with, "coverage upload with");
+    expect(coverageUploadWith.files).toBe("./coverage/lcov.info");
+    expect(coverageUploadWith.disable_search).toBe(true);
+    expect(coverageUploadWith.fail_ci_if_error).toBe(true);
+
+    const testResultsUploadWith = record(testResultsUpload.with, "test results upload with");
+    expect(testResultsUploadWith.report_type).toBe("test_results");
+    expect(testResultsUploadWith.disable_search).toBe(true);
+    expect(testResultsUploadWith.fail_ci_if_error).toBe(false);
+  });
+});
@@ -30,7 +30,7 @@ export default defineConfig({
       exclude: ["src/env.d.ts", "apps/**"],
       // Emit lcov for Codecov to compute patch (changed-lines) coverage.
       reporter: ["text", "lcov"],
-      // The 97% requirement now lives in codecov.yml as a *patch* gate (changed
+      // The 99% requirement now lives in codecov.yml as a *patch* gate (changed
       // lines only), which is compositional: merging one PR can't drop another
       // below the bar. These global thresholds are only a loose catastrophe net
       // (e.g. a deleted test file), set well below actual coverage so routine