From 49a741f141f7826bf6ebe13a465f0fe8de77bb0d Mon Sep 17 00:00:00 2001
From: Johnathan Falk <johnathan.falk@gmail.com>
Date: Mon, 5 Oct 2026 00:50:08 -0400
Subject: [PATCH] fix(ci): skip the dead Go coverage rerun without
 .testcoverage.yml; go-ci cap 40m

The Generate step re-ran the full suite for a checker that exits at once
when .testcoverage.yml is missing, pushing go-ci past its 30m cap.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017MtQ5LP2n3t3bs7AhptkKJ
---
 .github/workflows/reusable-ci.yml                 | 15 +++++++++++----
 .../20261005_go_ci_skip_dead_coverage_rerun.md    |  6 ++++++
 2 files changed, 17 insertions(+), 4 deletions(-)
 create mode 100644 changelog.d/20261005_go_ci_skip_dead_coverage_rerun.md

diff --git a/.github/workflows/reusable-ci.yml b/.github/workflows/reusable-ci.yml
index 9d050616..2254a442 100644
--- a/.github/workflows/reusable-ci.yml
+++ b/.github/workflows/reusable-ci.yml
@@ -1,5 +1,5 @@
 # file: .github/workflows/reusable-ci.yml
-# version: 1.14.0
+# version: 1.15.0
 # guid: reusable-ci-2025-09-24-core-workflow
 
 name: Reusable CI Workflow
@@ -409,7 +409,12 @@ jobs:
       fail-fast: false
       matrix: ${{ fromJSON(needs.load-config.outputs.go-matrix) }}
     runs-on: ${{ matrix.os }}
-    timeout-minutes: 30
+    # 40, not 30: the go-test helper runs `go test -race` with no -timeout, so
+    # a hung test binary is killed by Go's default 10m. The cap must exceed
+    # setup (~4.5m) + the slowest measured suite (~20.5m) + that 10m, or the
+    # runner cancels the job before Go prints the goroutine dump naming the
+    # stuck test.
+    timeout-minutes: 40
     permissions:
       contents: read
     env:
@@ -481,13 +486,15 @@ jobs:
           command: go-test
           coverage-threshold: ${{ inputs.coverage-threshold || needs.load-config.outputs.coverage-threshold }}
 
+      # Only feeds go-test-coverage, which exits at once without .testcoverage.yml;
+      # without the file this step was a dead full-suite rerun (6-8 min).
       - name: Generate Go coverage profile
-        if: ${{ !inputs.skip-tests }}
+        if: ${{ !inputs.skip-tests && hashFiles('.testcoverage.yml') != '' }}
         continue-on-error: true
         run: go test ./... -coverprofile=./cover.out -covermode=atomic -coverpkg=./...
 
       - name: Check Go test coverage
-        if: ${{ !inputs.skip-tests }}
+        if: ${{ !inputs.skip-tests && hashFiles('.testcoverage.yml') != '' }}
         continue-on-error: true
         uses: vladopajic/go-test-coverage@f94bcf0d6b9fa5fb8b783830b22648f6c17475e2 # v2.19.0
         with:
diff --git a/changelog.d/20261005_go_ci_skip_dead_coverage_rerun.md b/changelog.d/20261005_go_ci_skip_dead_coverage_rerun.md
new file mode 100644
index 00000000..ea75c92f
--- /dev/null
+++ b/changelog.d/20261005_go_ci_skip_dead_coverage_rerun.md
@@ -0,0 +1,6 @@
+### Fixed
+
+#### reusable-ci: no dead coverage rerun, 40-minute Go CI cap
+
+- "Generate Go coverage profile" and "Check Go test coverage" now run only when the repo has `.testcoverage.yml`. Without it the first step re-ran the whole suite (6–8 min) for a checker that exits immediately, which pushed audiobook-organizer's go-ci over its 30-minute cap.
+- `go-ci` `timeout-minutes` is now 40, so a hung test is killed by Go's 10-minute test timeout (with a goroutine dump naming it) before the runner cancels the job.