feat: implement issue #894 — [Fleet Monitor] petry-projects/.github — .github/workflows/ci.yml

.github/workflows/ci-tests.yml

@@ -0,0 +1,55 @@
+# Regression guard for the flaky "Lint GitHub Actions" download in ci.yml
+# (issue #894): every curl download in ci.yml must carry bounded retries so a
+# transient GitHub Releases blip retries instead of failing the job.
+# See test/workflows/ci/install-resilience.bats.
+name: CI Workflow Tests
+
+on:
+  pull_request:
+    paths:
+      - '.github/workflows/ci.yml'
+      - 'test/workflows/ci/**'
+      - '.github/workflows/ci-tests.yml'
+  push:
+    branches: [main]
+    paths:
+      - '.github/workflows/ci.yml'
+      - 'test/workflows/ci/**'
+      - '.github/workflows/ci-tests.yml'
+
+permissions: {}
+
+concurrency:
+  group: ci-tests-${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  test:
+    name: bats
+    runs-on: ubuntu-latest
+    timeout-minutes: 10
+    permissions:
+      contents: read
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7.0.0
+        with:
+          persist-credentials: false
+
+      - name: Install bats
+        run: |
+          set -euo pipefail
+          for attempt in {1..3}; do
+            if sudo apt-get update -qq && sudo apt-get install -y --no-install-recommends bats; then
+              exit 0
+            fi
+            if [ "$attempt" -lt 3 ]; then
+              echo "Apt install failed, retrying in 5 seconds..." >&2
+              sleep 5
+            fi
+          done
+          echo "Apt install failed after 3 attempts" >&2
+          exit 1
+
+      - name: bats
+        run: bats --print-output-on-failure test/workflows/ci/

.github/workflows/ci.yml

@@ -56,7 +56,14 @@ jobs:
           # Install actionlint via verified download (pinned version + checksum)
           ACTIONLINT_VERSION="1.7.7"
           ACTIONLINT_SHA="023070a287cd8cccd71515fedc843f1985bf96c436b7effaecce67290e7e0757"
-          curl -sLo actionlint.tar.gz "https://github.com/rhysd/actionlint/releases/download/v${ACTIONLINT_VERSION}/actionlint_${ACTIONLINT_VERSION}_linux_amd64.tar.gz"
+          # Bounded retries so a transient GitHub Releases 5xx / connection blip
+          # retries instead of failing the job (guards the fleet failure-rate
+          # metric — see test/workflows/ci/install-resilience.bats, issue #894).
+          curl -sSfL --proto '=https' --proto-redir '=https' \
+            --connect-timeout 10 --max-time 60 \
+            --retry 3 --retry-delay 2 --retry-connrefused --retry-all-errors \
+            -o actionlint.tar.gz \
+            "https://github.com/rhysd/actionlint/releases/download/v${ACTIONLINT_VERSION}/actionlint_${ACTIONLINT_VERSION}_linux_amd64.tar.gz"
           echo "${ACTIONLINT_SHA}  actionlint.tar.gz" | sha256sum -c -
           tar xzf actionlint.tar.gz actionlint
           chmod +x actionlint

test/workflows/ci/helpers/setup.bash

@@ -0,0 +1,10 @@
+#!/usr/bin/env bash
+# Common test helpers for the ci.yml bats suite.
+# Mirrors the pattern in test/workflows/pr-review-mention/helpers/setup.bash.
+
+# Repo root, regardless of where bats is invoked from.
+TT_REPO_ROOT="$(cd -- "$(dirname -- "${BASH_SOURCE[0]}")/../../../.." && pwd)"
+export TT_REPO_ROOT
+
+TT_WORKFLOW="${TT_REPO_ROOT}/.github/workflows/ci.yml"
+export TT_WORKFLOW

test/workflows/ci/install-resilience.bats

@@ -0,0 +1,52 @@
+#!/usr/bin/env bats
+# Regression guard for the flaky "Lint GitHub Actions" step in
+# .github/workflows/ci.yml.
+#
+# Pins issue #894: ci.yml intermittently fails (~14% of runs) because the pinned
+# rhysd/actionlint binary is fetched from GitHub Releases — which redirects to
+# objects.githubusercontent.com — with a bare `curl` that aborts on the first
+# transient network/5xx blip. This is the same failure mode fixed for
+# pr-review-mention-tests.yml under #739. AGENTS.md requires CI to be reliable;
+# the one unavoidable network fetch must therefore retry on transient errors.
+# These tests assert every download in ci.yml carries bounded curl retries, so a
+# future edit can't silently reintroduce a bare download.
+
+load 'helpers/setup'
+
+@test "install: the ci workflow exists" {
+  [ -f "$TT_WORKFLOW" ]
+}
+
+@test "install: every curl download in ci.yml uses bounded retries" {
+  [ -f "$TT_WORKFLOW" ]
+
+  # Parse the workflow joining backslash-continued lines so that a curl command
+  # split across multiple lines is treated as a single logical invocation.
+  # This avoids false failures when flags appear on continuation lines, and
+  # avoids false passes when a bare `curl` is split from its flags.
+  local curls=()
+  local current=""
+  while IFS= read -r line || [[ -n "$line" ]]; do
+    line="${line%$'\r'}"
+    if [[ "$line" == *\\ ]]; then
+      current+="${line%\\} "
+    else
+      current+="$line"
+      if [[ "$current" == *curl* ]] && ! [[ "$current" =~ ^[[:space:]]*# ]] && ! [[ "$current" =~ ^[[:space:]]*-?[[:space:]]*(name|uses|with): ]]; then
+        curls+=("$current")
+      fi
+      current=""
+    fi
+  done < "$TT_WORKFLOW"
+
+  # At least one curl invocation must be present so the test remains meaningful.
+  [ "${#curls[@]}" -ge 1 ]
+  for cmd in "${curls[@]}"; do
+    # A finite retry budget (not unbounded) so a truly-down mirror still fails fast.
+    [[ "$cmd" =~ --retry[[:space:]=][1-9] ]]
+    # Retry on connection refused, which curl otherwise treats as non-transient.
+    [[ "$cmd" == *"--retry-connrefused"* ]]
+    # Retry on transient HTTP 5xx too, not just connection-level errors.
+    [[ "$cmd" == *"--retry-all-errors"* ]]
+  done
+}