ci: stop running expensive jobs that cannot inform the result (#5528)

Three independent fixes that all avoid burning self-hosted runners on work
whose outcome is already determined. None of them changes what is tested.

- ci-docs-only: add a "Quick Checks" companion job. It is a prerequisite for
  gating ci.yml's expensive jobs behind quick-checks (rustfs/backlog#1599):
  once "Quick Checks" is a required check, a docs-only PR would otherwise wait
  on it forever. The steps are a byte-identical copy of ci.yml's quick-checks
  rather than an `echo`, so that on a mixed PR the two same-named check runs
  execute the same commands against the same merge ref and cannot disagree —
  GitHub has no written contract for how it picks between same-named required
  check runs, and the real job only takes 45-51s, leaving no timing margin to
  rely on.

- ci: guard uring-integration with the same `closed` check every other job
  already has. The pull_request trigger includes `closed` only so the
  concurrency group cancels in-flight runs; this job had no guard and no
  `needs`, so every closed or merged PR ran the full io_uring suite (4m17s,
  7m19s and 7m31s on runs 30678272341, 30678117601 and 30662728539).

- ci: gate s3-lifecycle-behavior-tests on e2e-tests, matching
  s3-implemented-tests. Both lanes only download the prebuilt debug binary, and
  s3-implemented-tests already finishes later, so a green PR's wall clock is
  unchanged; a red one stops holding a sm-standard-4 for up to 30 minutes.

Refs: rustfs/backlog#1598, rustfs/backlog#1599

.github/workflows/ci-docs-only.yml

@@ -12,18 +12,20 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-# Companion to ci.yml for the required "Test and Lint" status check.
+# Companion to ci.yml for the required "Test and Lint" and "Quick Checks"
+# status checks.
 #
 # ci.yml skips docs-only pull requests via paths-ignore, but the branch
-# ruleset requires a check named "Test and Lint" — without this workflow a
-# docs-only PR would wait on that check forever. This workflow triggers on
-# exactly the paths ci.yml ignores and reports an instant success under the
-# same job name. Mixed PRs trigger both workflows and the real check still
-# gates: a required check with any failing run blocks the merge.
+# ruleset requires checks named "Test and Lint" and "Quick Checks" — without
+# this workflow a docs-only PR would wait on those checks forever. This
+# workflow triggers on exactly the paths ci.yml ignores and reports success
+# under the same job names. Mixed PRs trigger both workflows and the real
+# checks still gate: a required check with any failing run blocks the merge.
 # https://docs.github.com/en/repositories/configuring-branches-and-merges-in-your-repository/defining-the-mergeability-of-pull-requests/troubleshooting-required-status-checks#handling-skipped-but-required-checks
 #
 # Keep the paths list below in sync with the pull_request paths-ignore list
-# in ci.yml.
+# in ci.yml, and keep the quick-checks steps below byte-identical to the
+# quick-checks job in ci.yml (see the comment on that job).
 
 name: Continuous Integration (docs only)
 
@@ -52,9 +54,63 @@ permissions:
   contents: read
 
 jobs:
+  # Deliberately NOT a bare `echo`. Once "Quick Checks" becomes a required
+  # check, ci.yml gates every expensive job behind it, so a mixed PR reports
+  # two check runs with this name: the real one (45-51s) and this companion.
+  # GitHub has no written contract for how it picks between same-named
+  # required check runs ("latest wins" vs "any failure blocks"), so instead of
+  # relying on ordering we make both runs execute the same commands against
+  # the same merge ref — their conclusions are then necessarily identical and
+  # the choice does not matter. Keep these steps byte-identical to the
+  # quick-checks job in ci.yml (a guard script that asserts this, and the paths
+  # sync below, is tracked in rustfs/backlog#1603).
+  #
+  # For a genuinely docs-only PR this adds no strictness (no code changed, so
+  # fmt and the guards always pass) and costs ~50s of ubuntu-latest.
+  quick-checks:
+    name: Quick Checks
+    runs-on: ubuntu-latest
+    timeout-minutes: 10
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7
+
+      - name: Install ripgrep
+        run: sudo apt-get update && sudo apt-get install -y ripgrep
+
+      - name: Install Rust toolchain
+        uses: dtolnay/rust-toolchain@29eef336d9b2848a0b548edc03f92a220660cdb8 # stable
+        with:
+          components: rustfmt
+
+      - name: Check code formatting
+        run: cargo fmt --all --check
+
+      - name: Check unsafe code allowances
+        run: ./scripts/check_unsafe_code_allowances.sh
+
+      - name: Check layered dependencies
+        run: ./scripts/check_layer_dependencies.sh
+
+      - name: Check architecture migration rules
+        run: ./scripts/check_architecture_migration_rules.sh
+
+      - name: Check tokio io-uring feature guard
+        run: ./scripts/check_no_tokio_io_uring.sh
+
+      - name: Check extension schema boundaries
+        run: ./scripts/check_extension_schema_boundaries.sh
+
+      - name: Check body-cache whitelist guard
+        run: ./scripts/check_body_cache_whitelist.sh
+
+      - name: Check no planning docs committed
+        run: ./scripts/check_no_planning_docs.sh
+
   test-and-lint:
     name: Test and Lint
     runs-on: ubuntu-latest
+    timeout-minutes: 10
     steps:
       - name: Checkout repository
         uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7

.github/workflows/ci.yml

@@ -448,6 +448,13 @@ jobs:
 
   uring-integration:
     name: io_uring Integration (real)
+    # The pull_request trigger includes `closed` purely so the concurrency
+    # group cancels in-flight runs of a closed PR; every other job opts out of
+    # that run with this guard (or is skipped through its `needs` chain). This
+    # job had neither, so each closed/merged PR really ran the whole io_uring
+    # suite (measured 4m17s / 7m19s / 7m31s on runs 30678272341 / 30678117601 /
+    # 30662728539) and kept the cancellation run in progress for minutes.
+    if: github.event_name != 'pull_request' || github.event.action != 'closed'
     # GitHub-hosted ubuntu-latest runs a recent kernel with io_uring and, unlike
     # a container, applies no seccomp filter that would block io_uring_setup — so
     # the probe succeeds and the tests exercise the real UringBackend/FdCache/
@@ -746,7 +753,13 @@ jobs:
   # evaluates ILM within ~2s of the due time, well inside the poll window.
   s3-lifecycle-behavior-tests:
     name: S3 Lifecycle Behavior Tests
-    needs: [ build-rustfs-debug-binary ]
+    # Also gated on e2e-tests, matching s3-implemented-tests: when the e2e smoke
+    # suite is already red this lane cannot tell us anything new, and it holds a
+    # sm-standard-4 for up to 30 minutes doing so. Both lanes only download the
+    # prebuilt debug binary (no cargo build), and s3-implemented-tests — which
+    # already waits on e2e-tests — finishes later anyway, so a green PR's total
+    # wall clock is unchanged.
+    needs: [ build-rustfs-debug-binary, e2e-tests ]
     runs-on: sm-standard-4
     timeout-minutes: 30
     steps: