Merge pull request #1282 from DeusData/feat/local-remote-parity

ci: revert runner-side Defender gating (RTP policy-locked off on hosted images)

.github/workflows/_smoke.yml

@@ -172,10 +172,6 @@ jobs:
     steps:
       - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7.0.0
 
-      - name: Ensure Windows Defender is active (venue parity)
-        shell: pwsh
-        run: scripts/ci/ensure-defender.ps1
-
       - uses: msys2/setup-msys2@66cd2cce69caa17b53920067426061ca1de3a884 # v2
         with:
           msystem: ${{ matrix.arch == 'arm64' && 'CLANGARM64' || 'CLANG64' }}
@@ -258,9 +254,7 @@ jobs:
             if ($code -eq 2) {
               Write-Host "BLOCKED: Windows Defender flagged $binary!"; exit 1
             } elseif ($code -ne 0) {
-              # ensure-defender gates this job earlier, so the engine MUST be
-              # available here: an engine failure is a red gate, never a skip.
-              Write-Host "BLOCKED: Defender scan engine could not run for $binary (exit $code) - Defender-ON parity violated"; exit 1
+              Write-Host "::warning::Windows Defender scan could not run for $binary (exit $code) - skipping AV gate on this runner"
             } else {
               Write-Host "=== Windows Defender: $binary clean ==="
             }

.github/workflows/_soak.yml

@@ -93,9 +93,6 @@ jobs:
     steps:
       - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7.0.0
 
-      - name: Ensure Windows Defender is active (venue parity)
-        shell: pwsh
-        run: scripts/ci/ensure-defender.ps1
       - uses: msys2/setup-msys2@66cd2cce69caa17b53920067426061ca1de3a884 # v2
         with:
           msystem: CLANG64
@@ -153,9 +150,6 @@ jobs:
     steps:
       - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7.0.0
 
-      - name: Ensure Windows Defender is active (venue parity)
-        shell: pwsh
-        run: scripts/ci/ensure-defender.ps1
       - uses: msys2/setup-msys2@66cd2cce69caa17b53920067426061ca1de3a884 # v2
         with:
           msystem: CLANGARM64
@@ -271,9 +265,6 @@ jobs:
     steps:
       - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7.0.0
 
-      - name: Ensure Windows Defender is active (venue parity)
-        shell: pwsh
-        run: scripts/ci/ensure-defender.ps1
       - uses: msys2/setup-msys2@66cd2cce69caa17b53920067426061ca1de3a884 # v2
         with:
           msystem: CLANG64

.github/workflows/_test.yml

@@ -154,7 +154,12 @@ jobs:
         with:
           name: shard-manifest-${{ matrix.os }}-${{ matrix.cc }}-${{ strategy.job-index }}
           path: build/c/test-logs/shard-manifest.txt
-          if-no-files-found: error
+          # The harness writes the manifest BEFORE any suite runs, so a
+          # missing file here means the job died before the harness started
+          # (contract step / build) — that failure is already the job's red;
+          # warn instead of stacking a second error on top of it. The
+          # shard-completeness job still gates the cross-shard union.
+          if-no-files-found: warn
 
       - name: Compiler cache (save)
         if: always() && (matrix.shard == null || startsWith(matrix.shard, '1/'))
@@ -247,11 +252,12 @@ jobs:
     steps:
       - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7.0.0
 
-      # Defender parity: every Windows venue (local VM and runner alike) tests
-      # with real-time protection ACTIVE — enabled + verified, never skipped.
-      - name: Ensure Windows Defender is active (venue parity)
-        shell: pwsh
-        run: scripts/ci/ensure-defender.ps1
+      # Defender posture: real-time protection is POLICY-LOCKED OFF on hosted
+      # Windows images (verified 2026-07-27: WinDefend starts and
+      # Set-MpPreference accepts, but RTP stays off — untamperable from the
+      # job). The local VM leg enforces Defender-ON via the canonical Defender
+      # preflight script, so AV-interaction coverage lives there; runners run
+      # without RTP by platform constraint.
 
       - uses: msys2/setup-msys2@66cd2cce69caa17b53920067426061ca1de3a884 # v2
         with:
@@ -329,7 +335,12 @@ jobs:
         with:
           name: shard-manifest-${{ matrix.os }}-${{ matrix.msystem }}-${{ strategy.job-index }}
           path: build/c/test-logs/shard-manifest.txt
-          if-no-files-found: error
+          # The harness writes the manifest BEFORE any suite runs, so a
+          # missing file here means the job died before the harness started
+          # (contract step / build) — that failure is already the job's red;
+          # warn instead of stacking a second error on top of it. The
+          # shard-completeness job still gates the cross-shard union.
+          if-no-files-found: warn
 
       # Shard 1 (or the unsharded job) saves the compiler cache; every shard
       # builds the identical runner, so one cache carries the full object set.
@@ -381,10 +392,6 @@ jobs:
     steps:
       - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7.0.0
 
-      - name: Ensure Windows Defender is active (venue parity)
-        shell: pwsh
-        run: scripts/ci/ensure-defender.ps1
-
       - uses: msys2/setup-msys2@66cd2cce69caa17b53920067426061ca1de3a884 # v2
         with:
           msystem: CLANG64

.github/workflows/pr.yml

@@ -84,11 +84,6 @@ jobs:
     steps:
       - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7.0.0
 
-      - name: Ensure Windows Defender is active (venue parity)
-        if: matrix.os == 'windows-latest'
-        shell: pwsh
-        run: scripts/ci/ensure-defender.ps1
-
       - name: Install deps (Ubuntu)
         if: matrix.os == 'ubuntu-latest'
         run: sudo apt-get update && sudo apt-get install -y zlib1g-dev ccache

.gitignore

@@ -66,3 +66,7 @@ graph-ui/.npm-cache-local/
 # Python bytecode from tests/windows/ harness
 __pycache__/
 *.pyc
+
+# Local soak-leg outputs (uploaded as CI artifacts, never committed)
+soak-results/
+soak-results-query-leak/

scripts/run-tests-parallel.sh

@@ -195,6 +195,22 @@ shard_filter < "$SER_FILE" > "$SER_FILE.shard" && mv "$SER_FILE.shard" "$SER_FIL
 SHARD_EXPECT="$LOGDIR/suites-shard.txt"
 cat "$PAR_FILE" "$SER_FILE" > "$SHARD_EXPECT"
 NSHARD=$(wc -l < "$SHARD_EXPECT" | tr -d ' ')
+
+# Machine-checkable manifest for CI's cross-shard completeness job: it
+# proves at runtime that the shards of one leg agree on N and on the full
+# suite list, and that the union of their slices IS that list — the guard
+# against a mis-plumbed CBM_TEST_SHARD (two jobs running the same slice
+# passes every per-shard check but silently drops a slice; only a
+# cross-shard view catches it). Written BEFORE any suite runs: the slice is
+# fully determined here, and a red run's manifest is exactly as load-bearing
+# as a green one's — CI uploads it if: always().
+{
+    echo "leg=${CBM_TEST_LEG:-local}"
+    echo "shard=${SHARD_INDEX}/${SHARD_TOTAL}"
+    echo "list_sha256=$(sort "$SUITES_FILE" | { sha256sum 2>/dev/null || shasum -a 256; } | awk '{print $1}')"
+    echo "--- slice ---"
+    cat "$SHARD_EXPECT"
+} > "$LOGDIR/shard-manifest.txt"
 echo "=== parallel test run: $NSHARD of $NSUITES suites (shard ${SHARD_INDEX}/${SHARD_TOTAL}, $(wc -l < "$SER_FILE" | tr -d ' ') serial-tail), $JOBS jobs ==="
 
 # Per-suite wall-clock ceilings make a wedged child fail loudly. The
@@ -255,20 +271,6 @@ stamp_windows_build_dir pre-tail
 run_wave "$FLEX_FILE" "${CBM_TAIL_JOBS:-2}"
 run_wave "$EXCL_FILE" 1
 
-# Machine-checkable manifest for CI's cross-shard completeness job: it
-# proves at runtime that the shards of one leg agree on N and on the full
-# suite list, and that the union of their slices IS that list — the guard
-# against a mis-plumbed CBM_TEST_SHARD (two jobs running the same slice
-# passes every per-shard check but silently drops a slice; only a
-# cross-shard view catches it).
-{
-    echo "leg=${CBM_TEST_LEG:-local}"
-    echo "shard=${SHARD_INDEX}/${SHARD_TOTAL}"
-    echo "list_sha256=$(sort "$SUITES_FILE" | { sha256sum 2>/dev/null || shasum -a 256; } | awk '{print $1}')"
-    echo "--- slice ---"
-    cat "$SHARD_EXPECT"
-} > "$LOGDIR/shard-manifest.txt"
-
 # ── Union guard: every suite in this shard's slice produced exactly one
 # result. The slice is deterministic, so N green shard jobs = full coverage;
 # a shard that ran anything more, less, or twice fails here. ──

test-infrastructure/vm/README.md

@@ -119,10 +119,12 @@ documented honestly where not:
 
 - **Preflight, every venue-grade command:** `clean-test-residue.ps1` sweeps
   temp roots/staged binaries and asserts a runner-like free-disk floor, then
-  `scripts/ci/ensure-defender.ps1` (the SAME script every Windows CI job runs)
-  enables + verifies Defender real-time protection. Defender posture is
-  matched ON on all venues — the runners enable it too; a venue where it
-  cannot be activated goes red, never silently tests a different OS shape.
+  `scripts/ci/ensure-defender.ps1` enables + verifies Defender real-time
+  protection ON THIS VM (fail-closed: a drifted-off Defender goes red).
+  Hosted GitHub runners CANNOT match this posture — RTP is policy-locked off
+  on their images (verified 2026-07-27: WinDefend starts, Set-MpPreference
+  accepts, RTP stays off) — so AV-interaction coverage is a deliberate LOCAL
+  superset, and runner release scans stay on-demand (MpCmdRun, fail-soft).
 - **Snapshot-revert is NOT scriptable via utmctl** (its verb set is
   version/list/status/start/suspend/stop — no snapshot support), so per-run
   revert-to-clean is not wired. The honest fallback is the sweep preflight

tests/test_venue_parity_contract.sh

@@ -249,32 +249,24 @@ REQUIRED = [
     ("pr.yml", r"vm-smoke\.sh", "PR CI smokes through the shared wrapper"),
     ("_build.yml", r"scripts/package-release\.sh",
      "release archives are produced by the canonical packaging entry"),
-    ("_test.yml", r"scripts/ci/ensure-defender\.ps1",
-     "Windows test legs enable+verify Defender (venue parity)"),
-    ("_smoke.yml", r"scripts/ci/ensure-defender\.ps1",
-     "Windows release smoke enables+verifies Defender (venue parity)"),
-    ("_soak.yml", r"scripts/ci/ensure-defender\.ps1",
-     "Windows soak legs enable+verify Defender (venue parity)"),
-    ("pr.yml", r"scripts/ci/ensure-defender\.ps1",
-     "PR Windows smoke enables+verifies Defender (venue parity)"),
 ]
 for name, pattern, why in REQUIRED:
     path = workflows / name
     if path.exists() and not re.search(pattern, path.read_text()):
         failures.append(f"{name}: missing required `{pattern}` — {why}")
 
-# Defender-ON parity: EVERY Windows job in every venue workflow enables and
-# verifies real-time protection — a single job dropping the step must fail,
-# so the expectation is an exact per-file count, not mere presence.
-DEFENDER_STEPS = {"_test.yml": 2, "_soak.yml": 3, "_smoke.yml": 1, "pr.yml": 1}
-for name, expected in DEFENDER_STEPS.items():
+# Defender posture: hosted Windows runners have real-time protection
+# POLICY-LOCKED OFF (verified 2026-07-27: WinDefend starts, Set-MpPreference
+# accepts, RTP stays off), so runner jobs must NOT gate on enabling it; the
+# LOCAL VM leg enforces Defender-ON via the same canonical script in its
+# preflight (see LOCAL_REQUIRED below) — superset coverage where the platform
+# allows it.
+for name in ("_test.yml", "_soak.yml", "_smoke.yml", "pr.yml"):
     path = workflows / name
-    if path.exists():
-        got = path.read_text().count("scripts/ci/ensure-defender.ps1")
-        if got != expected:
-            failures.append(
-                f"{name}: expected {expected} ensure-defender.ps1 step(s) "
-                f"(one per Windows job), found {got}")
+    if path.exists() and "ensure-defender.ps1" in path.read_text():
+        failures.append(
+            f"{name}: ensure-defender.ps1 must not gate hosted runners — RTP is "
+            f"policy-locked off there; the VM preflight owns Defender-ON coverage")
 
 # ── Layer 4: the local venues route through the same entries ──
 compose = root / "test-infrastructure" / "docker-compose.yml"