fix: flink verifier path bug — flink-streaming-java → flink-runtime

WindowOperator.java lives in flink-runtime/ not flink-streaming-java/
in the sg-evals/flink--0cc95fcc mirror. The verifier was checking the
wrong directory, causing legitimate agent work to score 0/6. Confirmed
via Sourcegraph search and agent trajectory analysis (12 edits to the
correct path).

Also adds configs/smoke_scaffold_4.json for the 4 DOE scaffold tasks.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: sjarmak

benchmarks/ccb_fix/flink-window-late-data-fix-001/CLAUDE.md

@@ -5,8 +5,8 @@
 Fix silent late data dropping in Flink's merging window operator side output path.
 
 ## Key Reference Files
-- `flink-streaming-java/src/main/java/org/apache/flink/streaming/runtime/operators/windowing/WindowOperator.java` — main target
-- `flink-streaming-java/src/main/java/org/apache/flink/streaming/runtime/operators/windowing/EvictingWindowOperator.java` — evicting variant
+- `flink-runtime/src/main/java/org/apache/flink/streaming/runtime/operators/windowing/WindowOperator.java` — main target
+- `flink-runtime/src/main/java/org/apache/flink/streaming/runtime/operators/windowing/EvictingWindowOperator.java` — evicting variant
 - `flink-runtime/src/main/java/org/apache/flink/streaming/api/windowing/assigners/MergingWindowAssigner.java` — merge base
 - `flink-streaming-java/src/test/java/org/apache/flink/streaming/runtime/operators/windowing/WindowOperatorTest.java` — tests
 

benchmarks/ccb_fix/flink-window-late-data-fix-001/instruction.md

@@ -20,7 +20,7 @@ When using event-time session windows with `allowedLateness(Time.seconds(0))` an
 2. **Fix the late element handling**:
    - Ensure late elements for merging windows check whether the element could be merged into an existing (non-expired) window BEFORE marking as late
    - OR ensure the side output path correctly emits the element via the registered OutputTag
-   - The fix should be in `flink-streaming-java/src/main/java/org/apache/flink/streaming/runtime/operators/windowing/`
+   - The fix should be in `flink-runtime/src/main/java/org/apache/flink/streaming/runtime/operators/windowing/`
 
 3. **Ensure correct OutputTag wiring**:
    - Verify that `sideOutput()` receives the correct `OutputTag<T>` for late data
@@ -32,8 +32,8 @@ When using event-time session windows with `allowedLateness(Time.seconds(0))` an
    - Verify: late element appears in the side output stream, not silently dropped
 
 ## Key Reference Files
-- `flink-streaming-java/src/main/java/org/apache/flink/streaming/runtime/operators/windowing/WindowOperator.java` — main window operator
-- `flink-streaming-java/src/main/java/org/apache/flink/streaming/runtime/operators/windowing/EvictingWindowOperator.java` — evicting variant
+- `flink-runtime/src/main/java/org/apache/flink/streaming/runtime/operators/windowing/WindowOperator.java` — main window operator
+- `flink-runtime/src/main/java/org/apache/flink/streaming/runtime/operators/windowing/EvictingWindowOperator.java` — evicting variant
 - `flink-runtime/src/main/java/org/apache/flink/streaming/api/windowing/assigners/MergingWindowAssigner.java` — merging window base
 - `flink-streaming-java/src/main/java/org/apache/flink/streaming/api/datastream/SingleOutputStreamOperator.java` — sideOutputLateData API
 - `flink-runtime/src/main/java/org/apache/flink/streaming/api/operators/AbstractStreamOperator.java` — base operator with output handling

benchmarks/ccb_fix/flink-window-late-data-fix-001/tests/test.sh

@@ -7,7 +7,7 @@ SCORE=0
 TOTAL=6
 WORKSPACE="${VERIFY_REPO:-/workspace}"
 
-WINDOW_OP_DIR="$WORKSPACE/flink-streaming-java/src/main/java/org/apache/flink/streaming/runtime/operators/windowing"
+WINDOW_OP_DIR="$WORKSPACE/flink-runtime/src/main/java/org/apache/flink/streaming/runtime/operators/windowing"
 WINDOW_TEST_DIR="$WORKSPACE/flink-streaming-java/src/test/java/org/apache/flink/streaming/runtime/operators/windowing"
 
 # Check 1: WindowOperator.processElement modified (late element handling)

configs/smoke_scaffold_4.json

@@ -0,0 +1,109 @@
+{
+  "metadata": {
+    "title": "Smoke test: 4 DOE rebalance scaffold tasks",
+    "description": "Targeted smoke for newly scaffolded tasks (3 ccb_feature + 1 ccb_fix). Run both baseline and sg_only modes.",
+    "generated_date": "2026-03-01",
+    "total_tasks": 4,
+    "note": "Use with validate_one_per_benchmark.sh --smoke-runtime (baseline) and --sg-only (MCP). Then run paired via run_selected_tasks.sh."
+  },
+  "methodology": {
+    "sdlc_suites": ["ccb_feature", "ccb_fix"]
+  },
+  "statistics": {
+    "total_tasks": 4,
+    "per_suite": {
+      "ccb_feature": 3,
+      "ccb_fix": 1
+    }
+  },
+  "tasks": [
+    {
+      "task_id": "postgres-copy-csv-header-feat-001",
+      "benchmark": "ccb_feature",
+      "sdlc_phase": "Implementation (feature)",
+      "language": "c",
+      "difficulty": "expert",
+      "category": "feature_implementation",
+      "repo": "postgres/postgres",
+      "mcp_benefit_score": 0.88,
+      "mcp_breakdown": {
+        "context_complexity": 0.95,
+        "cross_file_deps": 0.85,
+        "semantic_search_potential": 0.85,
+        "task_category_weight": 0.9
+      },
+      "selection_rationale": "DOE rebalance scaffold: C-language feature task for Neyman-optimal ccb_feature allocation (target 23)",
+      "task_dir": "ccb_feature/postgres-copy-csv-header-feat-001",
+      "context_length": 900000,
+      "context_length_source": "mcp_breakdown_proxy",
+      "files_count": 16,
+      "files_count_source": "mcp_breakdown_proxy"
+    },
+    {
+      "task_id": "servo-css-container-query-feat-001",
+      "benchmark": "ccb_feature",
+      "sdlc_phase": "Implementation (feature)",
+      "language": "rust",
+      "difficulty": "expert",
+      "category": "feature_implementation",
+      "repo": "servo/servo",
+      "mcp_benefit_score": 0.89,
+      "mcp_breakdown": {
+        "context_complexity": 0.95,
+        "cross_file_deps": 0.85,
+        "semantic_search_potential": 0.9,
+        "task_category_weight": 0.9
+      },
+      "selection_rationale": "DOE rebalance scaffold: Rust-language feature task for Neyman-optimal ccb_feature allocation (target 23)",
+      "task_dir": "ccb_feature/servo-css-container-query-feat-001",
+      "context_length": 950000,
+      "context_length_source": "mcp_breakdown_proxy",
+      "files_count": 16,
+      "files_count_source": "mcp_breakdown_proxy"
+    },
+    {
+      "task_id": "vscode-custom-fold-region-feat-001",
+      "benchmark": "ccb_feature",
+      "sdlc_phase": "Implementation (feature)",
+      "language": "typescript",
+      "difficulty": "hard",
+      "category": "feature_implementation",
+      "repo": "microsoft/vscode",
+      "mcp_benefit_score": 0.87,
+      "mcp_breakdown": {
+        "context_complexity": 0.9,
+        "cross_file_deps": 0.8,
+        "semantic_search_potential": 0.85,
+        "task_category_weight": 0.9
+      },
+      "selection_rationale": "DOE rebalance scaffold: TypeScript-language feature task for Neyman-optimal ccb_feature allocation (target 23)",
+      "task_dir": "ccb_feature/vscode-custom-fold-region-feat-001",
+      "context_length": 950000,
+      "context_length_source": "mcp_breakdown_proxy",
+      "files_count": 16,
+      "files_count_source": "mcp_breakdown_proxy"
+    },
+    {
+      "task_id": "flink-window-late-data-fix-001",
+      "benchmark": "ccb_fix",
+      "sdlc_phase": "Implementation (bug fix)",
+      "language": "java",
+      "difficulty": "hard",
+      "category": "bug_fix",
+      "repo": "apache/flink",
+      "mcp_benefit_score": 0.87,
+      "mcp_breakdown": {
+        "context_complexity": 0.9,
+        "cross_file_deps": 0.85,
+        "semantic_search_potential": 0.85,
+        "task_category_weight": 0.9
+      },
+      "selection_rationale": "DOE rebalance scaffold: Java-language fix task for Neyman-optimal ccb_fix allocation (target 26)",
+      "task_dir": "ccb_fix/flink-window-late-data-fix-001",
+      "context_length": 900000,
+      "context_length_source": "mcp_breakdown_proxy",
+      "files_count": 16,
+      "files_count_source": "mcp_breakdown_proxy"
+    }
+  ]
+}