core/src/main/java/io/questdb/PropServerConfiguration.java (1)

1655-1665: ⚠️ Potential issue | 🟡 Minor

Add lower-bound validation for interval knobs.
Non‑positive values for bracket depth or thresholds could yield surprising behavior even though the relative check passes. Consider guarding for >= 1 to fail fast.

🔧 Suggested validation addition

             this.sqlIntervalMaxBracketDepth = getInt(properties, env, PropertyKey.CAIRO_SQL_INTERVAL_MAX_BRACKET_DEPTH, 8);
             this.sqlIntervalMaxIntervalsAfterMerge = getInt(properties, env, PropertyKey.CAIRO_SQL_INTERVAL_MAX_INTERVALS_AFTER_MERGE, 1024);
             this.sqlIntervalIncrementalMergeThreshold = getInt(properties, env, PropertyKey.CAIRO_SQL_INTERVAL_INCREMENTAL_MERGE_THRESHOLD, 256);
+            if (this.sqlIntervalMaxBracketDepth < 1) {
+                throw new ServerConfigurationException(
+                        PropertyKey.CAIRO_SQL_INTERVAL_MAX_BRACKET_DEPTH.getPropertyPath()
+                                + " (" + this.sqlIntervalMaxBracketDepth + ") must be >= 1"
+                );
+            }
+            if (this.sqlIntervalIncrementalMergeThreshold < 1) {
+                throw new ServerConfigurationException(
+                        PropertyKey.CAIRO_SQL_INTERVAL_INCREMENTAL_MERGE_THRESHOLD.getPropertyPath()
+                                + " (" + this.sqlIntervalIncrementalMergeThreshold + ") must be >= 1"
+                );
+            }
             if (this.sqlIntervalMaxIntervalsAfterMerge <= this.sqlIntervalIncrementalMergeThreshold) {
                 throw new ServerConfigurationException(
                         PropertyKey.CAIRO_SQL_INTERVAL_MAX_INTERVALS_AFTER_MERGE.getPropertyPath()

core/src/main/java/io/questdb/griffin/engine/table/AsyncGroupByNotKeyedAtom.java (1)

172-188: ⚠️ Potential issue | 🟠 Major

Close leaks filtered memory records.

frameFilteredMemoryRecords and ownerPageFrameFilteredNoRandomAccessMemoryRecord are newly added but never freed in close(), which can retain native memory across cursor lifetimes.

🧹 Proposed fix

     Misc.free(ownerAllocator);
     Misc.freeObjList(perWorkerAllocators);
     Misc.free(ownerMapValue);
     Misc.freeObjList(perWorkerMapValues);
+    Misc.freeObjList(frameFilteredMemoryRecords);
+    Misc.free(ownerPageFrameFilteredNoRandomAccessMemoryRecord);
     if (perWorkerGroupByFunctions != null) {
         for (int i = 0, n = perWorkerGroupByFunctions.size(); i < n; i++) {
             Misc.freeObjList(perWorkerGroupByFunctions.getQuick(i));
         }
     }

core/src/main/java/io/questdb/griffin/engine/join/AsyncWindowJoinRecordCursorFactory.java (5)

959-978: ⚠️ Potential issue | 🟠 Major

Potential slot leak if exception occurs before try block.

The slotId is acquired at line 960, but the try block that releases it in finally doesn't start until line 978. If any operation between these points throws (e.g., populateFrameMemory(), record.init()), the slot will never be released.

🔧 Proposed fix: move try block earlier to cover slot acquisition

     final boolean owner = stealingFrameSequence != null && stealingFrameSequence == task.getFrameSequence();
     final int slotId = atom.maybeAcquire(workerId, owner, circuitBreaker);
 
+    try {
         final PageFrameMemory frameMemory;
         final boolean isParquetFrame = task.isParquetFrame();
         final boolean useLateMaterialization = atom.shouldUseLateMaterialization(slotId, isParquetFrame);
         if (useLateMaterialization) {
             frameMemory = task.populateFrameMemory(atom.getFilterUsedColumnIndexes());
         } else {
             frameMemory = task.populateFrameMemory();
         }
         record.init(frameMemory);
 
         final DirectLongList rows = task.getFilteredRows();
         rows.clear();
 
         final Function filter = atom.getMasterFilter(slotId);
         final CompiledFilter compiledFilter = atom.getCompiledMasterFilter();
 
-    try {
         if (compiledFilter == null || frameMemory.hasColumnTops()) {

---

1108-1127: ⚠️ Potential issue | 🟠 Major

Same slot leak pattern as filterAndAggregate.

Move the try block to immediately after slot acquisition to ensure proper cleanup.

---

1265-1284: ⚠️ Potential issue | 🟠 Major

Same slot leak pattern as other filter methods.

Move the try block to immediately after slot acquisition.

---

1421-1440: ⚠️ Potential issue | 🟠 Major

Same slot leak pattern as other filter methods.

Move the try block to immediately after slot acquisition.

---

1566-1585: ⚠️ Potential issue | 🟠 Major

Same slot leak pattern as other filter methods.

Move the try block to immediately after slot acquisition.