Remove excessively clever (and wrong) implementation of newBuffer()

core/src/main/scala/org/apache/spark/util/collection/ExternalSorter.scala

@@ -130,18 +130,16 @@ private[spark] class ExternalSorter[K, V, C](
   // grow internal data structures by growing + copying every time the number of objects doubles.
   private val serializerBatchSize = conf.getLong("spark.shuffle.spill.batchSize", 10000)
 
-  // The idea here is to reduce minimize the visibility of variables like `kvChunkSize` and
-  // `useSerializedPairBuffer` without recomputing them on each call to `newBuffer()`.
-  private def newBuffer: () => WritablePartitionedPairCollection[K, C] with SizeTracker = {
-    val useSerializedPairBuffer =
-      ordering.isEmpty &&
-        conf.getBoolean("spark.shuffle.sort.serializeMapOutputs", true) &&
-        ser.supportsRelocationOfSerializedObjects
+  private val useSerializedPairBuffer =
+    ordering.isEmpty &&
+      conf.getBoolean("spark.shuffle.sort.serializeMapOutputs", true) &&
+      ser.supportsRelocationOfSerializedObjects
+  private val kvChunkSize = conf.getInt("spark.shuffle.sort.kvChunkSize", 1 << 22) // 4 MB
+  private def newBuffer(): WritablePartitionedPairCollection[K, C] with SizeTracker = {
     if (useSerializedPairBuffer) {
-      val kvChunkSize = conf.getInt("spark.shuffle.sort.kvChunkSize", 1 << 22) // 4 MB
-      () => new PartitionedSerializedPairBuffer(metaInitialRecords = 256, kvChunkSize, serInstance)
+      new PartitionedSerializedPairBuffer(metaInitialRecords = 256, kvChunkSize, serInstance)
     } else {
-      () => new PartitionedPairBuffer[K, C]
+      new PartitionedPairBuffer[K, C]
     }
   }
   // Data structures to store in-memory objects before we spill. Depending on whether we have an