PartitionAlloc: Make in-place realloc() less likely to happen

Source/wtf/PartitionAllocTest.cpp

Index: Source/wtf/PartitionAllocTest.cpp
diff --git a/Source/wtf/PartitionAllocTest.cpp b/Source/wtf/PartitionAllocTest.cpp
index 9cbdce5652d9ac9c43b7046b41ae7c556f8bf633..7190d90ed220a55282c9bd36c5fd42d69d2361d1 100644
--- a/Source/wtf/PartitionAllocTest.cpp
+++ b/Source/wtf/PartitionAllocTest.cpp
@@ -692,14 +692,13 @@ TEST(PartitionAllocTest, Realloc)
 
     partitionFreeGeneric(genericAllocator.root(), ptr);
 
-    // Test that shrinking a direct mapped allocation happens in-place (even
-    // though the new size is smaller than kGenericMaxBucketed).
+    // Test that shrinking a direct mapped allocation happens in-place.
     size = WTF::kGenericMaxBucketed + 16 * WTF::kSystemPageSize;
     ptr = partitionAllocGeneric(genericAllocator.root(), size);
     size_t actualSize = partitionAllocGetSize(ptr);
-    ptr2 = partitionReallocGeneric(genericAllocator.root(), ptr, WTF::kGenericMaxBucketed - 16 * WTF::kSystemPageSize);
+    ptr2 = partitionReallocGeneric(genericAllocator.root(), ptr, WTF::kGenericMaxBucketed + 8 * WTF::kSystemPageSize);
     EXPECT_EQ(ptr, ptr2);
-    EXPECT_EQ(actualSize - 32 * WTF::kSystemPageSize, partitionAllocGetSize(ptr2));
+    EXPECT_EQ(actualSize - 8 * WTF::kSystemPageSize, partitionAllocGetSize(ptr2));

[Tom Sepez, 2014/04/29 20:27:02]

It would be nice to have a test case that shows, say, that going from say 2MB to
16B is not in-place.  Do we have this?

 
     // Test that a previously in-place shrunk direct mapped allocation can be
     // expanded up again within its original size.