Merge pull request #687 from Altinity/backports/24.8/74534_fix_ubsan_error_in_parquetdatabuffer

Enmk · web-flow · commit 40f384da5e93 · 2025-03-14T11:56:35.000+01:00
24.8 Backport of ClickHouse#74534 - Fix UBSan error in ParquetDataBuffer
diff --git a/src/Processors/Formats/Impl/Parquet/ParquetDataBuffer.h b/src/Processors/Formats/Impl/Parquet/ParquetDataBuffer.h
@@ -54,11 +54,10 @@ class ParquetDataBuffer
         auto necessary_bytes = count * sizeof(ParquetType);
         checkAvaible(necessary_bytes);
 
-        const ParquetType* src = reinterpret_cast<const ParquetType*>(data);
-
         for (std::size_t i = 0; i < count; i++)
         {
-            dst[i] = static_cast<TValue>(src[i]);
+            auto offset = i * sizeof(ParquetType);
+            dst[i] = unalignedLoad<TValue>(data + offset);
         }
 
         consume(necessary_bytes);

Original file line number	Diff line number	Diff line change
`@@ -54,11 +54,10 @@ class ParquetDataBuffer`
`54`	`54`	`auto necessary_bytes = count * sizeof(ParquetType);`
`55`	`55`	`checkAvaible(necessary_bytes);`
`56`	`56`
`57`		`- const ParquetType* src = reinterpret_cast<const ParquetType*>(data);`
`58`		`-`
`59`	`57`	`for (std::size_t i = 0; i < count; i++)`
`60`	`58`	`{`
`61`		`- dst[i] = static_cast<TValue>(src[i]);`
	`59`	`+ auto offset = i * sizeof(ParquetType);`
	`60`	`+ dst[i] = unalignedLoad<TValue>(data + offset);`
`62`	`61`	`}`
`63`	`62`
`64`	`63`	`consume(necessary_bytes);`