[impala] branch master updated: IMPALA-9762: Fix GCC7 shift-count-overflow in tuple-row-compare.cc

[joemcdonnell]

This is an automated email from the ASF dual-hosted git repository.

joemcdonnell pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/impala.git


The following commit(s) were added to refs/heads/master by this push:
     new f474b03  IMPALA-9762: Fix GCC7 shift-count-overflow in 
tuple-row-compare.cc
f474b03 is described below

commit f474b03dce35956d0762159eed16516b793903eb
Author: Joe McDonnell <joemcdonn...@cloudera.com>
AuthorDate: Tue May 19 18:36:48 2020 -0700

    IMPALA-9762: Fix GCC7 shift-count-overflow in tuple-row-compare.cc
    
    This fixes a GCC 7 compilation error for this code in
    TupleRowZOrderComparator's GetSharedIntRepresentation() and
    GetSharedFloatRepresentation():
    
      return (static_cast<U>(val) <<
          std::max((sizeof(U) - sizeof(T)) * 8, (uint64_t)0)) ^ mask;
    
    In this case, the std::max is running with uint64_t arguments. For
    template instatiations with sizeof(T) > sizeof(U), this results
    in integer overflow and a very large positive integer causing
    the shift-count-overflow. These instantiations are not used by
    Impala, but the compiler still needs to generate them.
    
    This changes the logic to use signed integers for the std::max,
    avoiding the shift-count-overflow.
    
    Testing:
     - Build on GCC 4.9.2 and GCC 7
     - Core tests
    
    Change-Id: I518e8bed1bb8d49d9cb76a33b07b665e15dfef87
    Reviewed-on: http://gerrit.cloudera.org:8080/15962
    Reviewed-by: Impala Public Jenkins <impala-public-jenk...@cloudera.com>
    Tested-by: Impala Public Jenkins <impala-public-jenk...@cloudera.com>
---
 be/src/util/tuple-row-compare.cc | 12 +++++++-----
 1 file changed, 7 insertions(+), 5 deletions(-)

diff --git a/be/src/util/tuple-row-compare.cc b/be/src/util/tuple-row-compare.cc
index d960bbe..099fa46 100644
--- a/be/src/util/tuple-row-compare.cc
+++ b/be/src/util/tuple-row-compare.cc
@@ -439,8 +439,9 @@ U TupleRowZOrderComparator::GetSharedRepresentation(void* 
val, ColumnType type)
 
 template <typename U, typename T>
 U inline TupleRowZOrderComparator::GetSharedIntRepresentation(const T val, U 
mask) const {
-  return (static_cast<U>(val) <<
-      std::max((sizeof(U) - sizeof(T)) * 8, (uint64_t)0)) ^ mask;
+  uint64_t shift_size = static_cast<uint64_t>(
+      std::max(static_cast<int64_t>((sizeof(U) - sizeof(T)) * 8), (int64_t) 
0));
+  return (static_cast<U>(val) << shift_size) ^ mask;
 }
 
 template <typename U, typename T>
@@ -449,13 +450,14 @@ U inline 
TupleRowZOrderComparator::GetSharedFloatRepresentation(void* val, U mas
   T floating_value = *reinterpret_cast<const T*>(val);
   memcpy(&tmp, &floating_value, sizeof(T));
   if (UNLIKELY(std::isnan(floating_value))) return 0;
+  uint64_t shift_size = static_cast<uint64_t>(
+      std::max(static_cast<int64_t>((sizeof(U) - sizeof(T)) * 8), (int64_t) 
0));
   if (floating_value < 0.0) {
     // Flipping all bits for negative values.
-    return static_cast<U>(~tmp) << std::max((sizeof(U) - sizeof(T)) * 8, 
(uint64_t)0);
+    return static_cast<U>(~tmp) << shift_size;
   } else {
     // Flipping only first bit.
-    return (static_cast<U>(tmp) << std::max((sizeof(U) - sizeof(T)) * 8, 
(uint64_t)0)) ^
-        mask;
+    return (static_cast<U>(tmp) << shift_size) ^ mask;
   }
 }