commit 2be451b434028f71d4e3620950b914072e4dfec7
author Zehua Zou <zehuazou2000@gmail.com> Thu Oct 16 21:35:56 2025 +0800
committer Gang Wu <ustcwg@gmail.com> Thu Oct 16 21:36:38 2025 +0800
tree 68389098bcab66ec879d35f6f70e091724ed1b2e
parent c4420c2369f15ac465e64b678fbe1af439fffd06

ORC-2027: [C++] Fix undefined behavior in DoubleColumnReader::readFloat()

### What changes were proposed in this pull request?

Unaligned reads are UB in C++. We can not guarantee that the `bufferPointer_` pointer is aligned by `alignof(int32_t)`.

### How was this patch tested?
Use UBsan to test in private repo.

### Was this patch authored or co-authored using generative AI tooling?
No.

Closes #2444 from HuaHuaY/fix_issue_2027.

Lead-authored-by: Zehua Zou <zehuazou2000@gmail.com>
Co-authored-by: Zehua Zou <41586196+HuaHuaY@users.noreply.github.com>
Signed-off-by: Gang Wu <ustcwg@gmail.com>
(cherry picked from commit c4fa9fa3d5b16630f5066d79bf4c36d171a2561d)
Signed-off-by: Gang Wu <ustcwg@gmail.com>

c++/src/ColumnReader.cc

diff --git a/c++/src/ColumnReader.cc b/c++/src/ColumnReader.cc
index af434c3..8c9ddb4 100644
--- a/c++/src/ColumnReader.cc
+++ b/c++/src/ColumnReader.cc
@@ -421,7 +421,7 @@
       int32_t bits = 0;
       if (bufferEnd_ - bufferPointer_ >= 4) {
         if (isLittleEndian) {
-          bits = *(reinterpret_cast<const int32_t*>(bufferPointer_));
+          memcpy(&bits, bufferPointer_, sizeof(bits));
         } else {
           bits = static_cast<unsigned char>(bufferPointer_[0]);
           bits |= static_cast<unsigned char>(bufferPointer_[1]) << 8;