From 43296460ff4d7ec1c914e27ddbba5a63c197fb00 Mon Sep 17 00:00:00 2001
From: wyb <wybb86@gmail.com>
Date: Thu, 8 Aug 2024 05:45:37 +0800
Subject: [PATCH] [BugFix] Fix orc tinyint on aarch64 (#49517)

Why I'm doing:
https://developer.arm.com/documentation/den0013/d/Porting/Miscellaneous-C-porting-issues/unsigned-char-and-signed-char
compiler treat char as unsigned on aarch64, negative tinyint will be wrong, and the query result from orc is null.

What I'm doing:
use int8_t not char to cast

Signed-off-by: wyb <wybb86@gmail.com>
(cherry picked from commit 6de3ab7f7199e5f2cf1eaa3e95088245a4cda49d)
---
 .../orc/apache-orc/c++/src/ColumnReader.cc    |   4 ++-
 test/sql/test_files/R/orc_tinyint_aarch64     |  27 ++++++++++++++++++
 test/sql/test_files/T/orc_tinyint_aarch64     |  14 +++++++++
 test/sql/test_files/orc_format/tinyint.orc    | Bin 0 -> 247 bytes
 4 files changed, 44 insertions(+), 1 deletion(-)
 create mode 100644 test/sql/test_files/R/orc_tinyint_aarch64
 create mode 100644 test/sql/test_files/T/orc_tinyint_aarch64
 create mode 100644 test/sql/test_files/orc_format/tinyint.orc

diff --git a/be/src/formats/orc/apache-orc/c++/src/ColumnReader.cc b/be/src/formats/orc/apache-orc/c++/src/ColumnReader.cc
index 35f35af941f05..50362043a561f 100644
--- a/be/src/formats/orc/apache-orc/c++/src/ColumnReader.cc
+++ b/be/src/formats/orc/apache-orc/c++/src/ColumnReader.cc
@@ -141,7 +141,9 @@ void ColumnReader::seekToRowGroup(PositionProviderMap* positions) {
    */
 void expandBytesToLongs(int64_t* buffer, uint64_t numValues) {
     for (size_t i = numValues - 1; i < numValues; --i) {
-        buffer[i] = reinterpret_cast<char*>(buffer)[i];
+        // compiler treat char as unsigned on aarch64, negative tinyint will be wrong,
+        // so use 'int8_t' not 'char' to cast
+        buffer[i] = reinterpret_cast<int8_t*>(buffer)[i];
     }
 }
 
diff --git a/test/sql/test_files/R/orc_tinyint_aarch64 b/test/sql/test_files/R/orc_tinyint_aarch64
new file mode 100644
index 0000000000000..21c33d70ab0a0
--- /dev/null
+++ b/test/sql/test_files/R/orc_tinyint_aarch64
@@ -0,0 +1,27 @@
+-- name: orc_tinyint_aarch64
+
+create database db_${uuid0};
+use db_${uuid0};
+
+shell: ossutil64 mkdir oss://${oss_bucket}/test_files/orc_format/${uuid0} >/dev/null || echo "exit 0" >/dev/null
+
+shell: ossutil64 cp --force ./sql/test_files/orc_format/tinyint.orc oss://${oss_bucket}/test_files/orc_format/${uuid0}/ | grep -Pv "(average|elapsed)"
+-- result:
+0
+
+Succeed: Total num: 1, size: 247. OK num: 1(upload 1 files).
+-- !result
+
+select * from files('path' = 'oss://${oss_bucket}/test_files/orc_format/${uuid0}/*', 'format' = 'orc');
+-- result:
+-128
+-- !result
+
+create table t1 as select * from files('path' = 'oss://${oss_bucket}/test_files/orc_format/${uuid0}/*', 'format' = 'orc');
+
+desc t1;
+-- result:
+k1	tinyint	YES	true	None	
+-- !result
+
+shell: ossutil64 rm -rf oss://${oss_bucket}/test_files/orc_format/${uuid0}/ > /dev/null
diff --git a/test/sql/test_files/T/orc_tinyint_aarch64 b/test/sql/test_files/T/orc_tinyint_aarch64
new file mode 100644
index 0000000000000..c59259b8608f4
--- /dev/null
+++ b/test/sql/test_files/T/orc_tinyint_aarch64
@@ -0,0 +1,14 @@
+-- name: orc_tinyint_aarch64
+
+create database db_${uuid0};
+use db_${uuid0};
+
+shell: ossutil64 mkdir oss://${oss_bucket}/test_files/orc_format/${uuid0} >/dev/null || echo "exit 0" >/dev/null
+shell: ossutil64 cp --force ./sql/test_files/orc_format/tinyint.orc oss://${oss_bucket}/test_files/orc_format/${uuid0}/ | grep -Pv "(average|elapsed)"
+
+select * from files('path' = 'oss://${oss_bucket}/test_files/orc_format/${uuid0}/*', 'format' = 'orc');
+
+create table t1 as select * from files('path' = 'oss://${oss_bucket}/test_files/orc_format/${uuid0}/*', 'format' = 'orc');
+desc t1;
+
+shell: ossutil64 rm -rf oss://${oss_bucket}/test_files/orc_format/${uuid0}/ > /dev/null
diff --git a/test/sql/test_files/orc_format/tinyint.orc b/test/sql/test_files/orc_format/tinyint.orc
new file mode 100644
index 0000000000000000000000000000000000000000..dc435b12d46d8264d4bea95a9cb11f4b0f95aa7f
GIT binary patch
literal 247
zcmZ9Gy$ZrG6ov0iOSuNY5Rk?$N~bDVszrxRf`cOXgZKhkDvF?xisBZ00w300i``t#
zcb4<BT#vaUZNe4sI%fc~Mkp&}m@rHkwt|o04mko8{fNpfhyoQWVj!SY&zj6P+~sA{
ze+OA&=J7(CinZDS`ZcObM30?%EeIe2p9o9W3AxIzTdgnYDG%@Zc+GTp%XEIcoTc39
u?KwSk<1p^-d&79p&mv5Io5}y?=(U(4J&NPT*gk`5(1@Jp+p?DHvHJynDkla2

literal 0
HcmV?d00001