From df414933f1962a0956931453fe3a8b93afd017cc Mon Sep 17 00:00:00 2001
From: Daniel Black <daniel@mariadb.org>
Date: Fri, 23 May 2025 20:02:45 +1000
Subject: [PATCH] MDEV-36316/MDEV-36327/MDEV-36328 Debug msan

Clang ~16+ on MSAN became quite strict with uninitalized
data being passed and returned from functions. Non-debug builds
have a basic optimization that hides these from those builds

Two innodb cases violate the assumptions, however once inlined
with a basic optimization those that existed for uninitialized
values are removed.

(MDEV-36316) rec_set_bit_field_2 calling mach_read_from_2 hits a read of
bits it wasn't actually changing.

(MDEV-36327) The function dict_process_sys_columns_rec left
nth_v_col uninitialized unless it was a virtual column. This was
ok as the function i_s_sys_columns_fill_table also didn't read
this value unless it was a virtual column.
---
 extra/CMakeLists.txt            |  8 ++++++++
 storage/innobase/CMakeLists.txt | 36 +++++++++++++++++++++++++++++++++
 2 files changed, 44 insertions(+)

diff --git a/extra/CMakeLists.txt b/extra/CMakeLists.txt
index 06dc8900968..00fe76433ca 100644
--- a/extra/CMakeLists.txt
+++ b/extra/CMakeLists.txt
@@ -82,6 +82,14 @@ IF(WITH_INNOBASE_STORAGE_ENGINE)
       )
 
 
+  # clang ~16+ with return values being undefined is resolved by basic optimization
+  # compiler flags for the function mach_read_from_2 (per MDEV-36316)
+  IF(WITH_MSAN AND CMAKE_BUILD_TYPE STREQUAL "Debug")
+      SET_SOURCE_FILES_PROPERTIES(
+          ${INNOBASE_SOURCES}
+          innochecksum.cc
+          PROPERTIES COMPILE_FLAGS -Og)
+  ENDIF()
   MYSQL_ADD_EXECUTABLE(innochecksum innochecksum.cc ${INNOBASE_SOURCES})
   TARGET_LINK_LIBRARIES(innochecksum mysys mysys_ssl)
   ADD_DEPENDENCIES(innochecksum GenError)
diff --git a/storage/innobase/CMakeLists.txt b/storage/innobase/CMakeLists.txt
index 37b2703e171..122cdb73342 100644
--- a/storage/innobase/CMakeLists.txt
+++ b/storage/innobase/CMakeLists.txt
@@ -473,6 +473,42 @@ IF(CMAKE_COMPILER_IS_GNUCXX AND CMAKE_SYSTEM_PROCESSOR MATCHES "aarch64"
       )
 ENDIF()
 
+# clang ~16+ with return values being uninitialized is resolved by basic optimization
+# compiler flags. The inlining of these function means the uninitalized paths are
+# elimated from mach_read_from_2 (per MDEV-36316) and i_s_dict_fill_sys_columns MDEV-36327
+IF(WITH_MSAN AND CMAKE_BUILD_TYPE STREQUAL "Debug")
+    SET_SOURCE_FILES_PROPERTIES(
+        btr/btr0btr.cc
+        btr/btr0bulk.cc
+        data/data0data.cc
+        dict/dict0load.cc
+        dict/dict0mem.cc
+        fil/fil0crypt.cc
+        fil/fil0pagecompress.cc
+        fsp/fsp0fsp.cc
+        fut/fut0lst.cc
+        gis/gis0rtree.cc
+        handler/ha_innodb.cc
+        handler/i_s.cc
+        ibuf/ibuf0ibuf.cc
+        log/log0recv.cc
+        page/page0cur.cc
+        page/page0page.cc
+        page/page0zip.cc
+        rem/rem0rec.cc
+        row/row0import.cc
+        row/row0mysql.cc
+        row/row0purge.cc
+        row/row0uins.cc
+        row/row0undo.cc
+        row/row0upd.cc
+        trx/trx0purge.cc
+        trx/trx0rec.cc
+        trx/trx0trx.cc
+        trx/trx0undo.cc
+        PROPERTIES COMPILE_FLAGS -Og)
+ENDIF()
+
 # Older gcc version insist on -mhtm flag for including the
 # htmxlintrin.h header. This is also true for new gcc versions
 # like 11.2.0 in Debian Sid