[Improvement] support build thirdparty and BE using clang

be/CMakeLists.txt

@@ -18,13 +18,6 @@
 cmake_minimum_required(VERSION 3.19.2)
 
 # set CMAKE_C_COMPILER, this must set before project command
-if (DEFINED ENV{DORIS_GCC_HOME})
-    set(CMAKE_C_COMPILER "$ENV{DORIS_GCC_HOME}/bin/gcc")
-    set(CMAKE_CXX_COMPILER "$ENV{DORIS_GCC_HOME}/bin/g++")
-    set(GCC_HOME $ENV{DORIS_GCC_HOME})
-else()
-    message(FATAL_ERROR "DORIS_GCC_HOME environment variable is not set")
-endif()
 
 project(doris CXX C)
 
@@ -89,16 +82,6 @@ endif()
 message(STATUS "make test: ${MAKE_TEST}")
 option(WITH_MYSQL "Support access MySQL" ON)
 
-# Check gcc
-if (CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
-    if (CMAKE_CXX_COMPILER_VERSION VERSION_LESS "7.3.0")
-        message(FATAL_ERROR "Need GCC version at least 7.3.0")
-    endif()
-    message(STATUS "GCC version is greater than 7.3.0, disable -Werror. Be careful with compile warnings.")
-elseif (NOT APPLE)
-    message(FATAL_ERROR "Compiler should be GNU")
-endif()
-
 set(PIC_LIB_PATH "${THIRDPARTY_DIR}")
 if(PIC_LIB_PATH)
     message(STATUS "defined PIC_LIB_PATH")

be/src/exec/olap_utils.h

@@ -75,6 +75,8 @@ inline CompareLargeFunc get_compare_func(PrimitiveType type) {
 
     default:
         DCHECK(false) << "Unsupported Compare type";
+        // Clang will report error if not all path have return value
+        return nullptr;
     }
 }
 

be/src/olap/column_vector.cpp

@@ -164,9 +164,6 @@ ScalarColumnVectorBatch<ScalarType>::ScalarColumnVectorBatch(const TypeInfo* typ
                                                              bool is_nullable)
         : ColumnVectorBatch(type_info, is_nullable), _data(0) {}
 
-template <class ScalarType>
-ScalarColumnVectorBatch<ScalarType>::~ScalarColumnVectorBatch() = default;
-
 template <class ScalarType>
 Status ScalarColumnVectorBatch<ScalarType>::resize(size_t new_cap) {
     if (capacity() < new_cap) { // before first init, _capacity is 0.
@@ -222,35 +219,4 @@ void ArrayColumnVectorBatch::prepare_for_read(size_t start_idx, size_t size, boo
     }
 }
 
-template <class T>
-DataBuffer<T>::DataBuffer(size_t new_size) : buf(nullptr), current_size(0), current_capacity(0) {
-    resize(new_size);
-}
-
-template <class T>
-DataBuffer<T>::~DataBuffer() {
-    for (uint64_t i = current_size; i > 0; --i) {
-        (buf + i - 1)->~T();
-    }
-    if (buf) {
-        std::free(buf);
-    }
-}
-
-template <class T>
-void DataBuffer<T>::resize(size_t new_size) {
-    if (new_size > current_capacity || !buf) {
-        if (buf) {
-            T* buf_old = buf;
-            buf = reinterpret_cast<T*>(std::malloc(sizeof(T) * new_size));
-            memcpy(buf, buf_old, sizeof(T) * current_size);
-            std::free(buf_old);
-        } else {
-            buf = reinterpret_cast<T*>(std::malloc(sizeof(T) * new_size));
-        }
-        current_capacity = new_size;
-    }
-    current_size = new_size;
-}
-
 } // namespace doris

be/src/olap/column_vector.h

@@ -51,6 +51,37 @@ class DataBuffer {
     void resize(size_t _size);
 };
 
+template <class T>
+DataBuffer<T>::DataBuffer(size_t new_size) : buf(nullptr), current_size(0), current_capacity(0) {
+    resize(new_size);
+}
+
+template <class T>
+DataBuffer<T>::~DataBuffer() {
+    for (uint64_t i = current_size; i > 0; --i) {
+        (buf + i - 1)->~T();
+    }
+    if (buf) {
+        std::free(buf);
+    }
+}
+
+template <class T>
+void DataBuffer<T>::resize(size_t new_size) {
+    if (new_size > current_capacity || !buf) {
+        if (buf) {
+            T* buf_old = buf;
+            buf = reinterpret_cast<T*>(std::malloc(sizeof(T) * new_size));
+            memcpy(buf, buf_old, sizeof(T) * current_size);
+            std::free(buf_old);
+        } else {
+            buf = reinterpret_cast<T*>(std::malloc(sizeof(T) * new_size));
+        }
+        current_capacity = new_size;
+    }
+    current_size = new_size;
+}
+
 template class DataBuffer<bool>;
 template class DataBuffer<int8_t>;
 template class DataBuffer<int16_t>;
@@ -161,6 +192,9 @@ class ScalarColumnVectorBatch : public ColumnVectorBatch {
     DataBuffer<ScalarCppType> _data;
 };
 
+template <class ScalarType>
+ScalarColumnVectorBatch<ScalarType>::~ScalarColumnVectorBatch() = default;
+
 // util class for read array's null signs.
 class ArrayNullColumnVectorBatch : public ColumnVectorBatch {
 public:

be/src/olap/types.h

@@ -420,7 +420,7 @@ struct CppTypeTraits<OLAP_FIELD_TYPE_UNSIGNED_BIGINT> {
 template <>
 struct CppTypeTraits<OLAP_FIELD_TYPE_LARGEINT> {
     using CppType = int128_t;
-    using UnsignedCppType = unsigned int128_t;
+    using UnsignedCppType = uint128_t;
 };
 template <>
 struct CppTypeTraits<OLAP_FIELD_TYPE_FLOAT> {

be/src/runtime/dpp_sink_internal.cpp

@@ -31,11 +31,12 @@ namespace doris {
 PartRangeKey PartRangeKey::_s_pos_infinite(1);
 PartRangeKey PartRangeKey::_s_neg_infinite(-1);
 
+// Clang will check constructor method not exist
 PartRange PartRange::_s_all_range = {
-        ._start_key = PartRangeKey::neg_infinite(),
-        ._end_key = PartRangeKey::pos_infinite(),
-        ._include_start_key = true,
-        ._include_end_key = true,
+        PartRangeKey::neg_infinite(),
+        PartRangeKey::pos_infinite(),
+        true,
+        true
 };
 
 RollupSchema::RollupSchema() {}

be/src/runtime/raw_value_ir.cpp

@@ -16,6 +16,7 @@
 // under the License.
 
 #include "runtime/raw_value.h"
+#include "runtime/string_value.hpp"
 #include "util/types.h"
 
 namespace doris {

be/src/util/bitmap_value.h

@@ -1734,10 +1734,6 @@ class BitmapValue {
 //  }
 class BitmapValueIterator {
 public:
-    BitmapValueIterator()
-       : _bitmap(BitmapValue()) {
-    }
-
     BitmapValueIterator(const BitmapValue& bitmap, bool end = false)
         : _bitmap(bitmap), _end(end) {
 

build.sh

@@ -38,10 +38,25 @@ set -eo pipefail
 ROOT=`dirname "$0"`
 ROOT=`cd "$ROOT"; pwd`
 
+
 export DORIS_HOME=${ROOT}
 
+#export CC="/data/common/clang-11/bin/clang-wrapper"
+#export CXX="/data/common/clang-11/bin/clang++-wrapper"
+
+#export CC="/usr/local/gcc-10.1.0/bin/gcc"
+#export CXX="/usr/local/gcc-10.1.0/bin/g++"
+
 . ${DORIS_HOME}/env.sh
 
+if [[ -z ${CC} ]]; then
+    if [[ -z ${DORIS_GCC_HOME} ]]; then
+        DORIS_GCC_HOME=$(dirname $(which gcc))/..
+    fi
+    export CC="${DORIS_GCC_HOME}/bin/gcc"
+    export CXX="${DORIS_GCC_HOME}/bin/g++"
+fi
+
 # Check args
 usage() {
   echo "

contrib/udf/CMakeLists.txt

@@ -18,13 +18,6 @@
 cmake_minimum_required(VERSION 3.19.2)
 
 # set CMAKE_C_COMPILER, this must set before project command
-if (DEFINED ENV{DORIS_GCC_HOME})
-    set(CMAKE_C_COMPILER "$ENV{DORIS_GCC_HOME}/bin/gcc")
-    set(CMAKE_CXX_COMPILER "$ENV{DORIS_GCC_HOME}/bin/g++")
-    set(GCC_HOME $ENV{DORIS_GCC_HOME})
-else()
-    message(FATAL_ERROR "DORIS_GCC_HOME environment variable is not set")
-endif()
 
 project(doris_udf)
 

env.sh

@@ -51,30 +51,25 @@ if ! ${PYTHON} --version; then
     fi
 fi
 
-# set GCC HOME
-if [[ -z ${DORIS_GCC_HOME} ]]; then
-    export DORIS_GCC_HOME=$(dirname $(which gcc))/..
-fi
-
-gcc_ver=$(${DORIS_GCC_HOME}/bin/gcc -dumpfullversion -dumpversion)
-required_ver="7.3.0"
-if [[ ! "$(printf '%s\n' "$required_ver" "$gcc_ver" | sort -V | head -n1)" = "$required_ver" ]]; then
-    echo "Error: GCC version (${gcc_ver}) must be greater than or equal to ${required_ver}"
+# register keyword is forbidden to use in C++17
+# the C++ code generated by flex that remove register keyword after version 2.6.0
+# so we need check flex version here to avoid compilation failed
+flex_ver=$(flex --version | awk '{print $2}')
+required_ver="2.6.0"
+if [[ ! "$(printf '%s\n' "$required_ver" "$flex_ver" | sort -V | head -n1)" = "$required_ver" ]]; then
+    echo "Error: flex version (${flex_ver}) must be greater than or equal to ${required_ver}"
     exit 1
 fi
 
-# find binutils
-if test -x ${DORIS_GCC_HOME}/bin/ld; then
-    export DORIS_BIN_UTILS=${DORIS_GCC_HOME}/bin/
+# check java home
+if [ -z "$JAVA_HOME" ]; then
+    export JAVACMD=$(which java)
+    JAVAP=$(which javap)
 else
-    export DORIS_BIN_UTILS=/usr/bin/
+    export JAVA="${JAVA_HOME}/bin/java"
+    JAVAP="${JAVA_HOME}/bin/javap"
 fi
 
-# export CLANG COMPATIBLE FLAGS
-export CLANG_COMPATIBLE_FLAGS=$(echo | ${DORIS_GCC_HOME}/bin/gcc -Wp,-v -xc++ - -fsyntax-only 2>&1 |
-    grep -E '^\s+/' | awk '{print "-I" $1}' | tr '\n' ' ')
-
-
 # if is called from build-thirdparty.sh, no need to check these tools
 if test -z "${BUILD_THIRDPARTY_WIP}"; then
     # register keyword is forbidden to use in C++17