[VL] Fix git version conflict issue in build script (apache#4007)

.github/workflows/velox_be.yml

@@ -488,7 +488,7 @@ jobs:
           cd /opt/gluten && \
           sudo -E ./dev/vcpkg/setup-build-depends.sh && \
           source ./dev/vcpkg/env.sh && \
-          ./dev/builddeps-veloxbe.sh --run_setup_script=OFF --build_tests=ON --build_benchmarks=ON --enable_s3=ON --enable_gcs=ON --enable_hdfs=ON --enable_abfs=OFF'
+          ./dev/builddeps-veloxbe.sh --enable_vcpkg=ON --build_tests=ON --build_benchmarks=ON --enable_s3=ON --enable_gcs=ON --enable_hdfs=ON --enable_abfs=OFF'
       - name: Build for Spark 3.2.2
         run: |
           docker exec static-build-test-$GITHUB_RUN_ID bash -c '

dev/builddeps-veloxbe.sh

@@ -24,6 +24,7 @@ ENABLE_ABFS=OFF
 ENABLE_EP_CACHE=OFF
 ARROW_ENABLE_CUSTOM_CODEC=OFF
 ENABLE_VCPKG=OFF
+RUN_SETUP_SCRIPT=ON
 
 for arg in "$@"
 do
@@ -89,6 +90,10 @@ do
         --enable_vcpkg=*)
         ENABLE_VCPKG=("${arg#*=}")
         shift # Remove argument name from processing
+        ;;
+        --run_setup_script=*)
+        RUN_SETUP_SCRIPT=("${arg#*=}")
+        shift # Remove argument name from processing
         ;;
 	      *)
         OTHER_ARGUMENTS+=("$1")
@@ -106,8 +111,8 @@ fi
 ##install velox
 cd $GLUTEN_DIR/ep/build-velox/src
 ./get_velox.sh --enable_hdfs=$ENABLE_HDFS --build_protobuf=$BUILD_PROTOBUF --enable_s3=$ENABLE_S3 --enable_gcs=$ENABLE_GCS --enable_abfs=$ENABLE_ABFS
-./build_velox.sh --enable_s3=$ENABLE_S3 --enable_gcs=$ENABLE_GCS --build_type=$BUILD_TYPE --enable_hdfs=$ENABLE_HDFS --enable_abfs=$ENABLE_ABFS \
-               --enable_ep_cache=$ENABLE_EP_CACHE --build_tests=$BUILD_TESTS --build_benchmarks=$BUILD_BENCHMARKS
+./build_velox.sh --run_setup_script=$RUN_SETUP_SCRIPT --enable_s3=$ENABLE_S3 --enable_gcs=$ENABLE_GCS --build_type=$BUILD_TYPE --enable_hdfs=$ENABLE_HDFS \
+                 --enable_abfs=$ENABLE_ABFS --enable_ep_cache=$ENABLE_EP_CACHE --build_tests=$BUILD_TESTS --build_benchmarks=$BUILD_BENCHMARKS
 
 ## compile gluten cpp
 cd $GLUTEN_DIR/cpp

ep/build-velox/src/build_velox.sh

@@ -90,7 +90,7 @@ function compile {
     elif [ $OS == 'Darwin' ]; then
       setup_macos
     else
-      echo "Unsupport kernel: $OS"
+      echo "Unsupported kernel: $OS"
       exit 1
     fi
   fi
@@ -126,7 +126,7 @@ function compile {
   elif [[ "$ARCH" == 'arm64' || "$ARCH" == 'aarch64' ]]; then
     CPU_TARGET=$ARCH make $COMPILE_TYPE EXTRA_CMAKE_FLAGS="${COMPILE_OPTION}"
   else
-    echo "Unsupport arch: $ARCH"
+    echo "Unsupported arch: $ARCH"
     exit 1
   fi
 
@@ -208,7 +208,7 @@ function setup_linux {
       set -u
       ;;
     *)
-      echo "Unsupport centos version: $LINUX_VERSION_ID"
+      echo "Unsupported centos version: $LINUX_VERSION_ID"
       exit 1
       ;;
     esac
@@ -223,20 +223,20 @@ function setup_linux {
       ;;
     3) scripts/setup-centos8.sh ;;
     *)
-      echo "Unsupport alinux version: $LINUX_VERSION_ID"
+      echo "Unsupported alinux version: $LINUX_VERSION_ID"
       exit 1
       ;;
     esac
   elif [[ "$LINUX_DISTRIBUTION" == "tencentos" ]]; then
     case "$LINUX_VERSION_ID" in
     3.2) scripts/setup-centos8.sh ;;
     *)
-      echo "Unsupport tencentos version: $LINUX_VERSION_ID"
+      echo "Unsupported tencentos version: $LINUX_VERSION_ID"
       exit 1
       ;;
     esac
   else
-    echo "Unsupport linux distribution: $LINUX_DISTRIBUTION"
+    echo "Unsupported linux distribution: $LINUX_DISTRIBUTION"
     exit 1
   fi
 }

ep/build-velox/src/get_velox.sh

@@ -75,8 +75,16 @@ for arg in "$@"; do
 done
 
 function process_setup_ubuntu {
-  # make this function Reentrantly
+  if [ -z "$(which git)" ]; then
+    sudo --preserve-env apt install -y git
+  fi
+  # make this function Reentrant
   git checkout scripts/setup-ubuntu.sh
+
+  # No need to re-install git.
+  sed -i '/git \\/d' scripts/setup-ubuntu.sh
+  # Do not install libunwind which can cause interruption when catching native exception.
+  sed -i 's/sudo --preserve-env apt install -y libunwind-dev && //' scripts/setup-ubuntu.sh
   sed -i '/libprotobuf-dev/d' scripts/setup-ubuntu.sh
   sed -i '/protobuf-compiler/d' scripts/setup-ubuntu.sh
   sed -i '/ccache/a\  *thrift* \\' scripts/setup-ubuntu.sh
@@ -113,8 +121,15 @@ function process_setup_ubuntu {
 }
 
 function process_setup_centos8 {
-  # make this function Reentrantly
+  # Allows other version of git already installed.
+  if [ -z "$(which git)" ]; then
+    dnf install -y -q --setopt=install_weak_deps=False git
+  fi
+  # make this function Reentrant
   git checkout scripts/setup-centos8.sh
+
+  # # No need to re-install git.
+  sed -i 's/dnf_install ninja-build ccache gcc-toolset-9 git/dnf_install ninja-build ccache gcc-toolset-9/' scripts/setup-centos8.sh
   sed -i '/^function dnf_install/i\DEPENDENCY_DIR=${DEPENDENCY_DIR:-$(pwd)}' scripts/setup-centos8.sh
   sed -i '/^dnf_install autoconf/a\dnf_install libxml2-devel libgsasl-devel libuuid-devel' scripts/setup-centos8.sh
   sed -i '/^function cmake_install_deps.*/i FB_OS_VERSION=v2022.11.14.00\n function install_folly {\n  github_checkout facebook/folly "${FB_OS_VERSION}"\n  cmake_install -DBUILD_TESTS=OFF -DFOLLY_HAVE_INT128_T=ON\n}\n'     scripts/setup-centos8.sh
@@ -144,19 +159,21 @@ function process_setup_centos8 {
 }
 
 function process_setup_centos7 {
-  # make this function Reentrantly
+  # Allows other version of git already installed.
+  if [ -z "$(which git)" ]; then
+    dnf install -y -q --setopt=install_weak_deps=False git
+  fi
+  # make this function Reentrant
   git checkout scripts/setup-centos7.sh
 
+  # No need to re-install git.
+  sed -i 's/dnf_install ccache git/dnf_install ccache/' scripts/setup-centos7.sh
   # cmake 3 and ninja should be installed
   sed -i '/^run_and_time install_cmake/d' scripts/setup-centos7.sh
   sed -i '/^run_and_time install_ninja/d' scripts/setup-centos7.sh
 
   # install gtest
   sed -i '/^  run_and_time install_fmt/a \ \ run_and_time install_gtest' scripts/setup-centos7.sh
-  # Don't force to install git.
-  sed -i 's/dnf_install ccache git/dnf_install ccache/' scripts/setup-centos7.sh
-  # Install git with --skip-broken to accept other version of git already installed.
-  sed -i '/^$SUDO dnf remove -y gflags/a \\dnf install -y -q --setopt=install_weak_deps=False --skip-broken git' scripts/setup-centos7.sh
 
   if [ $ENABLE_HDFS = "ON" ]; then
     sed -i '/^function install_protobuf.*/i function install_libhdfs3 {\n cd "\${DEPENDENCY_DIR}"\n github_checkout oap-project/libhdfs3 master \n cmake_install\n}\n' scripts/setup-centos7.sh

ep/build-velox/src/modify_velox.patch

@@ -79,19 +79,6 @@ index e2ff67fa5..e03b246e3 100644
  if(NOT TARGET gflags::gflags)
    # This is a bit convoluted, but we want to be able to use gflags::gflags as a
    # target even when velox is built as a subproject which uses
-diff --git a/scripts/setup-ubuntu.sh b/scripts/setup-ubuntu.sh
-index 42054e593..4c793b785 100755
---- a/scripts/setup-ubuntu.sh
-+++ b/scripts/setup-ubuntu.sh
-@@ -31,7 +31,7 @@ export CMAKE_BUILD_TYPE=Release
- # Install all velox and folly dependencies. 
- # The is an issue on 22.04 where a version conflict prevents glog install,
- # installing libunwind first fixes this.
--sudo --preserve-env apt update && sudo --preserve-env apt install -y libunwind-dev && \
-+sudo --preserve-env apt update && \
-   sudo --preserve-env apt install -y \
-   g++ \
-   cmake \
 diff --git a/third_party/CMakeLists.txt b/third_party/CMakeLists.txt
 index c7c4d8557..2715e625c 100644
 --- a/third_party/CMakeLists.txt