Fixes to build with CMake 3.2. Document what features require newer CMake in

the README. Add Docker task for CMake 3.2 from kszucs

cpp/CMakeLists.txt

@@ -16,6 +16,7 @@
 # under the License.
 
 cmake_minimum_required(VERSION 3.2)
+message(STATUS "Building using CMake version: ${CMAKE_VERSION}")
 
 # Extract Arrow version number
 file(READ "${CMAKE_CURRENT_SOURCE_DIR}/../java/pom.xml" POM_XML)
@@ -436,11 +437,12 @@ endif()
 ############################################################
 
 if(ARROW_BUILD_TESTS OR ARROW_BUILD_BENCHMARKS)
+  # Currently the compression tests require at least these libraries; bz2 and
+  # zstd are optional. See ARROW-3984
   set(ARROW_WITH_BROTLI ON)
   set(ARROW_WITH_LZ4 ON)
   set(ARROW_WITH_SNAPPY ON)
   set(ARROW_WITH_ZLIB ON)
-  set(ARROW_WITH_ZSTD ON)
 endif()
 
 if(ARROW_BUILD_TESTS)

cpp/Dockerfile

@@ -30,6 +30,7 @@ RUN apt-get update -y -q && \
         wget
 
 # install conda and required packages
+ARG EXTRA_CONDA_PKGS
 ENV PATH=/opt/conda/bin:$PATH \
     CONDA_PREFIX=/opt/conda
 ADD ci/docker_install_conda.sh \
@@ -39,7 +40,8 @@ ADD ci/docker_install_conda.sh \
 RUN arrow/ci/docker_install_conda.sh && \
     conda install -c conda-forge \
         --file arrow/ci/conda_env_cpp.yml \
-        --file arrow/ci/conda_env_unix.yml && \
+        --file arrow/ci/conda_env_unix.yml \
+        $EXTRA_CONDA_PKGS && \
     conda clean --all
 
 ENV CC=gcc \

cpp/README.md

@@ -30,7 +30,7 @@ in-source and out-of-source builds with the latter one being preferred.
 Building Arrow requires:
 
 * A C++11-enabled compiler. On Linux, gcc 4.8 and higher should be sufficient.
-* CMake
+* CMake 3.2 or higher
 * Boost
 
 On Ubuntu/Debian you can install the requirements with:
@@ -459,6 +459,14 @@ both of these options would be used rarely. Current known uses-cases when they a
 
 *  Parameterized tests in google test.
 
+## CMake version requirements
+
+We support CMake 3.2 and higher. Some features require a newer version of CMake:
+
+* Building the benchmarks requires 3.6 or higher
+* Building zstd from source requires 3.7 or higher
+* Building Gandiva JNI bindings requires 3.11 or higher
+
 [1]: https://brew.sh/
 [2]: https://github.com/apache/arrow/blob/master/cpp/apidoc/Windows.md
 [3]: https://google.github.io/styleguide/cppguide.html

cpp/cmake_modules/ThirdpartyToolchain.cmake

@@ -407,6 +407,13 @@ else()
     # disable autolinking in boost
     add_definitions(-DBOOST_ALL_NO_LIB)
   endif()
+
+  if (DEFINED ENV{BOOST_ROOT} OR DEFINED BOOST_ROOT)
+    # In older versions of CMake (such as 3.2), the system paths for Boost will
+    # be looked in first even if we set $BOOST_ROOT or pass -DBOOST_ROOT
+    set(Boost_NO_SYSTEM_PATHS ON)
+  endif()
+
   if (ARROW_BOOST_USE_SHARED)
     # Find shared Boost libraries.
     set(Boost_USE_STATIC_LIBS OFF)
@@ -629,8 +636,11 @@ if(ARROW_BUILD_TESTS OR ARROW_GANDIVA_BUILD_TESTS
 endif()
 
 if(ARROW_BUILD_BENCHMARKS)
-
   if("$ENV{GBENCHMARK_HOME}" STREQUAL "")
+    if(CMAKE_VERSION VERSION_LESS 3.6)
+      message(FATAL_ERROR "Building gbenchmark from source requires at least CMake 3.6")
+    endif()
+
     if(NOT MSVC)
       set(GBENCHMARK_CMAKE_CXX_FLAGS "-fPIC -std=c++11 ${EP_CXX_FLAGS}")
     endif()
@@ -1095,6 +1105,11 @@ if (ARROW_WITH_ZSTD)
           "-DCMAKE_C_FLAGS=${EP_C_FLAGS}")
     endif()
 
+    if(CMAKE_VERSION VERSION_LESS 3.7)
+      message(FATAL_ERROR "Building zstd using ExternalProject requires \
+at least CMake 3.7")
+    endif()
+
     ExternalProject_Add(zstd_ep
       ${EP_LOG_OPTIONS}
       CMAKE_ARGS ${ZSTD_CMAKE_ARGS}

cpp/src/arrow/util/compression-test.cc

@@ -448,17 +448,22 @@ TEST_P(CodecTest, StreamingRoundtrip) {
 
 INSTANTIATE_TEST_CASE_P(TestGZip, CodecTest, ::testing::Values(Compression::GZIP));
 
-INSTANTIATE_TEST_CASE_P(TestZSTD, CodecTest, ::testing::Values(Compression::ZSTD));
-
 INSTANTIATE_TEST_CASE_P(TestSnappy, CodecTest, ::testing::Values(Compression::SNAPPY));
 
 INSTANTIATE_TEST_CASE_P(TestLZ4, CodecTest, ::testing::Values(Compression::LZ4));
 
 INSTANTIATE_TEST_CASE_P(TestBrotli, CodecTest, ::testing::Values(Compression::BROTLI));
 
+// bz2 requires a binary installation, there is no ExternalProject
 #if ARROW_WITH_BZ2
 INSTANTIATE_TEST_CASE_P(TestBZ2, CodecTest, ::testing::Values(Compression::BZ2));
 #endif
 
+// The ExternalProject for zstd does not build on CMake < 3.7, so we do not
+// require it here
+#ifdef ARROW_WITH_ZSTD
+INSTANTIATE_TEST_CASE_P(TestZSTD, CodecTest, ::testing::Values(Compression::ZSTD));
+#endif
+
 }  // namespace util
 }  // namespace arrow

cpp/src/gandiva/jni/CMakeLists.txt

@@ -17,6 +17,10 @@
 
 project(gandiva_jni)
 
+if(CMAKE_VERSION VERSION_LESS 3.11)
+  message(FATAL_ERROR "Building the Gandiva JNI bindings requires CMake version >= 3.11")
+endif()
+
 # Find JNI
 find_package(JNI REQUIRED)
 

cpp/src/plasma/CMakeLists.txt

@@ -20,16 +20,12 @@ add_custom_target(plasma)
 # For the moment, Plasma is versioned like Arrow
 project(plasma VERSION "${ARROW_BASE_VERSION}")
 
-set(CMAKE_MODULE_PATH ${CMAKE_MODULE_PATH} "${CMAKE_SOURCE_DIR}/../python/cmake_modules")
-
-find_package(PythonLibsNew REQUIRED)
 find_package(Threads)
 
 # The SO version is also the ABI version
 set(PLASMA_SO_VERSION "${ARROW_SO_VERSION}")
 set(PLASMA_FULL_SO_VERSION "${ARROW_FULL_SO_VERSION}")
 
-include_directories(SYSTEM ${PYTHON_INCLUDE_DIRS})
 include_directories("${FLATBUFFERS_INCLUDE_DIR}" "${CMAKE_CURRENT_LIST_DIR}/" "${CMAKE_CURRENT_LIST_DIR}/thirdparty/" "${CMAKE_CURRENT_LIST_DIR}/../")
 
 set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -D_XOPEN_SOURCE=500 -D_POSIX_C_SOURCE=200809L")

cpp/thirdparty/download_dependencies.sh

@@ -38,7 +38,7 @@ download_dependency() {
 
   # --show-progress will not output to stdout, it is safe to pipe the result of
   # the script into eval.
-  wget --quiet --show-progress --continue --output-document="${out}" "${url}"
+  wget --quiet --continue --output-document="${out}" "${url}"
 }
 
 main() {

cpp/thirdparty/versions.txt

@@ -61,6 +61,7 @@ DEPENDENCIES=(
   "ARROW_ORC_URL orc-${ORC_VERSION}.tar.gz https://github.com/apache/orc/archive/rel/release-${ORC_VERSION}.tar.gz"
   "ARROW_PROTOBUF_URL protobuf-${PROTOBUF_VERSION}.tar.gz https://github.com/google/protobuf/releases/download/${PROTOBUF_VERSION}/protobuf-all-${PROTOBUF_VERSION:1}.tar.gz"
   "ARROW_RAPIDJSON_URL rapidjson-${RAPIDJSON_VERSION}.tar.gz https://github.com/miloyip/rapidjson/archive/${RAPIDJSON_VERSION}.tar.gz"
+  "ARROW_RE2_URL re2-${RE2_VERSION}.tar.gz https://github.com/google/re2/archive/${RE2_VERSION}.tar.gz"
   "ARROW_SNAPPY_URL snappy-${SNAPPY_VERSION}.tar.gz https://github.com/google/snappy/releases/download/${SNAPPY_VERSION}/snappy-${SNAPPY_VERSION}.tar.gz"
   "ARROW_THRIFT_URL thrift-${THRIFT_VERSION}.tar.gz http://archive.apache.org/dist/thrift/${THRIFT_VERSION}/thrift-${THRIFT_VERSION}.tar.gz"
   "ARROW_ZLIB_URL zlib-${ZLIB_VERSION}.tar.gz http://zlib.net/fossils/zlib-${ZLIB_VERSION}.tar.gz"

dev/tasks/tests.yml

@@ -22,6 +22,7 @@ groups:
     - docker-rust
     - docker-cpp
     - docker-cpp-alpine
+    - docker-cpp-cmake32
     - docker-c_glib
     - docker-go
     - docker-python-2.7
@@ -45,6 +46,7 @@ groups:
   cpp-python:
     - docker-cpp
     - docker-cpp-alpine
+    - docker-cpp-cmake32
     - docker-python-2.7
     - docker-python-2.7-alpine
     - docker-python-3.6
@@ -87,6 +89,14 @@ tasks:
         - docker-compose build cpp-alpine
         - docker-compose run cpp-alpine
 
+  docker-cpp-cmake32:
+    platform: linux
+    template: docker-tests/travis.linux.yml
+    params:
+      commands:
+        - docker-compose build cpp-cmake32
+        - docker-compose run cpp-cmake32
+
   docker-c_glib:
     platform: linux
     template: docker-tests/travis.linux.yml

docker-compose.yml

@@ -57,6 +57,22 @@ services:
       PARQUET_TEST_DATA: /arrow/cpp/submodules/parquet-testing/data
     volumes: *ubuntu-volumes
 
+  cpp-cmake32:
+    # Usage:
+    #   docker-compose build cpp-cmake32
+    #   docker-compose run cpp-cmake32
+    image: arrow:cpp-cmake32
+    shm_size: 2G
+    build:
+      context: .
+      dockerfile: cpp/Dockerfile
+      args:
+        EXTRA_CONDA_PKGS: cmake=3.2
+    environment:
+      ARROW_ORC: "OFF"
+      PARQUET_TEST_DATA: /arrow/cpp/submodules/parquet-testing/data
+    volumes: *ubuntu-volumes
+
   cpp-alpine:
     # Usage:
     #   docker-compose build cpp-alpine