我正在尝试使用 CMake 的外部项目功能在 linux 上构建 ceres-solver。
我有一个我构建的特定版本的 Eigen，其他依赖项是通过我的数据包管理器（BLAS、LAPACK 和 Suitesparse）安装的。
如果我直接从源代码构建 Ceres-solver：
cmake -G Ninja -DEIGEN_INCLUDE_DIR='<path-to-eigen>' -DEIGEN_BUILD_DIR='<path-to-eigen-build>' ..
它会自动找到 suitesparse 及其所有依赖项（BLAS、LAPACK、AMD、COLAMD 等），并且可以毫无问题地构建。
这是 CMake 输出：
-- Detected Ceres version: 2.0.0 from <path>/ceres-solver/include/ceres/version.h
-- Detected available Ceres threading models: [CXX11_THREADS, OPENMP, NO_THREADS]
-- Building with C++17
-- No preference for use of exported Eigen CMake configuration set, and no hints for include directory provided. Defaulting to preferring an installed/exported Eigen CMake configuration if available.
-- Found installed version of Eigen: <path>/eigen-3.3.5/share/eigen3/cmake
-- Found Eigen version 3.3.5: <path>/eigen-3.3.5/include/eigen3
-- Disabling use of Eigen as a sparse linear algebra library.
This does not affect the covariance estimation algorithm
which can still use the EIGEN_SPARSE_QR algorithm.
-- Found LAPACK library: /usr/lib/liblapack.so;/usr/lib/libblas.so
-- Found AMD headers in: /usr/include
-- Found AMD library: /usr/lib/libamd.so
-- Found CAMD headers in: /usr/include
-- Found CAMD library: /usr/lib/libcamd.so
-- Found COLAMD headers in: /usr/include
-- Found COLAMD library: /usr/lib/libcolamd.so
-- Found CCOLAMD headers in: /usr/include
-- Found CCOLAMD library: /usr/lib/libccolamd.so
-- Found CHOLMOD headers in: /usr/include
-- Found CHOLMOD library: /usr/lib/libcholmod.so
-- Found SUITESPARSEQR headers in: /usr/include
-- Found SUITESPARSEQR library: /usr/lib/libspqr.so
-- Found Intel Thread Building Blocks (TBB) library (2019.0) assuming SuiteSparseQR was compiled with TBB.
-- Found SUITESPARSE_CONFIG headers in: /usr/include
-- Found SUITESPARSE_CONFIG library: /usr/lib/libsuitesparseconfig.so
-- Found LIBRT library: /usr/lib/librt.so
-- Adding librt: /usr/lib/librt.so to SuiteSparse_config libraries (required on Linux & Unix [not OSX] if SuiteSparse is compiled with timing).
-- Found METIS library: /usr/lib/libmetis.so
-- Found SuiteSparse 5.4.0, building with SuiteSparse.
-- Building without CXSparse.
-- Building without Apple's Accelerate sparse support.
-- No preference for use of exported gflags CMake configuration set, and no hints for include/library directories provided. Defaulting to preferring an installed/exported gflags CMake configuration if available.
-- Found installed version of gflags: <path>
-- Detected gflags version: 2.2.2
-- Found Google Flags header in: <path>/include, in namespace: google
-- No preference for use of exported glog CMake configuration set, and no hints for include/library directories provided. Defaulting to preferring an installed/exported glog CMake configuration if available.
-- Found installed version of glog: <path>/glog
-- Detected glog version: 0.3.5
-- Found Google Log (glog). Assuming glog was built with gflags support as gflags was found. This will make gflags a public dependency of Ceres.
-- Using Ceres threading model: OPENMP
-- Building Ceres as a static library.
-- Enabling CERES_NO_CXSPARSE in Ceres config.h
-- Enabling CERES_NO_ACCELERATE_SPARSE in Ceres config.h
-- Enabling CERES_USE_OPENMP in Ceres config.h
-- Do not build any example.
-- Configuring done
-- Generating done
但是如果我将 Ceres-solver 构建为外部项目，它不会找到 BLAS（当我直接从源代码构建时会自动找到相同的 BLAS）。
ExternalProject_Add通过手动构建库与单独构建库时的行为是否存在差异？我的“主”项目（调用ExternalProject_Add者）中定义的选项和变量是否会泄漏到外部项目（如果是，我该如何防止这种情况发生？）。
我的电话ExternalProject_Add：
ExternalProject_Add(ceres-solver
PREFIX CeresSolver
BINARY_DIR ${CMAKE_CURRENT_BINARY_DIR}/ceres-solver
SOURCE_DIR ${CMAKE_CURRENT_SOURCE_DIR}/ceres-solver
INSTALL_DIR ${DEFAULT_INSTALL_ROOT}/ceres-solver
LOG_DIR ${DEFAULT_INSTALL_ROOT}/ceres-solver/log
LOG_CONFIGURE ON
CMAKE_ARGS
-DCERES_THREADING_MODEL=OPENMP
-DCMAKE_CXX_STANDARD=17
-DCMAKE_BUILD_TYPE=${CMAKE_BUILD_TYPE}
-DCMAKE_INSTALL_PREFIX=${DEFAULT_INSTALL_ROOT}/ceres-solver
-DBUILD_BENCHMARKS=OFF
-DBUILD_DOCUMENTATION=OFF
-DBUILD_EXAMPLES=OFF
-DBUILD_SHARED_LIBS=ON
-DBUILD_TESTING=OFF
-DCXSPARSE=OFF
-DEIGENSPARSE=OFF
-DEIGEN_INCLUDE_DIR=${EIGEN3_INCLUDE_DIR}
-DEigen3_DIR=${Eigen3_DIR}
-DSUITESPARSE=ON
-DLAPACK=ON)
以及 ceres-solver 的配置输出：
- Detected available Ceres threading models: [CXX11_THREADS, OPENMP, NO_THREADS]
-- Building with C++17
-- Found Eigen version 3.3.5: <path>
-- Disabling use of Eigen as a sparse linear algebra library.
This does not affect the covariance estimation algorithm
which can still use the EIGEN_SPARSE_QR algorithm.
-- Did not find LAPACK library, disabling LAPACK support.
-- Failed to find SuiteSparse - Did not find BLAS library (required for SuiteSparse).
-- Failed to find SuiteSparse - Did not find LAPACK library (required for SuiteSparse).
-- Failed to find SuiteSparse - Did not find AMD header (required SuiteSparse component).
-- Failed to find SuiteSparse - Did not find AMD library (required SuiteSparse component).
-- Failed to find SuiteSparse - Did not find CAMD header (required SuiteSparse component).
-- Failed to find SuiteSparse - Did not find CAMD library (required SuiteSparse component).
-- Failed to find SuiteSparse - Did not find COLAMD header (required SuiteSparse component).
-- Failed to find SuiteSparse - Did not find COLAMD library (required SuiteSparse component).
-- Failed to find SuiteSparse - Did not find CCOLAMD header (required SuiteSparse component).
-- Failed to find SuiteSparse - Did not find CCOLAMD library (required SuiteSparse component).
-- Failed to find SuiteSparse - Did not find CHOLMOD header (required SuiteSparse component).
-- Failed to find SuiteSparse - Did not find CHOLMOD library (required SuiteSparse component).
-- Failed to find SuiteSparse - Did not find SUITESPARSEQR header (required SuiteSparse component).
-- Failed to find SuiteSparse - Did not find SUITESPARSEQR library (required SuiteSparse component).
-- Failed to find SuiteSparse - Failed to find either: SuiteSparse_config header & library (should be present in all SuiteSparse >= v4 installs), or UFconfig header (should be present in all SuiteSparse < v4 installs).
-- Did not find all SuiteSparse dependencies, disabling SuiteSparse support.
-- Building without CXSparse.
-- Building without Apple's Accelerate sparse support.
===============================================================
Compiling without any sparse library: SuiteSparse, CXSparse
EigenSparse & Apple's Accelerate are all disabled or unavailable.
No sparse linear solvers (SPARSE_NORMAL_CHOLESKY & SPARSE_SCHUR)
will be available when Ceres is used.
===============================================================
-- Found Google Flags header in: <path>, in namespace: google
-- Found Google Log (glog). Assuming glog was built with gflags support as gflags was found. This will make gflags a public dependency of Ceres.
-- Using Ceres threading model: OPENMP
-- Building Ceres as a shared library.
-- Do not build any example.