Compile on Windows

[Andyjk1406]

I saw you have recently updated this, but I still have an issue in cmake with the gflw external:

---------------GLFW-------------------
Looking for pthread.h
Looking for pthread.h - not found
Found Threads: TRUE
Could NOT find Vulkan (missing: VULKAN_LIBRARY VULKAN_INCLUDE_DIR)
Using Win32 for window creation
CMake Error at ext/glfw/CMakeLists.txt:403 (install):
install INSTALL(EXPORT) given keyword "EXPORT_LINK_INTERFACE_LIBRARIES",
but target "glfw" does not have policy CMP0022 set to NEW.

This may be an issue with which version gets downloaded into the /ext folder ?

Can you help ?

BTW - I used vcpkg to build ceres (very easy), and then just make sure you have the -DTOOLCHAIN flag set in cmake to point to vcpk when building the solution. This makes the Ceres issue on Windows a pice of cake!

[jingxa]

Compiling always is in trouble! i can't help u ,sorry, Trying to again !

[adrelino]

Well since I am using submodules, the version in ext/glfw should be exactly the same that I have.

The only thing I can think of is an incompatibility with that (old) glfw version with a newer cmake version, hence the warning with setting a new policy because some variable was deprecated.

What cmake version are you using?

[adrelino]

Does this help:

http://www.glfw.org/docs/latest/compile_guide.html

Otherwise, instead of compiling glfw on windows using cmake, you could try to just use their precompiled .dlls downloadable here: http://www.glfw.org/download.html. They have dlls for multiple versions of the MSVC as well as the MinGW compiler, which one are you using?

[PiotrGizaB3D]

Hi everybody,
I've just compiled this project on Windows 10 (Visual Studio 2017). I've encountered several problems, mainly with external libraries compatibility:

1. newest release of G2O didn't work - finally used this one: https://github.com/RainerKuemmerle/g2o/releases/tag/20170730_git
2. new Sophus has slightly different syntax: Sophus::SE3Group has been renamed to Sophus::SE3,
3. it seems that Sophus doesn't provide internalJacobian method in Eigen::Map<const Sophus::SE3d> class, I've tried to replace this with Dx_this_mul_exp_x_at_0, but I have no idea if this is correct [file sophus_se3.h, function ComputeJacobian],
4. in file sophus_se3.h there was a problem with function getParametrization with this expression:
   return new ceres::AutoDiffLocalParameterization<SophusSE3Plus,Sophus::SE3::num_parameters, Sophus::SE3::DoF>;
   changed it to:
   return new ceres::AutoDiffLocalParameterization<SophusSE3Plus,Sophus::SE3d::num_parameters, Sophus::SE3d::DoF>;
   since num_parameters and DoF are hardcoded static constexpr, there should be no execution difference at all,
5. in file icp_ceres.cpp function Sophus::SE3d isoToSophus(const Isometry3d& pose) was erroneous. Changed this line:
   return Sophus::SE3d(pose); to return Sophus::SE3d(pose.rotation(), pose.translation());
   since Sophus::SE3d has no constructor that accepts Isometry3d object reference as input. No idea if this is correct semantically.

[georgezzj]

Hello, I don't quite understand how you solved the third question, I am very difficult to deal with this problem now