tksharpless

The unresolved references are just the nine flags DECLARE_ed in Calibration.h and DEFINE_ed in Calibration.cpp. And all those definitions are flagged with a warning about inconsistent DLL linkage. It now...

Well wouldn't you know it. Nice that it starts up without crashing, but then.... PS C:\Users\tommy\build-FB360-dep\bin\Debug> ./calibration ERROR: flag 'v' was defined more than once (in files 'C:\Users\tommy\vcpkg\buildtrees\glog\src\v0.4.0-46ccbc49a4\src\vlog_is_on.cc' and 'C:\FB-6dof\facebook360\source\calibration\GeometricCalibration.cpp')....

The source of the pathless references to double-conversion.lib is a line in the fb360-dep cmakelists.txt which adds that as a literal name to the link list for Folly. This is...

Finally have a working calibrate.exe -- testing it now on your sample data. The last piece of the gflags puzzle is solved by putting the definitions from calibrate.cpp into one...

The calibration test fails on my 24GB laptop when ceres solver tries to allocate 23GB of RAM. I am looking into ways to make the memory requirement more reasonable. Hopefully...

Rebuilt ceres with SuiteSparse. Now geometrical calibration demands less that 1GB of RAM. But it eventually fails after issuing the warning "final pass median error too high". Log attached. The...

Actually the process did not crash. It saves an updated rig file and exits normally. So I guess it can be said to be working. The numbers in this file...

I can now build all of facebook360-dep except test. The only real source code issues are dealing with the wchar strings that boost-win32 uses for path names, which is simple,...

The correct fix for wide boost::filesystem strings is the official boost solution, which is to imbue filesystem with the Windows UTF-8 codepage. This will require that all code which needs...

The situation with ISPC is a bit tricky. Your ISPC.cmake script creates projects for building the ispc libraries on Linux and OS X; but the process is different on Windows....