The current build system is a mess with a mix of cmake, premake and several nvidia build tools.
cmake wraps nvidia build tools which in turn wraps premake (along side other tools depending on the host/target platforms). Some limitations of the current system.
nvidia tools lack documentations
hardcoded build directory
not exactly hardcoded, but changing them require understanding nvidia's build tools
does not support installing
The main limitation we need to solve in order to release this is to make it support installation and packaging for omniverse.
Should we write custom scripts to perform installation or attempt to build this using cmake to make use of cmake install targets?
If we go the cmake route, how should we handle fetching of dependencies?
How should we package this so that it can find the ignition libraries? Note that we also need to set this up so that they do not use the system's ignition libs, as this requires custom build of ignition with pre cxx11 abi.
Static link those libraries
set rpath to find those libraries
make a launcher script which sets LD_LIBRARY_PATH (this is the approach the nvidia sample connector uses)
The current build system is a mess with a mix of cmake, premake and several nvidia build tools.
cmake wraps nvidia build tools which in turn wraps premake (along side other tools depending on the host/target platforms). Some limitations of the current system.
The main limitation we need to solve in order to release this is to make it support installation and packaging for omniverse.
LD_LIBRARY_PATH(this is the approach the nvidia sample connector uses)