linear_regression.cpp: undefined reference to `wrapper2_XXXX' about mlpack HOT 4 CLOSED

Just as an aside, life gets a little easier for linking if you move to mlpack 4.x, since mlpack is header-only and no longer has a boost dependency (so you could remove the -lboost_serialization -lmlpack).

The Armadillo configuration output looks correct, so I am wondering how you installed Armadillo. Did the Armadillo headers get installed into /usr/local/include/, and the Armadillo libraries get installed into /usr/local/lib/? As a guess, I wonder if adding -I/usr/local/include/ to the compilation command will help. Make sure there's only one version of Armadillo present on the system, and that the version being used by the compiler when including and when linking are the same.

from mlpack.

Thanks for your advice. I chose the old version because I want to use mlpack in a platform based on C++11.

I checked armardillo's location and found that it was installed in usr/ not in usr/local/, so I uninstall the whole cygwin environment and reinstall mlpack and its related dependencies in usr/local/, however, same issue occurred. Also, -I/usr/local/include/ doesn't work.

from mlpack.

And attached images are installed files in usr/local/include/ usr/local/lib/ and usr/local/bin/

from mlpack.

Old releases like mlpack 3.4.2 are no longer supported. We don't really have the bandwidth for that. If you wish to continue using mlpack 3.4.2, you're on your own.

Furthermore, this issue does not appear directly tied to mlpack. The core problem seems to be an installation problem you have with Armadillo on Cygwin, a Linux-like emulation layer on Windows. Cygwin has always been problematic, as at best it's a rough approximation of a proper Linux environment.

Suggest the following:

• manually remove all currently installed versions of Armadillo; this includes the runtime library and the header files; look in places like /usr/lib, /usr/lib64, /usr/local/lib, /usr/local/lib64, /usr/include, /usr/local/include, ...
• download the latest version of Armadillo, and reinstall it
• make sure the armadillo example program runs on your setup
• upgrade your compiler to allow use of C++14, preferably GCC 11 or 12; the C++14 standard is supported by GCC since GCC 5.x (released in 2015, about 8 years ago)
• move to mlpack 4.1

If you can't get things working on Cygwin, either use WSL, or a proper Linux installation like Ubuntu, Fedora, AlmaLinux, RockyLinux, Debian, etc.

from mlpack.