In the test case, I am using arma::arma_rng::set_seed(0) so that the second channel can specifically be dropped out (as in the image above). That allows me to directly compare the output with an expected matrix. This works when I test it on Mac and Linux. However, I noticed that the Windows build failed for the last commit when checking this. I think arma::arma_rng::set_seed(0) is not working the way I expect it to on the Windows machine. Any clues as to how I can go about resolving this? Thanks!

In the test case, I am using arma::arma_rng::set_seed(0) so that the second channel can specifically be dropped out (as in the image above). That allows me to directly compare the output with an expected matrix. This works when I test it on Mac and Linux. However, I noticed that the Windows build failed for the last commit when checking this. I think arma::arma_rng::set_seed(0) is not working the way I expect it to on the Windows machine. Any clues as to how I can go about resolving this? Thanks!

It's not guaranteed that the random number is the same across all systems, the only workaround I can see is to check all possible solutions, if that makes sense?

This looks alright. I'll merge it if you don't have anything else to add (take care of the conflicts as well).