Thanks! It looks like we've finaly got this solved. I hope porting RakGS to MinGW will be painless.

By "false positive" I meant that my other test system (without VC++/vcredist) did load my plugin under some circumstances while it should not have. The problem seemed to have been solved, while in fact the plugin still did not load on some systems.

Could you (just to be entirely sure this isn't because the vcredist wasn't uninstalled properly) verify that: running RakGS without installing vcredist fails?

That's hopefully the last thing I will need your help with, regarding this problem. Thanks!