I haven’t seen this internal compiler error before. Note that this is a bug in GCC, not a bug in Rosetta.
One option is to try a different compiler. A different version of GCC, or a different compiler entirely, is not likely to have the same problem.
A second option is to fiddle with the inlining limits. Many years ago, we set the compiler options for very greedy inlining, because at the time it improved performance. See this thread on how to fiddle with the inline limits: http://www.rosettacommons.org/node/2274
A third option is to try to disable or modify the code that fails to compile. I can’t reliably do this, since it doesn’t fail for me, but I can give you advice on attempting to excise just the failing code from ShapeComplementarityCalculator. (Note that this will mean you are missing some functionality; I’m guessing you only need that unit for docking based on its name, but I don’t know for sure.)