For reference, I did a './bootstrap.sh && ./configure && make msvc" on linux, copied over the igraph-0.6-msvc.zip and also the igraph header files to windows, loaded that up in VS 2008, applied the fixes above, and built igraph.lib.
This seemed to go well, but building the Python library, not so much. I had to monkeypatch graphobject.c because it is missing parameter 3 at line 3533:
as well as use the /FORCE option while building graphobject.obj because it could not resolve _igraph_assortativity_degree (and a few others).
I managed to install it in python to get the process down, but I know this hack isn't going to hold up well. Any suggestions to fix the igraph_get_shortest_paths_dijkstra call? I want to demo an analysis on Windows, but it needs some features of the 0.6 branch (like weighted calculations).