Does anyone know a way to speed up adjacency_list deallocation for debug builds?  I've done some timing tests (see code below - should build as-is in VC9) and found the adjacency_list creates quite quickly, but is terrible during deallocation (actually, it can be as bad on creation of the graph, too, if you are building up the graph incrementally using a vecS for your vertex storage that requires reallocation as the size grows).  In a release build, everything is blazing fast, but it's so slow when linked against the debug CRT that I essentially cannot debug my app.  And the graph doesn't have to be too big (see code below, 50K results in a deallocation time of 171 sec if I use vecS and 268 sec if I use listS, and that's on a quite good workstation; my production graphs are > 100K vertices).  BTW, I'm building with Visual C++ 9 (2008).