On Mon, 23 Sep 2013, potato_research wrote:
Jeremiah, I'm working on your suggestion of dumping the data and obtaining an MWE that we could make into a test (if it turns out there's a problem in the boost code). In the meanwhile, one of the assertions you added triggered! Here it is:
Assertion failed: (p_cur_label->b_is_valid), function r_c_shortest_paths_dispatch, file /Users/alberto/Applications/boost/boost/graph/r_c_shortest_paths.hpp, line 472.
It's good that the assertion triggered; it probably makes the issue easier to debug. Here are a couple of questions that might suggest places to look: 1. Do the cases that fail have multiple, partially-overlapping Pareto-optimal paths to the goal? Do they have multiple paths that are tied for "goodness"? 2. If you remove all of the clearing of b_is_valid, plus remove all of the object destruction and deallocation calls (to avoid the program crashing), does the answer seem reasonable? -- Jeremiah Willcock