Bizarre clue: the model crashes inside [TroutmodelSwarm stock] at the end of a run, even though it should not be there. (It executes a print statement I left at the top of -stock just before dying.) Likely symptom of memory being overwritten? Could produce different behavior on a different machine.
But no-- I tried it with GCC in Linux and got exactly same behavior.
Bizarre clue: the model crashes inside [TroutmodelSwarm stock] at the end of a run, even though it should not be there. (It executes a print statement I left at the top of -stock just before dying.) Likely symptom of memory being overwritten? Could produce different behavior on a different machine. But no-- I tried it with GCC in Linux and got exactly same behavior.