"Dmitry V. Pleganoff" <dimple at mail.primorye.ru> wrote: > By the way, rlm_checkval meets in most cases losses Please re-run valgrind after re-building, and post any leaks. They should be fixed, too. I've run valgrind on the CVS head, but I don't have every possible code path tested, so I can't find all of the leaks. Alan DeKok.