From: MenTaLguY Date: 2007-09-29T02:36:03+09:00 Subject: Re: gc.c -- possible logic error? On Fri, 28 Sep 2007 21:57:22 +0900, Hugh Sasse wrote: > I still haven't figured out why GC.start isn't freeing the memory in > Tom's cases, though. The collector is conservative with respect to the C stack -- should there be any byte sequences on the stack that happen to look like valid references to a Ruby object, that object and all the objects reachable from it will be uncollectable. Have you checked to make sure that there are no apparent lingering references on the stack? (gc_mark_rest() is probably a good place to instrument) -mental