This sounds like a bug. Explicit calls to `collect-garbage' shouldn't be needed for the garage collector to do its job.
Does anyone know an easy way for me to make the problem happen, so I can see what is going on? At Thu, 14 Jul 2011 12:47:54 -0400, Greg Hendershott wrote: > Once upon a time I noticed the Racket web server has a thread that > calls collect-garbage periodically (e.g. every 5 minutes). > > I've discovered I need to do the same, for any program that needs to > run for hours or days. Otherwise, the program eventually abends: Out > of memory. Sometimes with a "Racket virtual machine error" message. > This is my experience on both Windows 7 and on Linux, with Racket 5.0 > and 5.1.1. > > As a result the following has become a "magic spell" I now throw into > such programs: > > ;; Make a thread to do garbage collection every 5 minutes. > (thread > (lambda () > (let loop () > (sleep (* 5 60)) > (collect-garbage) > (loop)))) > > It seems weird to have such a magic spell, and to know it only from > nosing around the web server source. > > Instead, maybe the Racket runtime should do this automatically (at > least by default)? I think that's people would expect coming from some > other language environments with GC (at least it's what I expected). > > Or, at least it would be documented that this is by-design (say for > performance reasons, or there's no reasonable default, or whatever), > and that people should use such a magic spell for long-running > systems? > _________________________________________________ > For list-related administrative tasks: > http://lists.racket-lang.org/listinfo/users _________________________________________________ For list-related administrative tasks: http://lists.racket-lang.org/listinfo/users
