
mark at geekhive
May 2, 2011, 11:25 AM
Post #9 of 16
(2360 views)
Permalink
|
oops... Anyhow, to continue: http://search.cpan.org/dist/Devel-LeakTrace-Fast/lib/Devel/LeakTrace/Fast.pm If I had to guess I'd say bric_queued would be an easier debugging environment than mod_perl, and there's a pretty good chance that if we can fix the leak under the one it will cover the other as well, since all bric_queued does is call a burner. I wonder if it's worth giving up a few nights or a weekend for this. -mark On Mon, May 2, 2011 at 20:23, Mark Jaroski <mark [at] geekhive> wrote: > Hi Alex, > > That's a very good point. Maybe we should consider hunting the leak down > with > > > On Mon, May 2, 2011 at 17:47, Alex Krohn <alex [at] gt> wrote: > >> Hi Mark, >> >> > A memory leak is when a program requests memory from the system, and >> then >> > fails to give it back when it's done. As I understand there is some >> > controversy over whether the behaviour of Perl is considered a memory >> leak >> > or not, since apparently it fails to give the memory back by design. >> > >> > At least that's my recollection. >> > >> > In practice what it means is that you can't have a long-running perl >> > daemon that does anything very interesting. You have to kill it off from >> > time to time and restart it. >> >> Perl doesn't release the memory allocated, but does reuse it. So you can >> have Perl daemon's running for a long period of time and doing >> interesting things, just can't have them keep growing in memory (much >> like any language). =) >> >> Cheers, >> >> Alex >> >> -- >> Alex Krohn <alex [at] gt> >> >> >
|