Le Vendredi, 12 Mars 2010 17:38:37 -0800, Wes Hardaker <harda...@users.sourceforge.net> a écrit :
> You're right it's not. Generally there are a number of places in the > code (many many actually) where single-instance calls to memory > allocation are taken and never cleaned up like they should be at > shut-down time because we assume that the nice memory-protected > operating system will do cleanup at application shutdown. This is one > of the middle-casse where yes, a (small) string is duplicated and > never released but probably should be. > > If you want to submit a patch to our patch tracker to fix it, we'd be > happy to consider it. But it's probably not a serious problem for > most people. I've spotted one in 5.4.1 recently, a strdup() inside a printf(), that came from a Debian patch. Since the app (snmpd) never shuts down, memory cosumption was simply always going up, and pretty fast at that when under tests consisting of runnign snmpwalks against it. At first I had a daemon restarting snmpd after when it took more than 10 MB of RSS, but then I ended up searching for the leak and fixing it. This problem is not there in 5.5. Still, there's another leak in 5.4.1 whenever a WAN interface is brought down/up. Some users have seen a 1MB memory consumption over a week. Since these units runs 365 days a year, the lil' daemon watching over snmpd's memory is still there in function. One of these days I'll get down to find what that leak is exactly. ------------------------------------------------------------------------------ Download Intel® Parallel Studio Eval Try the new software tools for yourself. Speed compiling, find bugs proactively, and fine-tune applications for parallel performance. See why Intel Parallel Studio got high marks during beta. http://p.sf.net/sfu/intel-sw-dev _______________________________________________ Net-snmp-coders mailing list Net-snmp-coders@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/net-snmp-coders
