[Xastir-dev] Long Term CPU

[Tom Russo]

On Sat, Feb 19, 2005 at 11:49:30AM -0500, we recorded a bogon-computron collision of the <jerryc at netlab.org> flavor, containing:
> there is a small leak somewhere, after about 30 days up time it used all the
> memory and my swap went into overload.

Hmmm.  I've got an instance of xastir running here for 15 days and it's 
riding at about 25M resident size.  That's about what it is when I start it.
I'm not observing the leak, so there must be a very different set of data
you're getting.

> restart of xastir cleared it, It used all availible memory making my pine and
> other programs go to swap. It had sat for a week running as the displayed
> program.

Is this a recent CVS or a snapshot?  When you say it "sat" for a week, do you
mean that literally --- a fixed map view with no zoom/pan operations being
done?  (asking to see if it could be shapefile/rtree related leaks) 

What configure options?  What types of maps?  What sort of interfaces?
Are you displaying nation-wide wx alerts from the net, or just local RF?
(this would be a significant difference from what I do --- I hardly ever
see WX alerts since I only run RF and nobody 'round here gates wx alerts
to RF).

> Actually the operation was normal, do the hash tables stay in memory or swap
> out. I had not noticed this earlier, but then I never was able to keep it up
> long enough to notice without some other issue making a bigger issue.

> It grew over time. Are old lerts being purged and memory released?

I think Curt made some significant changes to how WX alerts are being handled
in recent weeks --- maybe more recently than your 30-day-uptime version.
Does this behavior persist in recent CVS versions?  

Could be an issue with the hash-table functions.  When I was implementing
the rtree stuff I noticed that the hash table library we use requires you to be
very careful when deleteing entries.  I had some leaks in the rtree-related 
hash tables early on because of how the hash functions do (and don't do) their 
cleanup.  I know that alerts are now in hash tables, so it could be there.

> I only have 128K real memory. Redhat 7.1 

Heh.  128K Linux box.  Takes ya back, it does.  Typo, I hope.

-- 
Tom Russo    KM5VY     SAR502  DM64ux         http://www.swcp.com/~russo/
Tijeras, NM  QRPL#1592 K2#398  SOC#236 AHTB#1 
 "When life gives you lemons, find someone with a paper cut."