Re: [Pan-users] 0.13.4 memory leaks

[Duncan]

On Sun 09 Mar 2003 15:41, Steven Ellis posted as excerpted below:
> 
> --- Brian Morrison <address@hidden> wrote:
> >
> > On Sun, 9 Mar 2003 07:00:48 -0500 in
> > address@hidden "Chris
> > Charlebois" <address@hidden> wrote:
> > > 
> > > I've noticed the problem too. To repeat the problem, try downloading a
> > > bunch of large binaries until the systems memory usage is 100% and the
> > > swap file is starting to be utilized. Then continue to download large
> > > binaries until the swap file is filled, once it is, boom!!.
> >
> > That sounds like the usual VM OOM problem rather
> > than Pan as such.
>
> What VM 00M problem? I use RH 7.2 with a custom 2.4.20
> kernel and I have the same problem with Pan 0.13.4 and
> 0.13.90. Eventually Pan eats all the memory on the box
> after performing a large number of big binary
> downloads.

Well, it is the VM dealing with it, sure, but that doesn't mean it's the VM 
causing the problem.

What sort of memory and swap U run, Chris & Steve?  Here, I haven't noticed 
the issue.. Base system 2.4.20 vanilla Marcelo kernel, Mandrake 9.1 Cooker, 
updated twice weekly or so, KDE 3.1 (Mdk cooker version), Athlon C 1.2GHz, 
512MB PC2100 memory (ECC), 1GB swap partition, ReiserFS on all partitions but 
swap and legacy FAT32.  I run a ksysguard kicker applet monitoring all sorts 
of stuff, and user memory and buffers usually take about half my real memory 
(256MB), with the rest cache.  Swap generally runs less than 200M (running 
130M now), and I seldom thrash unless I DO have a runaway program (like KView 
during the KDE betas).  I've not seen that change w/ the new PAN.

I do know PAN takes forever (well, about a minute, I timed it) to load, and 
has pretty much since I started running it except during those betas when 
they had the fast start code that ended up creating race conditions.  I also 
know that if I've just done something that cleared the memory fs cache, then 
left a bunch of free memory, by the time PAN loads, I have filled memory with 
file cache again, after it's read and parsed its message cache.  Actual user 
memory use doesn't go up so much, but it does read in all those files and 
fill up memory with file cache.

-- 
Duncan
"They that can give up essential liberty to obtain a little
temporary safety, deserve neither liberty nor safety." --
Benjamin Franklin