New job

I start a new job at Presilient on Monday. They work with
cool technology like Spring, Maven and Git so I'm hoping to
have stuff to blog about again soon.

System powers off during certain games

A couple of weeks ago I was ready for a new game and installed
Grand Theft Auto IV on the Windows XP side of my dual boot PC.
As I tried to play the game a very troubling symptom began to
occur... the display would freeze up and a few seconds later the
whole system would power off.

I tried checking the CPU temperature in the BIOS afterwards
but it was fine. However, this was after the system crashed and
rebooted so presumably it had some time to cool off.

I tried some other games and they worked fine, however, one
other game, STALKERS: Shadow of Chernobyl, was
getting the same failure mode, maybe a little less often.

Nothing in Linux was having any trouble including hi-res 3D games
like Urban Terror and DOOM 3.

I tried vacuuming out the case and all the fans, including the
fan on the graphics card. There really wasn't that much dust
accumulated and afterwards the system still crashed.

My gut is telling me the CPU is overheating so I ordered
one of these to try replacing the plain-vanilla CPU fan that
came with the CPU.

A colleague suggests my power supply might be too weak for all
the devices I'm trying to run off of it... something else to check.

The souped-up CPU fan arrived and I went to install it. There
was an unfortunate turn of events. I applied the thermal grease
to the fan as instructed and lowered it into place on top of the
CPU. I was in a hurry and careless and manhandled it into place.
Power on... nothing. I removed the fan again and found the
CPU stuck to the bottom of the fan, all cockeyed and with
numerous bent pins. What must have happened is the CPU
stuck to the thermal grease on the fan, got lifted out of its socket,
then was damaged as I tried to position the fan and lock it down.

I made a hasty visit to the computer hobby store. All they had
on hand in AM3 was an AMD 840. This is the low end of the
Phenom II 4X product line. It's a quad core but 3.2 GHz
instead of the 3.4 GHz of the AMD 965 I was replacing,
and it lacks a L3 cache.

The second, more cautious attempt at installing the fan was
successful. The temperature in the BIOS display is around 30 C
where before it was 60-70 C. But it's not really a good comparison
as it's a different CPU.

The problem I set out to solve appears to be fixed. I played both
games for a while with no crashes. I kind of think the cooling
theory was right, but something else about the old CPU could
have been causing the problem.

Update: A month or so went by and I was really noticing a
difference with the sluggish 840 CPU so I went out to Micro Center
and got an AMD Phenom II 975. It's basically the same as the
old 965 but 3.6 GHz instead of the 965's 3.4 GHz. I got this
installed without incident. GTA IV still runs without crashing
the system. This is more evidence that CPU overheating was
the problem. If anything this faster CPU should run hotter.

Amusing little irony... The AMD processor in a box comes with
the CPU, fan, thermal grease, instructions and a sticker.
The instructions advise that you will void your warranty if
you use your CPU with a fan other than the one provided,
yet it seems the stock fan is not really adequate for some
applications.

AMD ganged and unganged memory access

Last year I did a substantial upgrade on my desktop hardware.
Around that time I started noticing an issue with Ubuntu 9.10.
Maybe once a week or every other week the system would lock
up. Whatever was on the monitor would be frozen, keyboard
and mouse unresponsive, computer off the network, only
rebooting would fix it. If I was doing intensive DVD ripping
with dvd::rip it would happen a lot more often, like once a day.
I never observed this in Windows but I rarely have Windows
up for more than a few hours to play games. Upgrading to
Ubuntu 10.04 didn't fix the problem.

Since I've got some time and I'm doing intensive DVD ripping,
I thought I try to troubleshoot this.

(1) The BIOS (GA-MA790FXT-UD5P motherboard) has a 'load
fail-safe defaults' option to load conservative settings, tried that,
system still locks up.

(2) I looked at updating the BIOS. There's one version more
recent than what I've got on this one year old MB. It's a beta
and when I try to install it the flash application says it's got
a checksum error. Scratch that...

(3) Just cruising around in the BIOS menu I found a setting
for memory control mode: ganged or unganged.
Here's some discussion of what this means. Just kind of on
a whim I went ahead and changed this setting to ganged.

So after about 3 days of heavy DVD ripping the computer has
not locked up. Crossing my fingers. I won't really feel sure the
problem is fixed for a couple of weeks since it's so intermittent.

Ubuntu 10.04

It's the holiday season and that means OS upgrades.
I put the Ubuntu 10.10 CD in my home desktop and
tried to boot. It didn't even come up. I wonder if this
is the problem ESR mentioned. I put 10.10 on my two
grandchildren's laptops and didn't have any trouble.

I retreat back to 10.04. OK, boots fine. I'm a little
disappointed it doesn't detect my RAID but I know
how to do that manually.

It detects that I need a proprietary graphics driver
and downloads that... it figures out my USB wireless
device... a simple

mdadm -A /dev/md0

brings the RAID back up.

Wireless networking is unusably slow... there are
brief spikes of activity separated by long stretches with
no traffic flowing. I spent a couple of hours on this googling
and searching the forums. This magical incantation (I
wound up putting it in /etc/rc.local) fixed the problem:

sudo iwconfig wlan0 power off

See this thread.

Problem with hibernate using the 'Shut Down' applet
from the Gnome panel. Hibernation appears to work,
but resume from hibernate doesn't... system comes up
as from a cold reboot, but with networking disabled.
Problem is discussed at great length in bug 499940
and bug 577916, the latter of which appears to be still
open. None of the fixes described other than the
following workaround worked for me.

Workaround: I tried installing the 'hibernate' package
and running /usr/sbin/hibernate from the command line
(as root)... that seems to be working.

Slight bug in evince, see bug 572434. Fixed with

touch $HOME/.gnome2/evince/last_settings

I didn't need to do anything special to get Java and Flash
plugins installed into Mozilla, even running 64 bit. Nice.

It took me about a week to notice this but it seems the
distribution no longer includes the Lightning (calendar)
plugin for Mozilla Thunderbird. Just installing the
plugin from Tools->Add-ons didn't work. There was
some kind of incompatibility with 64-bit Thunderbird.
I was able to get it working by following this thread.

Update: my hibernate 'workaround' isn't really working.
This is keeping me from upgrading on my work box
where I use hibernate constantly switching back and
forth from Linux to Windows and I need this to be
fast and reliable. The last couple of times I tried
hibernate, when I went to resume I just wound up back
at the login screen.

Alternative to the "200 lines kernel patch that does wonders" that you can use right away

I configured this on my laptop a couple of days ago. I can't really
notice any difference so far, but on the other hand it didn't seem
to break anything. Getting ready to try it on my work computer
which is always struggling under the heavy load (VMware,
Eclipse, Weblogic server, SQL Developer, Firefox...)

http://www.webupd8.org/2010/11/alternative-to-200-lines-kernel-patch.html

No Fluff, Just Stuff, day 3

Best presentation of day 3 again goes to Venkat for 'Test Driving
Multithreaded Code'. Concurrent stuff is oddly fascinating.
Venkat: "When I read Goetz' Java Concurrency in Practice, I
realized all my code was broken"

I scored a cool NFJS LEATHER JACKET for attending 5 times.

No Fluff, Just Stuff, day 2

Cool presentations for day 2 were 'Gaelyk: Lightweight
Groovy on the Google App Engine' by Tim Berglund and
'How to Approach Refactoring' by Venkat Subramanian.

Links:

Try to knockout before you consider mock out
There's an archive of past presentations and code
from Venkat at Agile Developer.
DbUnit
MockRunner
Unitils
Cucumber