Fedora 11 on Thinkpad R61 (2009-09-24)
I couldn't really find too much on this so I'm posting it.
When installing Fedora 11 on a Thinkpad R61 that has the intel graphics chip set,
the initial boot screen will crash/hang at a blank grey screen. The reality is
it's just display all wrong. Just press "tab" then type "linux text"
and it'll proceed to install properly but in text mode.
Building your own TV antenna (2009-06-09)
Okay this is extensive enough to warrant it's own page.
Click on the picture to see my adventures in building my own antenna.
Comparing Thinkpad X61s L7500 with R61 T7100 (2009-05-02)
I recently bought and X61s to replace my R61 because the R61
is too large and bulky. The R61 has a 15" lcd where as the X61s
is 12" and the R61 is about 3 times as thick and 3 times as heavy too.
The thing I was wondering before I bought this was how the performance
of the machines would be affected - specifically for the things I do
which is work / software development. The machines are nearly identical,
they were so close that I could swap HD's with both windows and linux
and upon boot up neither os's noticed any components had changed at all.
Everything from the video card, to sound card, to wireless adapter, ram,
to hard drive, were all identical, the only thing different
were the processors, the R61 has a T7100 (core 2 duo, 1.8ghz, 2mb cache, 800mhz bus),
the X61s has a L7500 (core 2 duo, 1.6ghz, 4mb cache, 800mhz bus).
For the majority of people who use computers (not gamers), both machines
are over powered for what they would do and they wouldn't notice any
difference at all. For me, I do software development so I do things like
running database servers, application servers, and compiling source code.
The main question for me was... how is this going to impact my compilation
times as I compile fairly often, sometimes up to once every 5 minutes.
To make sure this test was fair, I removed and installed both the memory
and HD from one machine to the other with out reconfiguration. This
pretty much guaranteed both installations were the same as it's much harder
to install 2 systems the same unless I wipe both machines and start from scratch.
Other than the CPU's mentioned above, the other relevant configuration specs are :
- 3gigs of 667mhz ram
- HITACHI HTS541612J9SA00, 5400rpm, sata, 8mb buffer
- Fedora 10 32bit is the OS
- Sun Java 1.6 / maven 2.0.9 is the compiler
My test was to compile the Apache OpenJpa code base. I decided to
run the compiler with the following command "time mvn -Dmaven.test.skip=true clean package"
I ran the compiler sequentially 4 times on both machines, and I discarded the first run on both machines.
The average of the 2nd, 3rd, and 4th real times as provided with the time
command where :
- R61 T7100 @ 1.8ghz : 1minute, 56 seconds
- X61s L7500 @ 1.6ghz : 1minute, 43 seconds
This actually surprised me. I was expecting the 1.6 to be slightly slower, not
just because of the ghz but also because it's a ultra low voltage cpu and
sometimes they cut things out to make it smaller and lower power consumption. I
expected the 4mb cache to help make it close to the speed of the 1.8 processor
but I didn't expect it to exceed it, consistently on every run, by roughly 11%.
Like I said, most people won't know or care which is faster and the machine
style (15" v.s. 12") would be the deciding factor in choosing, but I couldn't
find any numbers so here's some posted so you can take it into consideration
if you need to.
Wipe MBR using linux (2009-04-26)
I always seem to have a need to nuke the MBR on disks...
don't ask why, I just do. For those of us who use linux
exclusively, we can't really use "fdisk /mbr" anymore.
Here's a linux version of nuking the mbr.
- dd if=/dev/zero of=/dev/sdb bs=512 count=1
Where sdb is the drive you want to nuke. Seems to always
be a usb drive for me, that's why it's my example.
Installing Fedora from USB drive (2008-08-01)
Finally, installing fedora from a usb drive is relatively easy. It's
previously been possible but enough of a pain that it wasn't feasible
and the cost of large usb flash drives was a problem. Now (approximately
2008-09-05) the cost of a 8gig usb drive is $10 or less so I finally
bought one and tried it again. Here's a simple how-to based on Fedora 9.
Prerequisites
- usb drive 4 gigs or more (I'm talking flash drives, I don't know if HD's will work or not)
- both the network install iso and the dvd iso for fedora, i.e. Fedora-9-i386-netinst.iso and Fedora-9-i386-DVD.iso
- assuming you're running linux, you need livecd-tools installed, i.e. yum install livecd-tools
Steps
- plug in your usb drive, if it mounts, unmount it (yes it's important)
- /sbin/fdisk /dev/sdb (replace sdb with your location), delete all partitions, create 1 big primary partition, mark it bootable, save changes and exit
- if your system mounted the new partition, unmount it
- /sbin/mkfs.ext3 /dev/sdb1 (replace sdb1 with your partition location, in theory this isn't required but it seems to be sometimes)
- if your system mounted the new partition, unmount it
- livecd-iso-to-disk Fedora-9-i386-netinst.iso /dev/sdb1 (this will make the usb bootable)
- now mount the usb partition
- cp Fedora-9-i386-DVD.iso /media/disk (replace /media/disk with where ever your usb is mounted)
- make sure your bio has boot to usb or boot to usb-hdd enabled
Now when you boot, select "install off HD" and just select "/dev/sdb1" and well... "works on my machine" :)
When I tried to make a Fedora 10 install usb I had to do one extra step.
For some reason the usb drive was missing files after the above commands.
You need to copy the files from the netinst iso to the usb drive manually.
i.e.
- mkdir /media/loop (to create a loop mount point if you don't already have one)
- mount -o loop Fedora-10-i386-netinst.iso /media/loop
- cp -r /media/loop/* /media/disk (where /media/disk is where your usb drive is mounted)
If you don't do that it complains about missing images/install.img or something like that.
Mutt
move deleted messages to a folder called deleted (2008-08-23)
in your .muttrc add
- folder-hook . 'macro index d <copy-message>=deleted\ny<delete-message> "Safe Delete"'
- folder-hook . 'macro pager d <copy-message>=deleted\ny<delete-message> "Safe Delete"'
- folder-hook 'deleted' 'macro index d <delete-message> "Delete Message"'
- folder-hook 'deleted' 'macro pager d <delete-message> "Delete Message"'
deleting old messages (2008-08-23)
To delete mail more than 30 days old... (i.e. to clean your deleted folder)
- T
- ~d >30d
- ;d
Linux monitor colour calibration (2008-11-01)
Normally I don't pay much attention to minor nuances of my monitor colour
but I am running a laptop with an external display on spanning mode
and the colour differences are slightly annoying and recently I've been
doing more with photos and tweaking the colour levels on the photos
and it became more noticeable that the photos didn't look the same
on the different display - and I didn't know which one was right.
On linux you just use the Argyll (or sometimes referred to as Argylcms)
application.
After a brief look into the subject, I bought a Spyder 2 Express.
Apparently, ColorVision is not an open source software friendly company
and I shouldn't have bought their product... I didn't know at the time.
Apparently Pantone Huey is OSS friendly and works well with Argyll. To
get the Spyder2 working (it looks like the only difference between the
spyder 2 express / suite / pro is the software and not the sensor) you
actually need the install cd / software. Argyll pulls a dll off there
because ColorVision won't tell them about something and won't let them
release the dll with the code. At any rate it's still fairly simple,
after downloading and installing / extracting Argyll :
- run 'spyd2en' (to patch Argyll with the spyder dll)
- run 'dispcal -h' (to show you the options and the monitors you have)
- run 'dispcal -v -d 1 -q h -y l ./external' (where -d 1 was my external monitor, -q h for high quality, -y l for lcd, ./external was my output filename, I tried different quality levels but the results were erratic)
- on the menu I just selected 'continue to calibration'. Don't forget to turn off your screen saver or it might turn on and screw it up.
- when it finishes there'll be a file, in my case called ./external.cal
- run 'dispwin -d 1 external.cal' (parameters are similar to dispcal)
Only thing left to do is do it all over again for the laptop display,
then find a good place to put the command. I've put it in
/etc/X11/xinit/xinitrc.d/localusr.sh so it runs when I login. In reality
I wrote a script that runs it once a minute then started the script there.
This is so when I unlock my screen from a screen saver it runs again soon-ish.
After the above, the two displays looked a lot closer, especially on photos
but I still noticed that my LCD display and my external display aren't
the same colours. Maybe I'll find a way to sort that out
someday or maybe that's what I get with cheap displays.
After all the above, I ended up picking an original Spyder off ebay for like $12
because it didn't come with the software. Considering I don't use the software
anyways it seemed like a good deal and it works just fine with Argyll too.
I'll just pawn the Spyder2 off on the used market.
Linux with HD's load cycle / continuous clicking (2007-06-06)
This seems to be a recurring issue that I have to deal with each time
I reinstall linux on my laptops. The HD goes into aggressive
standby / sleep mode every 3 seconds symptomised by a clicking
sound every 3 seconds. The real problem is that it parks the HD
head every 3 seconds. Not only is it a performance hit but
apparently most HD's are only rated for 200,000 load cycles
or something like that.
To verify you have this problem run 'smartctl --all /dev/sda'
(where sda is your drive device). Look for (or grep for)
'Load_Cycle_Count'. Under normal running conditions
this number should not go up. It might go up every time
you turn off your computer or put it into suspend or sleep etc.
so you'd expect maybe a few every day. If this number is going
up every few minutes (or seconds), you've got this glitch.
To fix this glitch, turn off HD hibernation. i.e. '/sbin/hdparm -B 254 /dev/sda'
for me. To verify the fix, look at the load cycle count again
and it should no longer be incrementing like a freaking clock.
Now the only thing you need to do is make sure that command
is run every time the machine boots up, i.e. a script
in /etc/rc.d/init.d (or be lazy and just put the command
into an existing script which you know is run like cpuspeed).