Maintenance

Started my first attempt at building Jenkins2 with my desktop VirtualBox. Steps:

• Created a new Ubuntu 64-bit VM with 2GB of RAM and an expanding 25GB HD.
• Booted and installed from a downloaded 64-bit PC (AMD64) server install CD ISO.
• Let the installer determine the partitioning (it used LVM and created three partitions).
• ERRONEOUSLY chose the "Tomcat Java Server" option to start with. Note that this is not necessary; in fact Tomcat gets in the way, and I had to uninstall it later.
• Rebooted after install, did updates, and then installed curl.
• Downloaded and installed the Oracle JDK (downloaded it locally, put it on a server, and curled it to the VM). NOTE: There are Sun Java packages in the repos, I think; it would probably be better to install them, but I couldn't quite figure out how and I was in a rush. I think I could have done sudo apt-get install sun-java6-jre and sudo apt-get install sun-java6-jdk.
• Created /usr/local/java, copied the JDK binary there, made it executable and executed it.
• Added the following to /etc/profile:
  • JAVA_HOME=/usr/local/java/jdk1.6.0_25
  • PATH=$PATH:$HOME/bin:$JAVA_HOME/bin
  • export JAVA_HOME
  • export PATH
• Configured Java through update-alternatives:
  • sudo update-alternatives --install "/usr/bin/java" "java" "/usr/local/java/jre1.6.0_25/bin/java" 1
  • sudo update-alternatives --set java /usr/local/java/jre1.6.0_25/bin/java
• Rebooted and tested with java -version that the JDK is now default the Java VM.
• Tried to do waiting updates with apt-get, but they were held back, so did them with aptitude. Why is this necessary?
• Couldn't find Tomcat, so did sudo apt-get install tomcat6, and installed it. This seemed to update the openjdk too, but it didn't mess with the configuration of the Oracle Java.
• After reading around a bit, I decided I'd rather use Tomcat7, so I created /usr/local/apache-tomcat-7.0.12, and used wget to download the tar.gz file, and then did tar xvzf to unwrap it.
• Added my usual utf8_startup.sh.
• Ran Tomcat and confirmed it was working by curling 127.0.0.1:8080.

Next I wanted to find a convenient way to access the VM's Tomcat from the host, so I ran this in a terminal on the host, after shutting down the vm:

VBoxManage modifyvm "Jenkins2" --natpf1 "tomcat,tcp,,9090,,8080"

This should set up port forwarding between the host desktop port 9090 and the Tomcat running on port 8080 in the VM.

The port-forwarding worked like a charm, but in the process I discovered that Tomcat 6 is set to run as a service. So I have Tomcat 7 available, but I'll stick with Tomcat 6 from the repos for the moment, unless I find a reason to abandon it.

NOTE: All of the Tomcat stuff was unnecessary, but the Sun/Oracle Java is required.

• Added the Jenkins key and repo as explained here, and installed Jenkins.
• Having done that, I realized that it's actually starting by itself on port 8080; it doesn't need to run inside Tomcat. So I edited its script in /etc/init.d/jenkins to add this to the arguments: --httpPort=8081. This should make it start on 8081 and not conflict with Tomcat.
• Then I shut down the VM and ran this on the host:

  VBoxManage modifyvm "Jenkins2" --natpf1 "jenkins,tcp,,9091,,8081"
  

  which should let me see Jenkins on 9091 on the host, without having to turn off Tomcat for the moment. However, in future, we should probably just leave Tomcat out of it.
• Despite my re-configuration, Jenkins kept trying to start on 8080. Meanwhile, I got sick of the limited terminal available in the headless vm, so I did sudo apt-get install openssh, and tested it, then shut down the VM and ran this:

  VBoxManage modifyvm "Jenkins2" --natpf1 "guestssh,tcp,,2222,,22"
  

  Now I can connect to the VM using:

  ssh -l hcmc -p 2222 localhost
  

  which means I can easily copy/paste from browsers, use a larger terminal, etc.
• So then I went back and got rid of Tomcat (sudo apt-get remove --purge tomcat6), and then copied my backup of the /etc/init.d/jenkins back over the edited one to remove the attempt to reconfigure the httpPort. Then a restart gave me.... A WORKING JENKINS on port 8080.

Next, install texlive and friends:

• In the user's home directory, mkdir texlive, then cd texlive and wget http://mirror.ctan.org/systems/texlive/tlnet/install-tl-unx.tar.gz.
• tar xvfz install-to-unx.tar.gz
• cd install.....
• sudo ./install-tl

Texlive installed. Then I had to do sudo apt-get install subversion, because I'll be needing that.

Next, I followed my own instructions to get TEI installed.

Then in trunk/P5 I did make dependencies, and then sudo apt-get install [the list of dependencies].

However, an attempt to build P5 failed:

make
Checking you have running XML tools and Perl before trying to run transform...
Perl:/usr/bin/perl
xmllint:make: *** [check.stamp] Error 1

Turns out xmllint is in libxml2-utils, so had to sudo apt-get install libxml2-utils. Now a make is proceeding as expected.

Next, I wanted to do some cleanup -- on login, there was a message to the effect that there were packages available for upgrade, but the message was wrong (see discussion here). So I did sudo rm /etc/motd.tail.

Next, I tried make pdf, which is one of the targets that failed for me on my desktop due to the absence of TeXLive 2010. It failed again for the same reason; although I've installed TeXLive 2010, the system has at least one symbolic link (from /usr/bin/xelatex to /usr/bin/xetex) which are from the old Debian 2009 Tex package.

So the instruction on the TEI wiki page for installing packages, which includes texlive, is wrong; that shouldn't be installed. I've now removed it (sudo apt-get remove texlive). Then I added the install location of the 2010 to my path, by doing sudo nano /etc/profile, and adding PATH=$PATH:/usr/local/texlive/2010/bin/x86_64-linux to it.

Then I tried make pdf again, and got the same error, so it's not caused by the texlive version after all. Although the error message suggests that the error is caused by a space in a font name, I think it's actually caused by the absence of the font. Reported this to SR, who modified the source so that the Minion Pro and Myriad Pro fonts are no longer required (they're proprietary anyway). Now I can build a PDF, but I'm still getting an error at the end.

Looking in job$.log (after deleting it, and re-running the make pdf to be sure it's fresh), I see that it still starts with this:

hcmc@Jenkins2:~/tei/trunk/P5$ more job\$.log
kpathsea: Invalid fontname `Myriad Pro', contains ' '
kpathsea: Invalid fontname `Myriad Pro', contains ' '
kpathsea: Invalid fontname `Myriad Pro', contains ' '
kpathsea: Invalid fontname `Myriad Pro', contains ' '
kpathsea: Invalid fontname `Myriad Pro/B', contains ' '
kpathsea: Invalid fontname `Myriad Pro', contains ' '
kpathsea: Invalid fontname `Myriad Pro/I', contains ' '
kpathsea: Invalid fontname `Myriad Pro', contains ' '
kpathsea: Invalid fontname `Myriad Pro/BI', contains ' '
kpathsea: Invalid fontname `Myriad Pro:', contains ' '

The rest of the file consists mainly of

** WARNING ** Failed to convert input string to UTF16...

but there are two more Myriad Pro errors towards the end, and finally there's this:

kpathsea: Invalid fontname `HAN NOM A', contains ' '
kpathsea: Invalid fontname `HAN NOM A', contains ' '
kpathsea: Invalid fontname `HAN NOM A', contains ' '
kpathsea: Invalid fontname `HAN NOM A/B', contains ' '
kpathsea: Invalid fontname `HAN NOM A', contains ' '
kpathsea: Invalid fontname `HAN NOM A/I', contains ' '
kpathsea: Invalid fontname `HAN NOM A', contains ' '
kpathsea: Invalid fontname `HAN NOM A/BI', contains ' '
kpathsea: Invalid fontname `HAN NOM A:', contains ' '

with lots more instances of the same thing. So I suspect there are still references to Myriad Pro in the code, and there's also this other font, which I've never heard of. It doesn't seem to be available in the repos either, as far as I can tell (there's no ttf-hannom, although Arch Linux seems to have one). Waiting to hear from SR where he got it from.

After JN had to check out the current source and build to get around a recent bug, I took the opportunity to go through the same process, create an eXist war file:

• mkdir exist_trunk
• cd exist_trunk
• svn co http://exist.svn.sourceforge.net/svnroot/exist/trunk/eXist .
• ./build.sh clean
• ./build.sh download-additional-jars
• cd extensions
• cp build.properties local.build.properties
• [Edit local.build.properties to turn on FOP.]
• cd ../
• ./build.sh
• ./build.sh -f build/scripts/jarsigner.xml
• ./build.sh dist-war
• mkdir [tomcat]/webapps/exist
• cp dist/exist-1.5....war [tomcat]/webapps/exist
• jar xfv [tomcat]/webapps/exist/exist-1.5...war

Then I started Tomcat 6.0.26, but eXist failed to start. So I installed the latest Tomcat (7.0.12). It took me a while to get this going, but I eventually figured out that all the .sh and all the .jar files in the bin directory need to be executable. I used my usual utf8-startup.sh file to start it up, and exist came up working fine. So the latest eXist seems to prefer the latest Tomcat.

SR has removed some build targets and components from the makefile, which means I can now build P5 without errors, but some of the issues will need to be handled at some point -- for instance, the error in building the PDF is due to the need for TeX Live 2010, whereas current Ubuntu versions are stuck at 2009. Still working on a detailed plan for creating a clone Jenkins machine, for which we'll need to solve all these problems. Here's some bits from email for the record (our suggested plan and SR's responses):

 -Start with an Ubuntu 10.04 headless server install.
  
   -Install the Oracle JDK.
  
   -Install Tomcat (any particular version? Latest? What's in repos?).
doesnt matter, version 6 or 7 latest

  
   -Install Jenkins from here:
     http://pkg.jenkins-ci.org/debian/
  
   -Install Tex Live 2010 from here:
     http://www.tug.org/texlive/acquire.html
  
   -Follow these instructions to get a working build setup:
     http://hcmc.uvic.ca/blogs/index.php?blog=15&p=8081
  
Jenkins will do its own checkout, so no need for some of that

   -???Profit?
pleasure

  
  I'm pretty sure we'll also have to install some extras (fonts and the 
  like),
note the result of 

    make dependencies

Background: lab machines are DHCP, therefore the IP address cannot be counted on, nor can the canonical name be set in DNS (apparently). rSnapshot requires a canonical something (obviously) so it know where to go to grab the snapshot.
Solution: put each client machine in the hosts file so rSnapshot knows where each machine is. This of creates a
Problem: if the client machine's IP address changes, the hosts file on the rSnapshot server needs to be adjusted.

I looked at using nmap/arp to do this job, but came to the conclusion that the results were unreliable. Occasionally, a machine would not report back to the request, making hte script unreliable. For posterity, this is how the script worked:
1) set var="MAC address" for each machine (e.g. machineName="00:DE:6F:34:34:00")
2) set ipArr=( $(nmap -sP 192.168.16.192/26|grep -oh '[0-9]\{3\}\.[0-9]\{3\}\.[0-9]\{3\}\.[0-9]\{3\}'| tr '\n' ' ') ) - which will output an array that looks something like this:
(192.168.16.197 192.168.16.198 192.168.16.203 192.168.16.204 192.168.16.206 192.168.16.214 192.168.16.217 192.168.16.220 192.168.16.223 192.168.16.228 192.168.16.230 192.168.16.237 192.168.16.241 192.168.16.242 192.168.16.250 192.168.16.254)
3) iterate through the array and run arp on each IP:
for (( i = 0 ; i < ${#active_IP_addresses[@]} ; i++ ))
do
echo `arp -a ${active_IP_addresses[$i]}`
done
~~~~
which will produce output that looks like this:
machine.name.as.found.in.remote.hosts.file (192.168.16.X) at XX:XX:XX:XX:XX:XX [ether] on eth0
ad nauseum

Once you get all that you can wrangle it in to a hosts file for the rSnapshot machine. As I said, though, sometimes it won't be complete - and I have no idea why; it remains a more elegant solution if it can be made more reliable. I say this because it would all run from the server. The solution I'm currently working on is a local/remote combo that I just don't like.

So the solution I have right now is to enter each machine in to the rSnapshot config file and set a cron job to grab a snapshot (for this example let's say that the daily snapshot will happen at 4pm).
On each client machine, set a cron job that sets the hostname and current IP address and sends it to the rSnapshot server. Here's how I'm doing this:
1) Set up SSH-keys on the client so it can send stuff to the server on its own.
2) Run this doozy on the client end to add *this* client's info to the rSnapshot server's /etc/hosts file, removing old entries for *this* machine along the way:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
rhost=$(h=$(echo -e $HOSTNAME);echo $h `ifconfig eth0|grep 'inet addr:'|cut -d ":" -f2|sed 's/ Bcast//gI'`) ; ssh rsnapshot.server.name "sed -i /\"$h\"/d /etc/hosts;echo $rhost >> /etc/hosts"
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Right now I'm working on getting it running from cron but I'm having some trouble with it. When I just dump the above in to a bash script and run it from the command line (or from cron)
I get an error: "sed: -e expression #1, char 2: unterminated address regex" which I do not currently understand.

However, here's how it works:
1) echo -e $HOSTNAME returns the name you gave the machine
2) `ifconfig eth0|grep 'inet addr:'|cut -d ":" -f2|sed 's/ Bcast//gI'` gets you the IP address for the machine by asking ifconfig for all the info it has on eth0 (the NIC), chopping it up and discarding extraneous stuff (the *cut* bit returns the second field after the colon, while the *sed* bit deletes the stuff after the IP address - in this case that's " Bcast").
3) wrap this up in a pair of variables (h being the hostname and rhost being the hostname IP address combination). This is the first part of the oneliner: rhost=$(h=$()).
4) once we have values for host name and IP address (h and rhost) we head over to the remote machine via ssh and send the remote machine (our rSnapshot server) a couple of commands. We do this by adding a quoted command to the end of the ssh invocation. We do this with double quotes; double-quotes will expand variables, single-quotes won't and we have variables.
5) the commands on the rSnapshot side go like this:
sed -i /\"$h\"/d /etc/hosts (which removes any lines, using the d command, which include our host name - this gets rid of old entries)
echo $rhost >> /etc/hosts (which adds an entry for the client to the hosts file)

Things I need to do to make this really useable:
1) make it work from inside a script that cron can successfully run (see above for the problem)
2) adjust the remote sed command to only remove lines that begin with hostname. Right now, this will remove *any* line that includes the hostname at all. For example, consider a file like this:

tron 192.168.1.1
megatron 192.168.1.2
neutron 192.168.1.3

If I want to edit the line for the machine called tron and I use my sed command up above I'll end up with an empty file because each machine includes the string "tron". I don't name machines in a way that this would be a problem, but it's a pretty big problem if I inadvertently do something stupid when naming a new machine.

Huge progress today.
I have a virtual rsnapshot environment working that automatically backs up a remote machine once per day. here's how:
1) install rsnapshot from the repos
2) configure it by editing the conf file at /etc/rsnapshot.conf
* make sure to edit "snapshot_root"
* uncomment "cmd_cp"
* uncomment "cmd_ssh"
* edit intervals to something useful, like "interval daily 5" - commenting out intervals not being used. This setting takes 5 snapshots per week.
* set log file location
* enable lockfile (rsnapshot should run as root)
* set includes and excludes if needed - I set up an excludes file
* set backup points - one for each machine: e.g. "backup user@machinename/ip:/home/netlink/ machinename/"

3) Prime the pot by manually running a backup: sudo rsnapshot daily
4) Then, set a crontab for running things every weekday at 6pm:
crontab -e
0 18 * * 1,2,3,4,5 /usr/bin/rsnapshot daily
5) Wait until 6pm

I've yet to test this on a real server/machine, but I'll set up Titchy as an rsnapshot server to take the next step.
Also to do: adjust rollover so that stale snapshots don't just get deleted, the latest version is always archived somewhere. Is this rdiff?
NOTE: rsnapshot handles the above issue on its own.

NOTE: when rsnapshot runs from the collector end (i.e. pulling *from* a client) it invokes a process on the client end that shows up as "rsync --server --sender -logDtprRe.iLsf --numeric-ids . /home/gregster/Desktop" where "/home/gregster/Desktop" is the path to the stuff getting backed up. This will be useful when locking down legal processes using single-use ssh-keys.

Small fix to the english grad apps program that allows reviewers to review PhD and MA applications online. The names of applicants are passed to a Javascript function that is triggered upon button click, like this (names used are not real):

<button onclick="showform('LASTNAME-Firstname-PhD')">Assess...</button>

However, if the name had an apostrophe, it was breaking the Javascript:

<button onclick="showform('LAST'NAME-Firstname-PhD')">Assess...</button>

So, I simply escaped the names with addslashes().

After discovering the excellent WineBottler package of Wine for the Mac, and testing it with one of my personal project (Markin) at home, I was hopeful that it might provide a means to run the Image Markup Tool on the Mac. GN and I thoroughly tested a variety of solutions: WineBottler, PlayOnMac, and Wineskin. All did more or less the same thing: install Wine (and sometimes XQuartz as well), and build a customized runtime environment (a "prefix") for the application, bundling it into a .app package. Everything worked well with the exception of the graphics handling. Opening a normal Mariage-sized file took several minutes, during which the app was unresponsive. This is in contraxt to IMT behaviour on Linux Wine, where everything works as normal (if anything, a bit snappier than Windows). Our conclusion is that the mapping of Windows graphics calls to the OSX graphics system is not as sophisticated as it is on Linux, so the fairly intensive operations the IMT does to resize and zoom images, which use the Graphics32 library, run extremely slowly. Maybe this will solve itself. We tested with Wine 1.1.4 and 1.2.2.

ER passed on some error messages from Tomcat-stable:

2011-01-25T13:40:03-08:00 
pear.hcmc.uvic.ca/pear.hcmc.uvic.ca/pear.hcmc.uvic.ca daemon.err tomcat: 
hcmc-stable
[ajp-8009-4]: Use of deprecated, since 2008-04-02, function 
fn:item-at(). It will be removed really soon. Please use
$x[1] instead.

It's not clear which project they're from, but item-at is used in xrequest.xq in Graves (line 365). I'll have to test replacing it.

I'm working on an updated version that includes stuff like setting up a printer by default, providing a choice of lab or developer build (extra apps, no ldap login).
I've got the printer going by creating a /etc/cups/printers.conf file with info on the printer (make, model etc.) and putting a pre-made PPD at /etc/cups/ppd/HCMC.ppd. Finally, printers.conf needs to be chgrp'd to group "lp" and the cups service needs to be restarted.
I'm now working on making it a bit more informative (echos out what is about to happen, provides choices etc.)

ANOTHER UPDATE: bizarre! When I first power-up and log in the script either doesn't fire or just doesn't work. If I log out and back in - it does! WTF?

UPDATE: this is an annoyingly difficult thing to nail down, but I think I've got it now.
I wrote a script called mice and put it in my path at: /usr/local/bin/mice
Then, I add it to my startup applications pref panel like this:
sh -c '/usr/local/bin/mice'
And it seems to work.
Here's the script I wrote
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
#!/bin/bash

# get ID and current set up for each input device
emID=`xinput list|grep 'Kensington Kensington Expert Mouse'|cut -d "[" -f1|grep -o [0-9].`
stID=`xinput list|grep 'Kensington Kensington Slimblade Trackball'|cut -d "[" -f1|grep -o [0-9].`
emButtonMap=`xinput --get-button-map 'Kensington Kensington Expert Mouse'`
stButtonMap=`xinput --get-button-map 'Kensington Kensington Slimblade Trackball'`
emAccel=`xinput --list-props 'Kensington Kensington Expert Mouse'|grep 'Constant'|tr "\t" "|"|cut -d "|" -f3`
stAccel=`xinput --list-props 'Kensington Kensington Slimblade Trackball'|grep 'Constant'|tr "\t" "|"|cut -d "|" -f3`
emBMap="3 2 1 5 4 6 7 8 9 10 11 12"
stBMap="1 8 3 4 5 6 7 2 9 10 11 12"
emDACD="3.000000"
stDACD="3.000000"

# check current settings against desired setup and change if necessary
if [ "$emButtonMap" != "$emBMap" ]; then
xinput --set-button-map $emID $emBMap
emBmapCurr=`xinput --get-button-map 'Kensington Kensington Expert Mouse'`
tsEm=`date +"%Y-%m-%d %H:%M:%S"`
echo "[$tsEm] set Expert Mouse (device ID: $emID) button map to $emBMapCurr">> /home/gregster/.hcmc/mice.log
fi

if [ "$stButtonMap" != "$stBMap" ]; then
xinput --set-button-map $stID $stBMap
stBMapCurr=`xinput --get-button-map 'Kensington Kensington Slimblade Trackball'`
tsSt=`date +"%Y-%m-%d %H:%M:%S"`
echo "[$tsSt] set Slimblade Trackball (device ID: $stID) button map to $stBMapCurr" >> /home/gregster/.hcmc/mice.log
fi

if [ "$emAccel" != "$emDACD" ]; then
xinput --set-prop $emID 'Device Accel Constant Deceleration' 3
emDACDCurr=`xinput --list-props 'Kensington Kensington Expert Mouse'|grep 'Constant'|tr "\t" "|"|cut -d "|" -f3`
tsEmAc=`date +"%Y-%m-%d %H:%M:%S"`
echo "[$tsEmAc] set Expert Mouse (device ID: $emID) acceleration to $emDACDCurr" >> /home/gregster/.hcmc/mice.log
fi

if [ "$stAccel" != "$stDACD" ]; then
xinput --set-prop $stID 'Device Accel Constant Deceleration' 3
stDACDCurr=`xinput --list-props 'Kensington Kensington Slimblade Trackball'|grep 'Constant'|tr "\t" "|"|cut -d "|" -f3`
tsStAc=`date +"%Y-%m-%d %H:%M:%S"`
echo "[$tsStAc] set Slimblade Trackball (device ID: $stID) acceleration to $stDACDCurr" >> /home/gregster/.hcmc/mice.log
fi

exit

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

In order to get my two trackballs working as I like (one left-handed, one right-handed) I tried a few ways but ended up going with adding instructions to the System > Preferences > Startup applications applet. (BTW, this adds .desktop files to ~/.config/autostart/ directory).

The trick is to get your syntax just right. Running something at the terminal is not simply duplicated in the Startup applications applet. I ended up creating one startup app for each instruction (doing many in one eluded me). This is what I put in the "command" field for each one (that is, each of the following lines is the meat of one startup app):
sh -c "xinput --set-prop 'Kensington Kensington Expert Mouse' 'Device Accel Constant Deceleration' 4"
sh -c 'xinput --set-button-map 'Kensington Kensington Expert Mouse' 3 2 1 5 4 6 7 8'
sh -c "xinput --set-prop 'Kensington Kensington Slimblade Trackball' 'Device Accel Constant Deceleration' 4"
sh -c 'xinput --set-button-map 'Kensington Kensington Slimblade Trackball' 1 8 3 4 5 6 7 2'

You have to use an invocation like sh to get it to run, and sh needs the -c flag to know that it's operating the stuff after it outside of the context of standard input (ie. running it directly from the CLI). Also, strings with spaces (e.g. Device Accel Constant Deceleration) need to be enclosed in SINGLE-QUOTES. Standard practice seems to put the entire command following sh -c in single-quotes, but in this case I use double-quotes to avoid painful escaping.

So, one of my .desktop files looks like this:

[Desktop Entry]
Type=Application
Exec=sh -c "xinput --set-prop 'Kensington Kensington Expert Mouse' 'Device Accel Constant Deceleration' 4"
Hidden=false
NoDisplay=false
X-GNOME-Autostart-enabled=true
Name[en_CA]=Add accel for ExpertMouse
Name=Add trackball configs
Comment[en_CA]=accel for ExpertMouse
Comment=eccel for ExpertMouse

NOTE: unfortunately, this does not work all the time. Re-investigating.

UPDATE: this is ridiculously hard. It looks like even if I put the commands in my .profile something else resets the button-mapping. I'm going to try udev again. Of course, the problem I see with this actually working is that it's a system level adjustment, and it *should* be per-user.

Anyway, here's the way I'm proceeding:

1) Gather info on the device by running this:
udevadm info -a -p $(udevadm info -q path -n /dev/input/by-id/usb-Kensington_Kensington_Slimblade_Trackball-mouse)
2) Unplugging/plugging a mouse while tailing the Xorg log gets this output:
[ 28709.701] (II) config/udev: removing device Kensington Kensington Expert Mouse
[ 28709.702] (II) Kensington Kensington Expert Mouse: Close
[ 28709.702] (II) UnloadModule: "evdev"
[ 28717.858] (II) config/udev: Adding input device Kensington Kensington Expert Mouse (/dev/input/mouse0)
[ 28717.859] (II) No input driver/identifier specified (ignoring)
[ 28717.859] (II) config/udev: Adding input device Kensington Kensington Expert Mouse (/dev/input/event2)
[ 28717.859] (**) Kensington Kensington Expert Mouse: Applying InputClass "evdev pointer catchall"
[ 28717.859] (**) Kensington Kensington Expert Mouse: always reports core events
[ 28717.859] (**) Kensington Kensington Expert Mouse: Device: "/dev/input/event2"
[ 28717.880] (II) Kensington Kensington Expert Mouse: Found 8 mouse buttons
[ 28717.880] (II) Kensington Kensington Expert Mouse: Found scroll wheel(s)
[ 28717.880] (II) Kensington Kensington Expert Mouse: Found relative axes
[ 28717.880] (II) Kensington Kensington Expert Mouse: Found x and y relative axes
[ 28717.880] (II) Kensington Kensington Expert Mouse: Configuring as mouse
[ 28717.880] (**) Kensington Kensington Expert Mouse: YAxisMapping: buttons 4 and 5
[ 28717.880] (**) Kensington Kensington Expert Mouse: EmulateWheelButton: 4, EmulateWheelInertia: 10, EmulateWheelTimeout: 200
[ 28717.881] (II) XINPUT: Adding extended input device "Kensington Kensington Expert Mouse" (type: MOUSE)
[ 28717.881] (II) Kensington Kensington Expert Mouse: initialized for relative axes.

Some links I've been using:
http://reactivated.net/writing_udev_rules.html
https://wiki.ubuntu.com/X/Config/Input
https://wiki.kubuntu.org/X/InputConfiguration