I just spent 5 hours learning that if you supply features to GBrowse 2 via DAS (for example with my custom python based pyDasD server), their names must not start with a dash (’-').

Otherwise the (weird, backwards compatibility) heuristic in Bio::Graphics::Panel->_do_add_track will silently eat the first feature, because it thinks it’s an option.
Hooray for in band data transmission, I’d say.

So long,
Flo

OpenGrok isn’t difficult to set up - if you have a running Tomcat with a Java runtime 1.5, that is. Now, that’s a bit unussual for the shop I work with, so I decided to implement a central code browser using VMWare (Player) and a virtual Debian.

Step 1: Get yourself a Debian vm
Either download the vm found on Vmware’s VM Applicances site, or get the Debian Netinstall. I went with the Netinstall (and VBX Creator), since I wanted my VM to have two virtual harddrives, one for the Debian system and one for our code. This will allow me easily to replace the code harddisk later on with a bigger one (this turned out to be necessary a few years later).

After answering the basic installations questions ( manual package selection, nothing ), I suggest apt-get install ssh, so you can access the vm from your favourite SSH client - I use Putty when on windows.

Step 2: Uprade to Debian Etch
Since Tomcat 5 isn’t in the current stable Debian (sarge) we’ll switch to testing (Etch) instead.
#first, my favourite editor…
apt-get install joe
joe /etc/apt/sources.list
add contrib…

Step 3: Java

Now, power off the VM and make a backup. During the package creation for Java, you’ll save some largish files that will enlarge your VM. I’ll only sketch the step here.
First, we’ll need the most current Java Runtime.
#Webbrowser
apt-get install w3m
w3m http://java.sun.com/javase/downloads/index.jsp
Fetch the java se runtime

Now, we’ll create an installable java package
apt-get install java-package
su nonRootUser
fakeroot fakeroot make-jpkg jre-1_5_0_06-linux-i586.bin
(obviously change the file name for your java runtime)

SCP the created .deb file to a different machine, throw away the VM and start your backup.
Install the .dev here.

Step 4: Install Tomcat
apt-get install tomcat 5

Step 5: Fetch and install openGrok
Download openGrok and move the source.jar to /var/lib/tomcat5/webapps
Move openGrok.jar and lib to /opt/openGrok
Restart Tomcat (/etc/init.d/tomcat 5 restart )

Step 6: Mount the second harddrive
fdisk /dev/hdb # one partition is enough
mkfs.ext3 /dev/hdb5
joe /etc/fstab
…
mount /var/lib/tomcat5/webapps/source/code/
mkdir /var/lib/tomcat5/webapps/source/code/src
mkdir /var/lib/tomcat5/webapps/source/code/data

Step 6: Configure Tomcat
If your’re paranoid like me, Tomcat Apps need explicit permissions.
So place the following in your /etc/tomcat5/policies/04webapps.policy

grant codeBase “file:${catalina.home}/webapps/source/-” {
  permission java.util.PropertyPermission “subversion.native.library”, “read”;
permission java.lang.RuntimePermission “loadLibrary.svnjavahl-1″;
  permission java.lang.RuntimePermission “loadLibrary.libsvnjavahl-1″;
  permission java.lang.RuntimePermission “loadLibrary.svnjavahl”;
  permission java.util.PropertyPermission “disableLuceneLocks”, “read”;
  permission java.util.PropertyPermission “catalina.home”, “read”;
permission java.util.PropertyPermission “java.io.tmpdir”, “read”;
  permission java.util.PropertyPermission “org.apache.lucene.lockdir”, “read”;
  permission java.util.PropertyPermission “org.apache.lucene.writeLockTimeout”, “read”;
  permission java.util.PropertyPermission “org.apache.lucene.commitLockTimeout”, “read”;
  permission java.util.PropertyPermission “org.apache.lucene.mergeFactor”, “read”;
  permission java.util.PropertyPermission “org.apache.lucene.minMergeDocs”, “read”;
  permission java.util.PropertyPermission “org.apache.lucene.*”, “read”;
  permission java.io.FilePermission “/var/lib/tomcat5/temp”, “read”;
  permission java.io.FilePermission “/var/lib/tomcat5/temp/*”, “write”;
  permission java.io.FilePermission “/var/lib/tomcat5/temp/*”, “delete”;
};

grant codeBase “file:${catalina.home}/webapps/source/WEB-INF/lib/-” {
  permission java.util.PropertyPermission “subversion.native.library”, “read”;
  permission java.lang.RuntimePermission “loadLibrary.svnjavahl-1″;

  permission java.util.PropertyPermission “disableLuceneLocks”, “read”;
permission java.util.PropertyPermission “catalina.home”, “read”;
permission java.util.PropertyPermission “java.io.tmpdir”, “read”;
};

grant codeBase “file:${catalina.home}/webapps/source/WEB-INF/classes/-” {
permission java.util.PropertyPermission “subversion.native.library”, “read”;
  permission java.lang.RuntimePermission “loadLibrary.svnjavahl-1″;
  permission java.util.PropertyPermission “disableLuceneLocks”, “read”;
  permission java.util.PropertyPermission “catalina.home”, “read”;
  permission java.util.PropertyPermission “java.io.tmpdir”, “read”;
};

Restart Tomcat

Step 7: Fetch code
I use python and subversion (command line) to fetch our repositories each night, but then I already had a webbased repository list at hand.
apt-get install python subversion
Basically, I just call svn update on each working copy (or delete it and do a new checkout if that fails for some reason).
I could be using pysvn, but I never got it to work with Umlauts in the svn repository. Apperantly it simply ignores the locale I set…

Step 8: Create a shellscript and a cron
joe /opt/fetchAndIndexCode.sh
python /opt/fetchCode.py
java -jar /opt/openGrok.jar -s //var/lib/tomcat5/webapps/source/code/src /var/lib/tomcat5/webapps/source/code/data

chmod +x fetchAndIndexCode.sh
crontab -e

Step 9:
Run fetchAndIndexCode

Step 10: Visit http://yourip:8888/source

Step 11: Make tomcat accessible via Port 80
This is optional - for something more secure, I suggest using Apache and mod_proxy

I’d love to make it accessible from /, but apperantly there are some hardcoded paths in the webapp - and it’s not annyoing enough for me to work on that. I simple placed a meta redirect in /var/lib/tomcat5/webapps/ROOT/index.html

That’s it, I hope this quick guide has helped you.
I’d also love to get some comments, expieriences, improvements, etc.

So long,
Flo

Here’s a quick German guide how to not only cite Wikipedia correctly, but how to cite a page as it was on the day you looked at it.

Components (ie. required):

• a Sony Ericsson K700i (other Sony Ericsson’s should work - but I have none at hand to test them)
• Python
• A linux system with a running bluez stack (and an appropriate bluetooth dongle) (I used debian)
• AT Command reference for the K700i - available from http://developer.sonyericsson.com/
• Bluetooth (bluez) support for python

 

From small to even more bloody detailed…

AT commands were used back in the day to talk to modems (way before broadband ever was)… modern day cell phones still speak them, and they’re the easiest way to talk to your phone.

Bluetooth requires devices to be coupled. You’ll also need to know your phone’s bluetooth address. So turn on bluetooth on your phone and execute ‘hcitool scan’ in your shell.
You should see a line like 00:12:EE:55:10:F2       Name of your phone

00:12:EE:55:10:F2 would be your address. Note that for later.

Now, we’ll talk to your phone via bluetooth.
Fire up python, type the following.

from bluetooth import *;
s = BluetoothSocket (RFCOMM)
strAddr = “00:12:EE:55:10:F2″
x = s.connect ( ( strAddr , 2) ) # 2 is the channel. This might change from phone to phone.
s.send(”AT\r\n”)
print s.recv(1000)

you should get an ‘AT\r\nOK’ back.

If your phone wants you to pair, and you don’t know the pin, you might need to dig through your bluez settings.
On a shell only (no X) debian system, the standard pin tool bluez-pin won’t work (it requires GTK). I wrote a quick “echo PIN:1234″ shell script and (after chmod +x) set that in /etc/bluetooth/hcid.conf as my pin_helper

Now to getting your messages
Step 1
Choose the appropriate phone memory (SIM or built in).
The command in question is AT+CPMS=”ME” (don’t forget the newlines when sending!).

Step 2
Ask for all SMS - AT+CMGL=4.

Step 3
Get mad because your recv() returns without the phone having send all data.

Step 4 
Learn about select(). Then define your own wrapper
def readFromSock (s):
    i = [s]
    w = []
    e = [s]
    out = ”
    while True:
        ir, wr, er = select.select(i,w,e,3)
        if len(ir) == 0:
           break
        if len(er) > 0:
            break
        out += i[0].recv(1000)
        if (out.find(”OK\r\n”) != -1):
            return out
    return out

Step 5
Wonder at what gibberish you’re getting back.

No worries - the phone is dutifully returning all your SMS. Unfortunatly, they’re encoded in a format called PDU, defined in GSM 03.40 and GSM 03.38. Now there are some examples for pdu decoding out there. But still, it’s a headache ;-).

Koders.com, an open source code search engine, to the rescue:
I found a basic gpl’ed SMS decoder class written by a guy called Dominik Pytlewski.
It worked for some SMS., but it failed on some others, particulary outgoing SMS.
Attached you’ll find an improved version that seems to handle all my SMS ok. (out of sheer paranoia, I still store the PDU with every message, just in case of a decoding error).

Step 6
Explode output on “CMGL: “, strip, seperate body and header (message id - you’ll need this later to delete a particular message), turn into SMS decoder objects. Save the message id along with the decoded SMS.

Step 7
I split my messages into incoming and outgoing messages - but that’s just because the phone doesn’t provide a date for outgoing message, and I use a seperate deletion scheme for them.

Step 8
Do some clever ‘have I seen this message before’ caching (basically: Store last 1000 messages in PDU and use them as they keys for a dictionary. If has_key(), don’t bother mailing this message again)

Step 9
Assemble messages into a mail text, use smtplib to mail them wherever you like. Look up numbers from the cell’s phone book (retrieve (lazy evaluation) via AT+CPBS=”ME” (memory) and AT+CPBR=1,255 (get phonebook). Then split and store in a dictionary. Finally, look the phone numbers up (beware - they might be stored with a 0, but you get the international (+xy) number . Or the other way around. Think up some heuristic. Personally, I try both variants, then I remove the first three characters and search for a string match ).

Step 10
Decide which SMS you want to remove from the phone (I delete incoming SMS after seven days and keep only 10 sent SMS. But that’s a configuration option in the other attached file - blueMailSuck.py. Yeah, that’s what it does. It sucks SMS from your phone ;-).

Step 11
For each SMS to remove, send a AT+CMGD=message_id.

Step 12
Play a neat sound on the phone (AT+CRMP=5,0,1 for the alarm sound). I’d have displayed an image during the procedure, but apperantly the AT*EAPP command is broken and interprets only 4 letters of the text to pass in. Stupid if you *need* to specify a folder like images/whatever. (If you were playing the sound before another command, make sure to use socket.recv() instead of the select() version - otherwise your select will timeout and you’ll send the next command before the phone is ready).

I believe that’s pretty much it…
Feedback of course is always welcome.

So long,
Flo

Attached files:

• smsdecoder.py
• blueMailSuck.py

They not only got it published in the
Proceedings of the 4th European Conference on Computer Vision-Volume II in 1996,
no, they even got grant money for it!

You’re probably either here because I directed you here myself,
or because google thought I had a good article ;-).

So, exepect the usual… ramblings, tech, human interest, plain old weirdness.

Comments are on, but I’ll have to aprove your first comment.

 The logo image was created using

a photo of a butterfly © Michael Jastremski for openphoto.net which is under the CC:Attribution-ShareAlike licence
and another one - the ice cube/cherry creation - , gracefully provided to me by my favourite Punk Rock Princess Miri