Pop Qvarnström: Attention Surplus

Ubuntu 10.04.3 Server LTS on Hyper-V part 2

2012-01-19T09:22:00.000+01:00

Adding a few details about the installation and the immediate things after installation:

When configuring the network, if you installed without a legacy network adapter, use eth, not seth.
The console via Hyper-V manager is painstakingly slow. To remedy this, blacklist the frame buffer module: add the line blacklist vga16fb to the file /etc/modprobe.d/blacklist-framebuffer.conf and restart the vm.

[Re]sources/credits:

http://serverfault.com/questions/138110/ubuntu-10-04-server-on-hyper-v-server-r2-has-sluggish-install-and-command-line

http://jasonbos.co/fix-slow-refresh-rate-on-hyper-v-console-runn

http://screamingadmin.wordpress.com/2010/07/08/vga16fb-and-ubuntu-10-04-server/

http://solyaris.wordpress.com/2010/05/01/ubuntu-10-04-on-intel-driver-and-server-frame-buffer/

Installing Ubuntu 10.04.3 LTS on a Hyper-V virtual Machine

2012-01-18T18:01:00.000+01:00

tl;dr Either configure a legacy network adapter or skip installing any extra packages until after installing the Hyper-V kernel modules and configuring the synthetic network adapter.

I just made my first contact with Hyper-V on Windows Server 2008 R2, which is the virtualization environment of choice at the company I work for. The production environment for the service I'm working on is a Xen based virtual machine running Ubuntu Server 10.04.3 LTS, and we need a local test server with a similar setup. Installing Ubuntu on a virtual machine has never before been a problem (VMware, VirtualBox), so I expected it to be a smooth ride: there are Linux kernel modules for Hyper-V available, and in newer Ubuntu releases they are included on the distribution image. In 10.04 the kernel modules are not included on the distribution image, but are available separately. There are also a few brief guides to installing 10.04 on Hyper-V available on the net, but none mentioned any real problems.

The first thing I ran into was that no network adapter was found. This was mentioned in pretty much all installation guides, so no biggie - after installation is done we'll just install the Hyper-V kernel modules and do a little bit of configuration and the synthetic network adapter will be found. Next, I chose a few packages to install: postfix, postgres, tomcat. Installation started, a few configuration questions popped up for postfix. Then: BAM! Big, red screen telling me an installation step failed. Meh. Retried the installation step, same result. Meh.

Popping out to a console and checking the installation log (syslog, really) revealed what had happened: during postfix configuration, no network interface was found (as expected). This made the configuration step fail for the postfix installation, which in turn caused the whole installation to fail. Meh.

Solution? Either configure a legacy network adapter to use during installation, or don't install any extra packages (at least not postfix) until after installing the Hyper-V kernel modules and configuring the network.

Resources:
http://social.technet.microsoft.com/wiki/contents/articles/961.aspx
http://fawzi.wordpress.com/2010/05/03/ubuntu-server-10-4-in-hyper-v/
http://blogs.technet.com/b/pascals/archive/2010/05/03/ubuntu-server-10-4-dans-hyper-v-avec-les-drivers-hyper-v.aspx
http://blogs.technet.com/b/port25/archive/2009/07/22/introduction-to-the-linux-integration-components.aspx
http://blog.allanglesit.com/2010/05/ubuntu-and-hyper-v-the-paths-to-enlightenment/
http://blogs.technet.com/b/enterprise_admin/archive/2010/03/09/linux-and-hyper-v-the-easy-way-brown-and-serve.aspx

Wordfeud on Android ICE broken

2012-01-13T12:52:00.000+01:00

Since a google search shows no results, I must create one: currently Wordfeud does not work on Android Ice Cream Sandwich. Or, it does, but not on 3G connections - if you're on wifi it will work.

The error manifests itself by showing a small "Logged in" text, quickly followed by a series of popups (at least 3 at a time for me) with the text "Login failed Please try again later" and an Ok button.

Try to do anything after dismissing the popups, an error message shows: "Server error Make sure you're [sic!] Wordfeud installation is up to date. Error details: com.hbwares.wordfeud.lib.WerdFeudService$ProtocolException: Unexpected error type: login_required Message;".

Solution? Wait for Wordfeud to fix this or be on wifi.

Meh.

Email from Wordfeud support:

Hi,

We are aware of an issue with Android 4 Ice Cream Sandwich that causes problems with logging in to Wordfeud when connected by 3G, and are working to fix it.
WiFi login should work as normal.

Best regards,
Odin Borg Nordgård
Wordfeud.com

Bankid (Nexus Personal) on Ubuntu 11.04 64-bit

2011-06-11T00:02:00.000+02:00

As usual, when I need my bankid it does not work. Usually this is because I got a new computer or installed a new OS or something equally disruptive to bankid functionality. Every time I have to search the internet for information on how to get it to work, and every time I find pieces of information in different places. That is annoying. Here are my reminder notes for next time.

What I needed to do:

1. Download and install Nexus Personal

This was straightforward: go to https://install.bankid.com/ and follow the instructions to download the latest version. Unpack the tar.gz file, switch working dir to the directory with the archive contents and run the install script (install.<version>.sh i) as root.

2. Copy my bankid from another computer, or get a new one from my bank if it has expired.

This was somewhat more convoluted: The linux GUI does not include import/export functionality, but refers to the program persadm. For import you use persadm import, which prompts you for a file name and, if the file is found, a password/pin for the bankid in that file.

I had an export file of my active bankid, with a name like (110502 20.47) Pop Qvarnström - BankID på fil.nge. Normally, you will have to escape the spaces and parentheses in the file name, so using command completion to get the full path to the file will generate something perfectly reasonable like Desktop/sticka/bankid/$110502\ 20.47$\ Pop\ Qvarnström\ -\ BankID\ på\ fil.nge. Sadly, copying this will fail. No escaping allowed in persadm paths. You need to provide the full path, no escapes, and the program does not support completion.

Easy enough to provide the file name, next up is the password. Can't help you there, but if you - like me - get an error even though you provide a correct password, the reason might be that you are only allowed to import an exported bankid once. Any following import attempts will fail. You will not get any hints about that in the error message: Error importing token.

Create a new export file and things will work.

3. Make sure that nspluginwrapper knows about this 32-bit plugin: sudo nspluginwrapper -i /usr/local/lib/personal/libplugins.so

Oh, and you also need to have ia32-libs installed.

Restart firefox and you should be good to go.

Skype problems on Ubuntu 11.04 64-bit

2011-05-30T01:49:00.000+02:00

tl;dr If Skype is refusing to start correctly, crashing a few seconds after startup, remove ~/.Skype/shared.xml and restart Skype.

Since Friday morning Skype has been acting up, refusing to start (or rather crashing after a few seconds). This caused serious problems for me, trying to manage a geographically dispersed team spread over several countries using skype as the main means of communication.

Starting Skype from a terminal gave a lot of error messages, mostly regarding 32-bit expectations and 64-bit files:

me@mycomputer:~/$ skype
/usr/lib/gtk-2.0/2.10.0/menuproxies/libappmenu.so: wrong ELF class: ELFCLASS64
(:13947): Gtk-WARNING **: Failed to load type module: /usr/lib/gtk-2.0/2.10.0/menuproxies/libappmenu.so
/usr/lib/gtk-2.0/2.10.0/menuproxies/libappmenu.so: wrong ELF class: ELFCLASS64
(:13947): Gtk-WARNING **: Failed to load type module: /usr/lib/gtk-2.0/2.10.0/menuproxies/libappmenu.so
(:13947): Gtk-WARNING **: /usr/lib/gtk-2.0/2.10.0/immodules/im-ibus.so: wrong ELF class: ELFCLASS64
(:13947): Gtk-WARNING **: Loading IM context type 'ibus' failed
(:13947): Gtk-WARNING **: /usr/lib/gtk-2.0/2.10.0/immodules/im-ibus.so: wrong ELF class: ELFCLASS64
(:13947): Gtk-WARNING **: Loading IM context type 'ibus' failed
/usr/lib/gtk-2.0/2.10.0/menuproxies/libappmenu.so: wrong ELF class: ELFCLASS64
(:13947): Gtk-WARNING **: Failed to load type module: /usr/lib/gtk-2.0/2.10.0/menuproxies/libappmenu.so
/usr/lib/gtk-2.0/2.10.0/menuproxies/libappmenu.so: wrong ELF class: ELFCLASS64
(:13947): Gtk-WARNING **: Failed to load type module: /usr/lib/gtk-2.0/2.10.0/menuproxies/libappmenu.so
(:13947): Gtk-WARNING **: /usr/lib/gtk-2.0/2.10.0/immodules/im-ibus.so: wrong ELF class: ELFCLASS64
(:13947): Gtk-WARNING **: Loading IM context type 'ibus' failed
Aborted

Reinstalling Skype or purging configuration did nothing. Installing a freshly downloaded version from skype.com did nothing. Checking the ia32-libs install did nothing. Bah. A bit of searching revealed the solution: remove ~/.Skype/shared.xml and restart Skype.

Now, if only the Skype statusbar icon would come back, or grow a bit larger than 1 pixel. That would be super, really.

Sources:
http://heartbeat.skype.com/2011/05/an_update_on_todays_downtime.html
https://bugs.launchpad.net/ubuntu/+source/skype/+bug/788573/comments/37
http://heartbeat.skype.com/2011/05/problems_signing_into_skype_an.html

https://bugs.launchpad.net/ubuntu/+source/skype/+bug/764473

sl -a

2011-05-23T16:58:00.000+02:00

Funniest find on my Linux box for a while: sl -a. I'm obiously a sloppy typer, and this beats any error message I've seen so far. You get trains! Install and enjoy.

Flash in FF4 on 64-bit Ubuntu 10.10

2011-03-27T03:52:00.000+02:00

Meh.

Flash refused to work with Firefox 4, wasn't even listed in about:plugins. Why? Because the linux download offered from mozilla.com was a 32-bit binary, and my flash plugin was a 64-bit binary. Über-meh.

Grabbing the 64-bit version from ftp://ftp.mozilla.org/pub/firefox/releases/latest-4.0/linux-x86_64/en-US/ did the trick. I really should have read the console output much, much sooner.

(LoadPlugin: failed to initialize shared library /home/popq/.mozilla/plugins/libflashplayer.so [/home/popq/.mozilla/plugins/libflashplayer.so: wrong ELF class: ELFCLASS64] was a pretty strong hint.)

Converting a 2-up PDF into 1-up

2011-03-23T14:34:00.000+01:00

TL;DR - If you want to split pages in a PDF, try briss.

This morning I downloaded a free PDF book, planning to read it on my Kindle DX. Most PDF documents show just fine on the DX with default zoom, but 2-up documents, i.e. documents with two pages per PDF page, tend to be a little hard to read without resorting to zooming and thus displaying only part of a page at a time. This particular book turned out to be created that way. That sucks, so I started to look for a tool that could split PDF pages.

Turns out there are lots of tools that can manipulate pdf's in all sorts of useful ways, but few that do what I wanted.

After looking at pdftk, pdfmerge, pdfshuffler, pdfedit, unpaper, poster and more, I stumbled over briss, which did exactly what I wanted.

Simple usage: start briss, open pdf, mark crop areas (each area becomes a separate page), save.

Note: pdfshuffler also allowed me to do this, but required more work: easier crop (creating a pdf with left pages and one with right pages), but manual merge (inserting a left page next to each right) in gui made it too much work for a document with several hundred pages.

Seting default editor in Ubuntu 10.10 to the reasonable value: vim

2011-03-21T12:45:00.000+01:00

For some reason, the default editor in Ubuntu 10.10 seems to be nano. Why on earth anyone would prefer that over vim is a complete mystery to me, but luckily it's easy enough to configure this.

Assuming you have access to root privileges:

1. Scripted
Check for vim in alternatives and use update-alternatives --set if a vim alternative is available (will use the current value, not the "best" alternative). A 4-line script:

VIM=$(update-alternatives --query vim | sed -n '/Value:/ { s/Value:\s$.*$/\1/; p}')
if [ "$VIM " -ne " " ]; then
sudo update-alternatives --set editor $VIM
fi

2. Manual

Use update-alternatives --config editor:

sudo update-alternatives --config editor

This command will display a menu like the one below. Choose the number representing vim in your installation.

There are 5 choices for the alternative editor (providing /usr/bin/editor).

  Selection Path Priority Status
------------------------------------------------------------
* 0 /bin/nano 40 auto mode
  1 /bin/ed -100 manual mode
  2 /bin/nano 40 manual mode
  3 /usr/bin/emacs23 0 manual mode
  4 /usr/bin/vim.basic 30 manual mode
  5 /usr/bin/vim.tiny 10 manual mode

Press enter to keep the current choice[*], or type selection number: 4

3. Without root access

No root privileges? Resort to using the EDITOR environment value. Make sure vim is in your path and put this in your .bashrc:

export EDITOR=vim

A little bit on Dyncon 2011, longish on the utter shame of careless tweeting

2011-03-16T09:37:00.001+01:00

tl;dr - Dyncon 2011 rocked. JRuby is an absolutely awesome piece of software that you should be using.

Saturday was the first day of Dyncon 2011, with great presentations on Erlang, Groovy, Ruby, Testing with Python, Seaside, SproutCore, Clojure and Javascript BDD. Sunday brought more of everything (Ruby and JRuby, Node.js, Cloud9, Erlang, Smalltalk, Groovy, Unhosted, Clojure) and closed with a fantastic session on caching and performance tuning for web sites. I enjoyed it immensely. If you weren't there, you definitely missed out.

The twitter flow was a nice addition to the talks, and I happily contributed my short comments and brief summaries of noteworthy, interesting or funny things said during and after the talks. However, as it happens, some of my tweets could rather be described as "highly shortened versions of other peoples statements taken completely out of context", which indeed does sound much worse than "comments and brief summaries".

One of those careless tweets has annoyingly become one of my most retweeted ever:

@xlson: Groovy is for java devs who want to get things done, JRuby is for ruby devs who want to be on the jvm. #dyncon"

Now, that would all be great and unicorns and sunshine if it wasn't for the small fact that I do not really agree with that statement. I honestly think that all java developers would benefit from at least giving JRuby a try, or as Charles Oliver Nutter so eloquently puts it:

JRuby is capable of great things, and every JVM dev would be remiss if they did not give it a try.

A reasonable question is of course "Why the heck did you tweet something you didn't really mean, stupid?"

Well.

At the time, it seemed to be a perfectly reasonable short summary of Leonard Axelsson's (@xlson) answer to the question "Could you describe the difference between Groovy, JRuby and Mirah?", asked by a member of the audience during Axelsson's presentation on Groovy at Dyncon.

Axelsson actually gave a much better answer than my tweet implies: first, he was very upfront about the fact that he hadn't really used the other languages, and thus didn't really regard himself as qualified to answer this in depth. Then he went on to give a pretty good description of how Mirah differs from the others, following up with a few sentences on his take on the positioning of Groovy and JRuby. I quoted one of these sentences, since I thought it was a catchy phrase that gave a fun description of Groovy. In hindsight, it's easy to see that I managed to convey something that neither I nor @xlson actually meant. Once again in the words of Mr Nutter:

My biggest problem with your statement is that it tells Java people "don't bother with Ruby." They should not ignore Ruby.

True.

Thanks to Charles Nutter for helping me see that I did something stupid, and to Leonard Axelsson for a good presentation on the Groovy Ecosystem.

For those interested, the whole conversation is available on twitter. And here, as an image, for your convenience:

Broken Blackberry

2011-03-10T23:18:00.000+01:00

Tech gone bananas. Sort of. Hehe.

Kindle edition more expensive than print. Wtf?

2011-02-28T08:59:00.000+01:00

I find this weird. Anyone care to explain?

wishes.xml - my settings/modifications

2011-02-24T00:20:00.001+01:00

During my setup of the Apple Magic Trackpad, I modified the wishes.xml from the Ubuntu Multitouch Wiki on Apple Magic Trackpad. Here's a summary of my settings. You can also download my modified file.

Global settings

2 finger tap: right click

3 finger tap: middle click

2 finger drag up/down/left/right: scroll up/down/left/right
3 finger pinch: maximize / restore window

3 finger drag left: switch to next app (Shift + Super + Tab)
3 finger drag right: ringswitcher (Shift + Super + s)
4 finger tap: Exposé (Super + E)
4 finger drag down: window wall, all windows (Super + A)

4 finger drag up: show desktop (Super + D)

4 finger drag left: switch to desktop right of current (Ctrl + Alt + Right)

4 finger drag right: switch to desktop left of current (Ctrl + Alt + Left)

4 finger pinch: zoom in/out (Super + button 4/5 (scroll))

Application settings

firefox + google-chrome
5 finger drag left: back (Alt + Left)
5 finger drag right: forward (Alt + Right)
3 finger drag up: previous tab, scroll through (Ctrl + Shift + Tab)
3 finger drag down: next tab, scroll through (Ctrl + Tab)
2 finger pinch: zoom in/out (Ctrl + KP_Add/KP_Subtract)

inkscape
2 finger pinch: KP_Add/KP_Subtract
2 finger drag: Ctrl + Left/Right/Up/Down

evince (untested)

2 finger pinch: Ctrl + KP_Subtract/KP_Add
2 finger rotate cw: Ctrl + Right
2 finger rotate ccw: Ctrl + Left

okular
2 finger pinch: Zoom in/out (Ctrl + KP_Subtract/KP_Add)

openoffice.org-writer
2 finger pinch: Zoom in/out (Ctrl + button 4/5)

eog (untested)
2 finger pinch: Zoom in/out (Ctrl + KP_Add/KP_Subtract)
2 finger drag left/right: prev/next (Page_Up/Page_Down)
2 finger rotate cw: rotate (Ctrl + R)
2 finger rotate ccw: rotate (Ctrl + Shift + R)

fennec (untested)

2 finger pinch: (Ctrl + Up/Down)

References:

Ubuntu Multitouch Wiki on Apple Magic Trackpad
wishes.xml at Ubuntu Multitouch Wiki on Apple Magic Trackpad

wishes.xml at popqvarnstrom.com

Apple Magic Trackpad on Ubuntu 10.10 x86_64

2011-02-23T16:46:00.004+01:00

Updated 23-Feb-2011: Added reference links, formatting
Updated 24-Feb-2011: Reference to /etc/ginn/wishes.xml instead of /usr/share/ginn/wishes.xml

The Apple Magic Trackpad I recently ordered to use with my docked laptop arrived today. Nice little piece of hardware it is! Since it doesn't work ootb with full multitouch, here's my log of how to get it working on my dev machine.

Pairing
Easy peasy: use the bluetooth applet. Just remember to set the PIN options to "Fixed PIN - 0000". After a successful connection, you should be able to move the mouse pointer around using the trackpad. Multitouch will most probably not work.

Enabling multitouch
The Internet is full of information on how to do this, much of it not very good. The resource that worked (for me) was the Ubuntu Wiki on Apple MagicTrackpad, with a few modifications. Below are instructions based on information from the Ubuntu Wiki on Apple MagicTrackpad, with my modifications.

If pairing went ok, you should now have a file in /usr/share/X11/xorg.conf.d called 60-magictrackpad.conf. If not, log out and then back in again. This configuration file by default assigns the evdev driver for the trackpad. You now have two options: switch to the synaptics driver, which has limited but proven support for multitouch, or install multitouch support for evdev.

Using the synaptics driver
In /usr/share/X11/xorg.conf.d/60-magictrackpad.conf, change the line
   Driver "evdev"
to read
   Driver "synaptics"

The complete file should look similar to this:

Section "InputClass"
   Identifier "Magic Trackpad"
   MatchUSBID "05ac:030e"
   MatchIsTouchpad "on"
   Driver "evdev"
EndSection

Log out and then back in again. Open your mouse settings, and you should have all options available for the trackpad that are available for a synaptics touchpad, including two-finger scroll. You might want to install the "Pointing devices" control panel (sudo apt-get install gpointing-device-settings) unless it is already installed.

Using the evdev driver

You will need to install multitouch support, including Ginn, for multitouch to work with the evdev driver in 10.10:

sudo apt-get install utouch input-utils
sudo apt-add-repository ppa:utouch-team/utouch
sudo apt-get update
sudo apt-get install ginn

Download the Ginn configuration file wishes.xml from the Ubuntu Wiki on Apple MagicTrackpad or one with my modifications from popqvarnstrom.com/files/wishes.xml. Replace ~~/usr/share/ginn/wishes.xml~~ /etc/ginn/wishes.xml with the downloaded version.
Execute ginn. Voila!

(You will probably also want to set up ginn to autostart.)

References:
Ubuntu Wiki on Apple MagicTrackpad
Ginn at Launchpad
Ubuntu Wiki on Multitouch
uTouch at Launchpad

Color printing on Canon iR C2880 from Ubuntu 10.10 x86_64

2011-02-16T01:39:00.000+01:00

Finding and adding the printer was easy as nothing, but no driver matched the printer model exactly. Canon does publish Linux drivers, and a quick download of a .deb from http://software.canon-europe.com/products/0010365.asp later the printer worked as a charm, except for the small matter of interpreting "color" as "greyscale". Not unlike Ford once, the iR seemed to say "You can have your printouts any color you like, as long as it's black".

The driver (CQue 2.0.0 Linux Driver DEB 64-bit (v2.0.0) ) supported setting after setting for stapling and duplex and paper sources, all working perfectly, but no matter how much I configured color printing, papers came out with black toner only. Turning to the Internet for help informed me that yes, this seemed to be a problem with that driver, and that by using the other driver supplied by Canon (UFRII/UFRII LT Printer Driver for Linux v2.20 (2.20)), color would fill my prints. Since the UFRII driver comes as rpm only for 64-bit linux, a few extra steps were needed.

In the interest of making the information more available, I repeat the instructions from http://ubuntuforums.org/archive/index.php/t-809246.html here, slightly modified:

If you only want to print in black & white then it is fairly straight forward.

Black & White printing:
1. Goto System->Administration->Printing
2. Add your printer, either by
   a) browsing the network and adding from there, or by
   b) setting up a new printer queue
   c) Select the type of connection to the printer
   d) Input the printer settings eg: 10.0.0.1/NameOfPrinter and authentication if needed
3. For drivers use Generic->PCL 5c->Generic PCL 5c Printer Foomatic/hpijs
4. Give it a name and you're set to go

If you would also like to print in colour then this is where the fun begins.

You will need the UFR Printer Drivers from Canon. The PostScript drivers are also there but I haven't tried them but I can't see why this guide won't work for them.
(If you're using the iRC 2880 then you can go straight here: http://software.canon-europe.com/software/0040355.asp?model=). This guide should work for most Canon printers and most versions of Ubuntu.

Once you have the drivers and have extracted them then follow these steps.

Color printing for a 32-bit system:
1. If you try and run the deb file in the 32-Bit_Driver/Debian folder in Ubuntu 10.10 it might give a dependency error regarding libcupsys2-gnutls10. If so:
2. Run sudo apt-get install alien
3. Go to the extracted drivers 32-Bit_Driver/RPM directory in a Terminal
4. Run sudo alien -c *.rpm (The -c will include any scripts in the package)
5. Run your newly created deb packages. First the cups common one then the driver one.
6. Follow the intructions above for the setting up of only black & white printing up to and including step 2
7. For drivers use Canon->iR C2880/C3380 UFR II->iR C2880/C3380 UFR II (or whatever model you are using)
8. Give it a name and you should now be printing in colour

Color printing for a 64-bit system:

No deb files are provided.

1. Run sudo apt-get install alien

2. Go to the extracted drivers 64-Bit_Driver/RPM directory in a Terminal

3. Run sudo alien -c *.rpm (The -c will include any scripts in the package)

4. Run your newly created deb packages. First the cups common one then the driver one.

5. Follow the instructions above for the setting up of only black & white printing up to and including step 2

7. For drivers use Canon->iR C2880/C3380 UFR II->iR C2880/C3380 UFR II (or whatever model you are using)

8. Give it a name and you should now be printing in colour

(Oh, all this is for cups)

References:
http://ubuntuforums.org/archive/index.php/t-613979.html, post #5 by Andy
http://ubuntuforums.org/archive/index.php/t-809246.html, post #1 by T0N3

Duokan installation on Kindle DXG

2011-02-14T13:55:00.000+01:00

This was simple enough, thanks to flip and this informative image, telling me how to change language to english. A few things to be prepared for:

'unrar e' will extract files without paths. 'unrar x' is what you want. This is probably a non-issue for anyone either using rar regularly or actually reading the help info or man page.
The 'pre-configured english language rar' displayed all text in chinese until I changed the language setting to English.
The instructions for how to set the language did not match the layout on my screen. For me, it was Home - Menu - Settings (0) - Page 2 in Settings, Letter i, Number 2. This image helped a lot, since it shows the chinese writing representing the menu items.

References: flip's page on duokan installation, image, Duokan, mobileread forum on Duokan

Kindle DXG - first impressions

2011-02-14T13:31:00.000+01:00

Amazon delivered a bit earlier than promised - yay! - and I've enjoyed my Kindle for about a week now. This post is a short summary of my impressions so far.

The good, nah - great

Amazing screen. Looks like paper, but easier to read and nicer to my eyes than actual paper. I've spent several hours a day reading on this thing for a week now. Sunlight? Only improves the reading experience.
Easy to use. The kindle is made for reading, and for this purpose, it works great. I love reading on this thing.
Battery life. I've charged it once, and the battery is still almost fully charged.
Speed. Definitely speedy enough for reading on. Page turns are much faster than I expected with the latest firmware (2.5.7 at the time of writing).
Wikipedia is always accessible. Now, the only thing missing is the label "Don't worry!".

Things that really should be improved

No 3g Internet access in Sweden, except for Wikipedia and the Kindle store. This means no Google Reader, no access to online comics. No blogs. Disappointed.
Only en.wikipedia.org is allowed, all other languages are blocked. This is one of the most annoying restrictions.
No possibility to enable 3g internet access, even if I wanted to pay.
Somewhat quirky zooming in the pdf viewer: most O'Reilly or Pragmatic Bookshelf books fits the screen perfectly at 150% zoom, if the text is centered on the screen. To center the page on the screen is not as easy as it should be, but takes a lot of "nudging".
Limited file type support. Most of my books are PDF or multi-format (mobi/ePub/PDF), which works fine. However, I also read a fair amount of djvu publications. Converting them to PDF typically increases the size 10's of times, and takes a lot of time.

Things I've done to remedy these annoyances

Installed Duokan for DJVU (and possibly improved PDF) support.
Jailbreak and installed USB network support.

Things I plan to do to remedy more annoyances

Crack the device open and switch sim to get 3g internet access.
Nag Amazon about internet access, or at least multi-language wikipedia access.

Stupid Amazon

2011-01-26T23:02:00.001+01:00

So, after much hesitation, especially after the whole Wikileaks debacle, I decided to order a Kindle DX. It seemed reasonable to order a leather cover at the same time, to protect the device. Now, the leather cover had no known delivery date, and the Kindle DX had an estimated delivery in 6-8 weeks. I wanted them delivered together, of course, or if that wasn't possible the DX before the cover. This seems reasonable, right?

Not so according to Amazon, it seems. I now have a leather cover for a gadget that - with a little luck - will show up in March. Why, amazon. Why? I wanted these items shipped together, and I told you so. I have absolutely no use for a kindle cover without a kindle to cover.

Stupid Amazon.

ANTLRWorks and multibyte characters patch

2010-12-14T06:52:00.002+01:00

Update 2010-12-28: patch integrated in antlrworks source at github since 2010-12-15.

If you are using multibyte characters somewhere in your grammar, and wonder why there are weird (null) chars added to the end of your grammars every time you open them, here's a patch: http://popq.pastebin.com/q3d9s9AX

Tested with 1.4.1 and 1.4.2.

Update: new link to updated paste.

Project Euler #34

2010-09-03T03:21:00.007+02:00

Edit 2010-Sep-07: added solution in Haskell
Edit: added link to projecteuler.net problem page
Problem statement:

145 is a curious number, as 1! + 4! + 5! = 1 + 24 + 120 = 145.
Find the sum of all numbers which are equal to the sum of the factorial of their digits.

Note: as 1! = 1 and 2! = 2 are not sums they are not included.

Solution in [j]ruby:
Nothing special here, but I'd really like to do this without the string cast.

class Fixnum
def fact
    if self < 0
      raise "Factorial is only defined for positive integers"
    end
    self == 0 || self == 1 ? 1 : (2..self).reduce(1,:*)
end
end

(10..100000).reduce(0) {|s,x| x.to_s.split("").reduce(0) {|u,v| u + v.to_i.fact} == x ? s + x : s}

Without using string/split:

class Fixnum

  def fact

    if self < 0 

      raise "Factorial is only defined for positive integers"

    end

    self == 0 || self == 1 ? 1 : (2..self).reduce(1,:*)

  end

  def fact_sum

    0.upto(Math.log10(self).floor).reduce(0) do |sum, x|

      sum + (self / 10 ** x % 10).fact

    end

  end

end

(10..1000000).reduce(0) {|s,x| if x.fact_sum == x then s + x else s end}

(Very small speed difference though)

Solution in (noob's) Haskell:

import Data.List

import Char

fact 0 = 1

fact 1 = 1

fact n = foldl' (*) 1 [2..n]

euler34 =
let fact_sum n = sum [ fact $ digitToInt a | a <- show n ]
in foldl' (+) 0 $ filter (\x -> x == fact_sum x) [10..100000]

(In Haskell, as in jruby, this was faster using int->[char] ->int than div/mod loop to find each digit.)

Ref: http://projecteuler.net/index.php?section=problems&id=34

Project Euler #1

2010-09-03T00:40:00.000+02:00

Problem statement:

If we list all the natural numbers below 10 that are multiples of 3 or 5, we get 3, 5, 6 and 9. The sum of these multiples is 23.

Find the sum of all the multiples of 3 or 5 below 1000.

Solution in [j]ruby:
This was simple. Select the multiples of 3 or 5 from the range 3..999 and add them.

(3..999).select {|x| x % 3 == 0 || x % 5 == 0} .reduce(0,:+)

Solution in Haskell:
Since I'm a total noob at Haskell, this is probably a weird solution. Also, I'm still annoyed that things that work perfectly when typed in ghci do not work when sent to ghci from an emacs buffer using haskell-mode. Meh. Anyhow, here is basically the same solution in Haskell:

(sum . filter (\x -> mod x 3 == 0 || mod x 5 == 0)) [3..999]

Ref: http://projecteuler.net/index.php?section=problems&id=1

OSB 11g - Perspective missing in Eclipse

2010-09-02T17:44:00.000+02:00

I have on multiple occasions had Eclipse denying me the Oracle Service Bus perspective for OSB 11g, even though there were no problems during the installation.

Checking forums.oracle.com, I found others in the same situation. I shared my solution there, and here for convenience:

I have found two ways to solve this: either reinstall everything, or follow the script below.

Move the file ECLIPSE_HOME/dropins/oracle.osb.link to a different, safe location.
Edit ECLIPSE_HOME/eclipse.ini and add -clean on its own line right after the -vm <path> entry. Also add -console on its own line, below -clean.
Start Eclipse, and revert the configuration to a point before the OSB Tooling was installed (Help / About Eclipse SDK -> Installation Details -> Installation History -> select a configuration from before OSB Tooling was installed -> Revert). Make sure to leave OEPE installed.
Shut down Eclipse
Move the file oracle.osb.link back to ECLIPSE_HOME/dropins/
Start Eclipse. The Oracle Service Bus perspective might already be working now, so it might be worth checking. If it does, congratulations. Otherwise, continue.
Open the file ECLIPSE_HOME/configuration/org.eclipse.equinox.simpleconfigurator/bundles.info in your favourite text editor.
In the Eclipse [osgi] Console window, type 'ss alsb' (without the quotes), and look for any entries with status INSTALLED.
Choose one of these entries, note the id in the first column and type 'diag <id>' without the quotes, replacing <id> with the actual id for the entry you chose.
In the result, look for any unresolved constraints and pick the last 'Missing required bundle com.bea.alsb.xxx' line (or the most common missing bundle).
Find the matching folder in OSB_HOME/eclipse/plugins, i.e. for the line 'Missing required bundle com.bea.alsb.libs_0.0.0.', find the folder OSB_HOME/eclipse/plugins/com.bea.alsb.libs_1.1.0 (or any number higher than 0.0.0).
Copy the location of that folder, and in the Eclipse Console type 'install file:///<folder path>' (once again, without the quotes). For the example above, with an OSB_HOME path of /opt/OSB11g/Oracle_OSB1 that would be install /opt/OSB11g/Oracle_OSB1/eclipse/plugins/com.bea.alsb.libs_1.1.0/
You should get a bundle id as response. Take note of this bundle id.
In your favourite text editor, modify the file you opened in step 7 (ECLIPSE_HOME/configuration/org.eclipse.equinox.simpleconfigurator/bundles.info) to include a line for the bundle you just installed. It is easiest to copy an existing line and modify that copy. The bundle name will be in the first column, version in the second and the relative path to the actual files (the folder you used in step 12) in the third last column. The entries in the file are sorted by name, and I find it convenient to keep it that way.
Use the bundle id from step 13 to start the bundle: type 'start <bundle id>' at the Eclipse Console prompt (no, still no quotes).
If the response says ok, continue from step 8 above and repeat until all bundles are in any of the states <<LAZY>>, RESOLVED or ACTIVE. If the response is an exception telling you the bundle could not be resolved, repeat from step 11 above for the missing required bundle mentioned in the exception.

When done, save the file ECLIPSE_HOME/configuration/org.eclipse.equinox.simpleconfigurator/bundles.info, restart Eclipse and check that everything is working.

Restore the ECLIPSE_HOME/eclipse.ini file by removing the -clean and -console entries you added earlier.

Spotify links, Firefox 3+ and wine

2010-04-30T03:28:00.002+02:00

Following the instructions on spotify.com isn't quite enough to make spotify links work in Firefox, at least on Ubuntu with FF 3+. you need to add an extra entry in about:config : boolean value network.protocol-handler.expose.spotify, set to false.

In short:
Create an executable script ~/.browser2spotify that does

exec wine "C:\Program Files\Spotify\spotify.exe" /uri "$@"

(of course, modify the path if needed.)

Add the following values in about:config

boolean network.protocol-handler.external.spotify - true
string network.protocol-handler.app.spotify - ~/.browser2spotify
boolean network.protocol-handler.expose.spotify - false

References:
www.spotify.com
Comment by "grus" at http://atheistblogger.com/2008/12/17/using-spotify-links-in-linux/#comments

Unit testing Gtk# stuff

2010-04-24T03:11:00.002+02:00

While experimenting a little bit with C# on Mono, I found myself extending GtkSourceView.SourceBuffer. An example:

  public abstract class DocumentBuffer : SourceBuffer
  {
    public DocumentBuffer() : base(new Gtk.TextTagTable())
    {
    }

    public virtual bool Empty
    {
      get { return StartIter.IsEnd; }
    }
  }

  public class TxtBuffer : DocumentBuffer
  {
    public TxtBuffer() : base()
    {
      HighlightMatchingBrackets = true;
      StyleScheme = SourceStyleSchemeManager.Default.GetScheme("classic");
    }
  }

I created nunit tests for my extensions, compiled, ran the tests, and got nothing in monodevelop (though it claimed to be running tests).

The console, on the other hand, showed

(/usr/lib/monodevelop/bin/mdhost.exe:14191): GLib-GObject-CRITICAL **: /build/buildd/glib2.0-2.24.0/gobject/gtype.c:2706: You forgot to call g_type_init()

(/usr/lib/monodevelop/bin/mdhost.exe:14191): GLib-CRITICAL **: g_once_init_leave: assertion `initialization_value != 0' failed

(/usr/lib/monodevelop/bin/mdhost.exe:14191): GLib-GObject-CRITICAL **: g_object_new: assertion `G_TYPE_IS_OBJECT (object_type)' failed

(/usr/lib/monodevelop/bin/mdhost.exe:14191): GLib-GObject-CRITICAL **: /build/buildd/glib2.0-2.24.0/gobject/gtype.c:2706: You forgot to call g_type_init()

(/usr/lib/monodevelop/bin/mdhost.exe:14191): GLib-CRITICAL **: g_once_init_leave: assertion `initialization_value != 0' failed

Being somewhat new to Gtk and Gtk#, I wasn't sure what I was missing. The app ran just fine, but the tests just hung. The original test looked something like this:

        [Test(), Description("Test for DocumentBuffer.Empty")]
        [Category("DocumentBuffer")]
        public void TestEmpty ()
        {
            DocumentBuffer buf = new TxtBuffer();
            Assert.That(buf.Empty,"New TxtBuffer should be empty.");
            buf.Text = "Example contents";
            Assert.IsFalse(buf.Empty, "TxtBuffer with content should not be empty.");
            buf.Text = "";
            Assert.That(buf.Empty, "TxtBuffer with no contents should be empty.");
            buf.Dispose();
        }

Removing most of it left me with just creating a new instance:

        [Test(), Description("Test for DocumentBuffer.Empty")]
        [Category("DocumentBuffer")]
        public void TestEmpty ()
        {
            DocumentBuffer buf = new TxtBuffer();
        }

No difference. Still hanging in monodevelop, still the same console output.

Time to read, and find out that I need to call Gtk.Application.Init() before executing the tests. Adding the set up code below to my test class fixed it.

       [TestFixtureSetUp]
        public void SetUp() {
            Gtk.Application.Init();
        }

VirtualBox on Ubuntu Lucid beta 2 - segfault / error 4 in libGL.so

2010-04-22T18:16:00.000+02:00

Abstract

Problems: Ubuntu 10.04 beta2 64-bit, problems with VirtualBox, screensaver, desktop effects: segfault, error 4 in libGL.so.1.2 in syslog or dmesg.

Solution: Possibly rebuild the VirtualBox kernel modules using "sudo /etc/init.d/vboxdrv setup". Update the ATI/AMD proprietary FGLRX driver, using the "Hardware Drivers" tool.

Longwinded version

After a somewhat rash dist-upgrade, moving my lenovo T400 laptop from Ubuntu 9.10 64-bit to beta2 of Ubuntu 10.04 (Lucid Lynx) 64-bit, frequent error message pop-ups started to show, mostly concerning gnome-screensav. Segfault. Meh. Since I was now running a beta release, and my concern for whether the screen saver is functional or not is miniscule, these error messages were promptly ignored after being reported.

More annoying, but still easily ignorable, was the fact that I could not enable any desktop effects, like scale and expo. No obvious error messages here, and not worth my time investigating. Old school, no-fx desktop. Works for me.

"Kernel panic" as a substitute for "Resume from Suspend" is pretty annoying, but I sorted that in the "broken screensaver" department and learned how to save often. (Yes, I can be pretty lazy when it comes to my laptop and beta os installations and problems that aren't showstoppers.)

When VirtualBox started acting up, now that was a completely different scenario, attention-wise. Starting VirtualBox went fine, but almost immediately after trying to start any virtual machine, it crashed. Repeated, reproducable crashes. Segfault. Meh. Rebuilding the drivers (/etc/init.d/vboxdrv setup) did nothing. Double meh.

dmesg told me something along the line

VirtualBox[10241]: segfault at 4 ip 00007fbff514612e sp 00007fbfe0ea8ed0 error 4 in libGL.so.1.2[7fbff50f0000+a2000]

This reminded me very much of the screensaver error:

gnome-screensav[13519]: segfault at 4 ip 00007f6c8823012e sp 00007fffdd877a70 error 4 in libGL.so.1.2[7f6c881da000+a2000]

Lucid has some changes regarding default video drivers, but not for my video board (I'm running a proprietary (sadly) ATI/AMD FGLRX driver). A quick check in the Hardware Drivers tool revealed that a different version was in use, than the one preferred by the tool. Updating to the recommended version removed all libGL-related errors, i.e. desktop effects, screen savers and - most importantly - Virtual Box are all back in a functional state. Joy!