What 'OK' button?

Netbooks are great. I love the portability, and the fact that it's possible to make devices that small now-a-days, but they still have sufficient power for typical daily work. One of the trade-offs for that smaller size though, is a smaller screen size. Typical resolution for many of these newer netbooks is 1024x576.

Unfortunately, most developers aren't running at anything near that small anymore. The side effect of this is that devices that small often go forgotten, or untested at least. I saw a few of these recently when digging through the preferences dialogs in an application and found one that was not resizable, and pushed straight off the bottom of the screen. Fortunately, this isn't too hard to work around. If you ever run into this, simply hold down alt, and click in the windows somewhere to drag it around to a better position so that you can see the buttons. But more importantly - REPORT THE BUG! And if you're a developer, please think about things like this when designing user interfaces.

I'd like to focus on the test aspect of this for a moment though. I'm working on a testsuite for testing these sorts of things, but the way it works for now requires that you actually be running at the maximum resolution you support. Problem is, I don't have a netbook at the moment that defaults to 1024x576.

xrandr to the rescue
Here's a simple way to do this, and it's even working great on my UNR installation running under virtualbox. You'll need to have the VirtualBox guest extensions installed on the VM for this to work properly.

Use cvt to get the modeline
$ cvt 1024 576
Modeline "1024x576_60.00" 46.50 1024 1064 1160 1296 576 579 584 599 -hsync +vsync

Next use xrandr to add the new mode
$ xrandr --newmode "1024x576_60.00" 46.50 1024 1064 1160 1296 576 579 584 599 -hsync +vsync
$ xrandr --addmode VBOX1 "1024x576_60.00"
$ xrandr -s 1024x576_60.00

And that's it! At this point, you should see the screen resolution change. One thing to note, if you are not running under virtualbox, replace the VBOX1 with LVDS on the addmode line. If that still doesn't work, run xrandr by itself and look for the line that says "connected" to find the output name to use.

The more things change, the more they stay the same

So I'm on to new stomping grounds at Canonical now, working on the Mobile team. One of the things we work on is Ubuntu Netbook Remix. If you have a netbook, check it out. You can even run it straight off a USB stick to try it out first.

The other major thing we're working on right now is the ARM port of Ubuntu. Lots of cool stuff going on here too. Jaunty went out with support for iMX51 boards, and even more great ARM stuff coming for the next release. Stay tuned, lots of fun stuff coming with this.

SystemTap Redpaper

I participated in an IBM Redbook assignment recently with a small team (3 of us). The paper is called: SystemTap: Instrumenting the Linux Kernel for Analyzing Performance and Functional Problems. If you are interested, check it out at: http://www.redbooks.ibm.com/abstracts/redp4469.html. I was really impressed with the Redbook process and this was a very enjoyable experience for me to work with two guys from Brazil on this. As with anything, it always seems like time is too short and so much more I wanted to do, but I guess that's just my personality.

Migrating Thunderbird/Lightning when reinstalling

I recently migrated one of my systems from FC9 to Ubuntu 8.04 (Hardy) and ran into some minor annoyances with Mozilla Thunderbird and the Lightning extension. I eventually found the answer after some digging, but thought I might post it in case anyone else runs into this. Or more likely, so that I can remind myself next time I hit this problem! :)

It seemed simple enough, I did an apt-get install thunderbird and left my old .thunderbird directory in place from the FC9 install. When I brought up Thunderbird though, it was as if it was coming up for the first time.

Problem 1 - Thunderbird did not see the old .thunderbird directory
Fix - This was the easy one to spot, apparently the Thunderbird under Ubuntu looks for a .mozilla-thunderbird directory. A quick rename fixed the problem

Next though, I noticed that my calendar didn't work AT ALL. All I saw was the shell of what it would should be, no days were even displayed, no calendars, not even an option to create a new calendar. The calendar info is stored under your profile dir in the .mozilla-thunderbird dir, in a file called storage.sdb. I renamed my storage.sdb file to preserve it, removed the addon for lightning, and reinstalled the addons from the latest .9 version of lightning from http://addons.mozilla.org. After restarting Thunderbird though, I still didn't see a calendar. I didn't want to use the version that shipped with Ubuntu since it was version .7, but it looks like that one works fine. So if you don't care to be using the latest version, then this is a viable workaround, but not the fix I was looking for. As it turns out, the lightning extension I downloaded relies on an older version of libstdc++.

Problem 2 - Empty and non-functional lightning .9
Fix 2:

1. apt-get install libstdc++5
2. Uninstall the lightning addon
3. reinstall the lightning addon downloaded from http://addons.mozilla.org
4. replace storage.sdb file with your old, saved copy

It all works great now.

A couple of fun things

Here are a couple of fun things I found recently that I thought were interesting:

Gist - Most people have probably heard of github, but Gist seems to be a new service they are offering that will let you post code snipits, kind of like pastebin. But with Gist, you can share, make changes, fork, add files, etc. via git! To actually modify what you put there via git, you'll have to have a free account, but sharing code and pasting via the web interface requires nothing at all.

Ohloh - I have to admit, that I just don't get social networking yet. I can't seem to bring myself to really get into all the facebook and myspace and all that. I'm probably just getting old. This is one site that actually makes me think there is hope for the concept though, as it actually provides interesting metrics and information about open source projects. It found contributions I had made from years ago, and let me tie them to my account.

Fault injection with Systemtap

A while back, I started learning and using Systemtap to help in understanding code flow, and in debugging. It's a really neat tool, but looking at it from the perspective of a tester, I immediately realized that it was going to be useful for fault injection. After some searching, the only thing I could find was a reference to the SCSI fault injection test harness. This was a good start, but obviously the same concept would be useful to apply to other areas of the kernel as well. Today, I noticed that someone else has finally started along that path.

It's fairly new, but the authors claim to have ported the SCSI fault injection test harness to their new fault injection framework. It is small, easy to understand, and they have provided some examples already. It also looks like it will have a home in the Systemtap tree fairly soon as well. This is one of those things I wish I had about 7 years ago!