Save it into ~/Library/Application Support/Xcode/Color Themes/, restart Xcode, and open its Preferences. You will be able to pick the Railscasts Color Theme in the Fonts & Colors tab.

I dug this theme up after realizing that I haven’t actually saved the Railscasts theme I’d replicated from several months ago and was too lazy to set up another one in my other Mac.

Koen Van Der Auwera did a similar theme too so if you find that mine doesn’t work so well (I’ve only used it for Cocoa Touch development, not at all for developing Mac apps), scoot over to his blog post.

After being a happy Xen user for several years now, I’ve recently had to switch to an alternative virtualization solution. My colleague Arun (@iamclovin) actually struggled for a week with Xen VMs that locked up on Hardy; we’ve had much success with Hardy and Xen before, so we attributed it to a hardware problem since these were our first blade servers.

Out of ideas, we tried Karmic (Ubuntu 9.10) only to discover that Xen support via the apt package system is gone. I went down the path of compiling a paravirt_ops Dom0 kernel (this article was very useful) but ended up deciding the process took far too long despite being successful.

With KVM gaining official support from Ubuntu as the virtualization solution, I ended up ditching Xen and switching to KVM for these new servers on Karmic. The rest of the entry is a step-by-step guide on setting up KVM VMs on a Ubuntu server; I’m putting this down because like all wikis, the Ubuntu KVM wiki has grown a little too organically to be useful.

Preparing a host server for KVM

1. Update and upgrade apt packages (use your own discretion on whether this is necessary):

   aptitude update && aptitude dist-upgrade

2. Check whether CPU supports hardware virtualization:

   egrep '(vmx|svm)' --color=always /proc/cpuinfo

   You should see lines with either “vmx” or “svm” highlighted.

3. Install these packages:

   aptitude install kvm libvirt-bin ubuntu-vm-builder bridge-utils

   If you see a FATAL: Error inserting kvm_intel message during installation, it means that virtualization is not enabled in your machine’s BIOS. You’ll need to reboot your machine, enter the BIOS setup and enable virtualization (you’ll have to hunt for the option).
   
   After enabling virtualization in the BIOS and rebooting, run:

   modprobe kvm-intel

   There should be no error shown (in fact, no console response).

4. Optionally, install virt-top, a top-like tool for your VMs:

   aptitude install virt-top

5. Verify that you can connect to the hypervisor:

   virsh -c qemu:///system list

   You should see something like this:

   Connecting to uri: qemu:///system
    Id Name                 State
    ----------------------------------

6. Setup a network bridge on the server for VMs. Edit /etc/network/interfaces so it looks like this (use your own IPs):

   auto lo
   iface lo inet loopback
   
   auto eth0
   iface eth0 inet manual
   
   auto br0
   iface br0 inet static
    address 192.168.1.222
    netmask 255.255.255.0
    network 192.168.1.0
    broadcast 192.168.1.255
    gateway 192.168.1.167
    bridge_ports eth0
    bridge_stp off
    bridge_fd 9
    bridge_hello 2
    bridge_maxage 12
    bridge_maxwait 0

7. Make sure that you have a direct console to the server because you’re going to restart networking:

   /etc/init.d/networking restart

8. Verify that your changes took place with ifconfig. You should see 2 entries like these:

   br0       Link encap:Ethernet  HWaddr 00:11:22:33:44:55
             inet addr:192.168.1.215  Bcast:192.168.1.255  Mask:255.255.255.0
             inet6 addr: fe80::223:aeff:fefe:1f14/64 Scope:Link
             UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
             RX packets:1099 errors:0 dropped:0 overruns:0 frame:0
             TX packets:50 errors:0 dropped:0 overruns:0 carrier:0
             collisions:0 txqueuelen:0
             RX bytes:74665 (74.6 KB)  TX bytes:6223 (6.2 KB)
   
   eth0      Link encap:Ethernet  HWaddr 00:66:77:88:99:00
             inet6 addr: fe80::223:aeff:fefe:1f14/64 Scope:Link
             UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
             RX packets:4939 errors:0 dropped:0 overruns:0 frame:0
             TX packets:39 errors:0 dropped:0 overruns:0 carrier:0
             collisions:0 txqueuelen:1000
             RX bytes:532798 (532.7 KB)  TX bytes:5585 (5.5 KB)
             Interrupt:36 Memory:da000000-da012800

Setting up a VM

1. Setup a VM with vmbuilder:

   vmbuilder kvm ubuntu \
       -v \
       --suite=karmic \
       --libvirt=qemu:///system \
       --arch=amd64 \
       --cpus=2 \
       --mem=2048 \
       --swapsize=4096 \
       --rootsize=20480 \
       --flavour=server \
       --hostname=billiejean \
       --ip=192.168.1.240 \
       --mask=255.255.255.0 \
       --net=192.168.1.0 \
       --bcast=192.168.1.255 \
       --gw=192.168.1.167 \
       --dns='202.157.163.157 202.157.131.118' \
       --bridge=br0 \
       --mirror=http://archive.ubuntu.com/ubuntu \
       --components='main,universe' \
       --addpkg=openssh-server \
       --user=administrator \
       --pass=icanhaspasswd \
       --dest=/root/vm-billiejean \
       --tmpfs=-

   The options you need to care about are:

   1. suite: Version of Ubuntu to install (e.g. karmic, hardy).
   2. cpus: Number of CPUs to assign to VM.
   3. mem: Amount of RAM in MB to assign to VM.
   4. swapsize: Size of swap in MB of VM.
   5. rootsize: Size of root filesystem in MB of VM.
   6. flavour: The “flavour” of kernel to use in the VM. Either “virtual” or “server”.
   7. hostname: Hostname of VM.
   8. ip: IP address of VM.
   9. mask: Netmask of VM.
   10. net: Network of VM.
   11. bcast: Broadcast address of VM.
   12. gw: Gateway of VM.
   13. dns: DNS server(s) for VM.
   14. addpkg: APT packages to install in the VM. openssh-server is needed so that we can login to the VM to setup the virsh console.
   15. user and pass: User account that’s setup for you to access the VM.
   16. dest: Destination directory on server where VM disk image will reside.
2. If your VM is created successfully, there’ll be a config file for the VM in /etc/libvirt/qemu/ (e.g. /etc/libvirt/qemu/billiejean.xml), and a disk image in the directory specified in the --dest option (e.g. /root/vm-billiejean/disk0.qcow2).
3. You can verify that it works by starting the VM and SSHing into it (virsh console will not work yet).

   virsh start billiejean

Converting Disk Images to LVM Logical Volumes

Now, we have the VM setup but it’s running off a disk image. For better performance, running the VM off a LVM logical volume will optimize disk IO.

vmbuilder is supposed to support the --raw option to write the VM to a block device (such as a LVM logical volume), but I’ve had no success with it (as does Mark Imbriaco, sysadmin of 37signals: http://twitter.com/markimbriaco/status/7437688341 and http://twitter.com/markimbriaco/status/7437699338). We’re going to convert the disk images using qemu-img and write the bits into a LVM logical volume instead.

1. Stop the VM if it’s running:

   virsh shutdown billiejean

2. Convert the VM’s qcow2 (QEMU image format) disk image to a raw disk image:

   qemu-img convert disk0.qcow2 -O raw disk0.raw

3. Create a logical volume to house the VM, making sure it’s big enough for the VM’s rootsize and swapsize options:

   lvcreate -L24GB -n <logical_volume_name> <volume_group_name>

4. Copy raw image into the logical volume:

   dd if=disk0.raw of=/dev/<volume_group_name>/<logical_volume_name> bs=1M

   This will take awhile (the bigger your image, the longer it takes).

5. Edit the VM’s config so that it uses your new logical volume:

   virsh edit billiejean

   Change <disk> to point to the logical volume:

   <disk type='block' device='disk'>
     <source dev='/dev/<volume_group_name>/<logical_volume_name>'/>
     <target dev='hda' bus='ide'/>
   </disk>

6. Startup the VM. You might wanna rename the original disk0.qcow2 image first just to make sure your VM isn’t still using it.
7. Once you’re sure your VM is running off your LVM logical volume, you can delete or backup the original qcow2 disk image.

Getting a console to your VM from the Host Server

Now, we have to setup the VM so that virsh console works. This is a console to the VM from the host server that works even when the networking in the VM is not.

1. Edit the VM’s settings:

   virsh edit billiejean

   In the <devices> block, add:

   <serial type='pty'>
     <target port='0'/>
   </serial>

2. Startup your VM:

   virsh start billiejean

3. SSH into VM and create a file /etc/init/ttyS0.conf:

   start on stopped rc RUNLEVEL=[2345]
   stop on runlevel [!2345]
   
   respawn
   exec /sbin/getty -8 38400 ttyS0 vt102

   Start the tty with:

   start ttyS0

4. Still in the VM, install acpid so that the VM will respond to shutdown commands from the server:

   aptitude install acpid

5. Reboot the VM.
6. Verify the console works by opening a console to the VM from the server:

   virsh console billiejean

   You may have to hit “Enter” before you see any console output.

Miscellaneous VM Setup

That’s it, your VM is ready! You’ll probably want to do these:

• Set a root password and possibly delete the user you setup using vmbuilder.
• Set the timezone with dpkg-reconfigure tzdata.

While a page is loading, the button acts as a Stop button:

When it’s done loading, it becomes a Reload button.

I’m always up for 1 less unnecessary button. I think this feature is copied from Safari (I might be wrong).

As an example, here’s the Youtube webpage of the owner of Maru the Cat:

To open the thumbnail image of Maru in a new tab, right-click the image, and then middle-click the View Image item from the context menu:

It should open up in a new tab:

You can also do the same for background images (middle-click View Background Image from the right-click context menu).

Anyway, I submitted it to the iPhone App Store for approval as a free app and named it Singtel Data Usage. No points for creativity with the name but I thought it’d be nice to know exactly what the app did from just reading its name. After one keyword rejection (which was unfounded but I wasn’t going to argue with the app reviewers), it finally got approved a few days ago.

Yeah that’s what it looks like. I stole part of the design from another iPhone app – can anyone guess which one? I think I spent 80% of development time changing the design over and over again, tweaking font sizes, and agonizing over which icons to use.

Thanks to folks on Twitter (I’m @chuyeow – do follow me if you’re reading my blog), word got around and by the 3rd day of it being approved, my app was the Top Free app on the Singapore App Store. Going to take a time out for the vanity shot:

A similar app, iBBOM, was approved around the same time but I believe I got a head start because of all the awesome people on Twitter who retweeted my tweet and the kind folks who gave glowing reviews on the App Store (thanks especially to Andy Croll for the 1st review). iBBOM actually looks much better than my app (I can’t help tweeting about it) and works perfectly (only thing I don’t like are the ads). I believe I couldn’t have gotten to the top of the App Store without the guys and gals who spread the word on Twitter.

On to the not so fun part of this post. Yet another similar app, SG BBOM, was pulled from the App Store by Singtel (the telco) recently so I’m expecting to hear from Singtel really soon. We shall see.

In the meantime, if you own an iPhone and have a Singtel data plan, do check out my app (App Store link) and give an honest review on the App Store. If you’ve already downloaded it, I’d like to know what you think in a review as well. Thanks!

Here’s a before and after shot:

I think it’s much better, definitely more modern and minimalist.

The new blog theme is basically a heavily modified version of the
wonderful and free Compositio WordPress theme –
you can check out a demo of Compositio here to see how different it is.

The only thing I kept from my old theme, which I affectionately called “Clover”, was the same clover logo. I did that with the shape tool in Photoshop – it’s a simple logo and no skill was required to create it, obviously!

If you’re reading this in a feed reader or one of those aggreggator-type sites, you can help me out by checking out the blog itself and giving me some good old constructive criticism.

I attributed it to the recent jailbreak (I’ve already removed the jailbreak trying to figure out this problem) at first. I only realized something was really wrong when the iPhone’s battery started draining while it was charging via USB!

Looking around on teh interwebs, I managed to solve it (easily!) by following the suggestions from this thread at the Apple support forums. Here’s what worked for me:

1. Go to Settings on your iPhone.
2. Turn Notifications off. I’m not sure if this is necessary but did so just in case.
3. Go to Mail, Contacts, Calendars and delete all your accounts (I deleted even Calendar accounts just in case). You may want to write down or backup your account settings first.
4. Go to Fetch New Data (also in the same Mail, Contacts, Calendars settings) and make sure Push is Off.
5. It’s probably not necessary but you may want to restart your iPhone (power it on and off) here.
6. Add back all your accounts. You can turn Push back on Fetch New Data now.
7. You can also turn Notifications back on if you wish.

It seems like the iPhone 3.0 OS update has a bug where email accounts with active push notifications were set to keep fetching new email even if push is turned off explicitly. Hopefully this post will help solve your battery problems, if not, you’ve gotta keep looking – the Apple iPhone Support forums is a good place to start.

It probably sounds like shameless self-promotion that I’m blogging about these changes since I’d contributed them. Well, yeah that’s true, but I’m also doing so because I doubt these changes will ever be made known since Curb doesn’t publicize any updates or changelogs.

Curb is my current number 1 HTTP client for Ruby so the more love it gets, the better.

Passing cookies as a string in Curb requests

Curl veterans will probably know how to do this with the curl binary:

curl -b "auth=abcdef; ASP.NET_SessionId=lotsatext;" example.com

This sends a request to example.com with 2 cookies named “auth” and “ASP.NET_SessionId” (I hate those big-ass ASP.NET cookies btw). There wasn’t a way to set this in Curb, so I looked up the libcurl C API docs and replicated the same option in Curb (commit on Github). An example:

curl = Curl::Easy.new('http://example.com/')
curl.cookies = 'auth=abcdef; ASP.NET_SessionId=big-wall-of-text;'
curl.perform

Of course, the cookies will more often be retrieved/constructed rather than a literal like in the example above. In my case, I was proxying cookies while trying to wrap an API around a site that doesn’t have one.

Passing cookies as a file via the “cookiefile” option

The second change is the new cookiefile option. This replicates curl commands like these:

curl -b cookies-to-send.txt www.example.com

with something like this in ruby:

curl = Curl::Easy.new('http://example.com/')
curl.cookiefile = '/path/to/cookies-to-send.txt'
curl.perform

The cookies file looks like this (you can get a sample with the --cookie-jar option to curl, e.g. curl --cookie-jar cookies.txt www.wego.com):

www.wego.com	FALSE	/	FALSE	0	lang
www.wego.com	FALSE	/	FALSE	0	user_country_code	SG

Check out the commit on Github if you’re interested.

Why would I use these?

Now you might be wondering how these 2 changes are useful – well, they are totally irrelevant to you if you’re not expecting any cookie support in Curb. However, if you’re accessing or scraping a site that uses cookie-based authentication, these changes allow you to keep your Curb client authenticated across sessions, even when doing HTTP POSTs (Curb doesn’t send cookies properly in POST requests even if curl.enable_cookies is set).

I’ve found these changes to Curb particularly useful since I vastly prefer Curb to most HTTP clients for its speed and lightweight implementation, and heavier scraper-type HTTP clients like Mechanize are a last resort.

So anyway, I am one of those laggards and only wrote a Rails Metal piece not too long ago for a Rails app on Wego.com in an effort to optimize some high volume requests. It looked something like this:

class Adamantine
  def self.call(env)
    if env['PATH_INFO'] =~ ROUTE_REGEX
      location = Location.find(1) # Use ActiveRecord.

      [200, { 'Content-Type' => 'text/html' }, [location.to_json]]
    else
      # Leave it to Rails to deal with the request.
      [404, { 'Content-Type' => 'text/html' }, ['Not Found']]
    end
  end
end

Notice the use of an ActiveRecord model. It ran for quite awhile, almost 30 minutes, in production until I started getting notification emails about “Too many open database connections” to the MySQL server! The change was promptly rolled back and there was no cheezburger for me.

As it turns out, Rails doesn’t take care of certain things in Rails Metal pieces, including the releasing of connections back to the database connection pool. A bit of googling turned up this bug in Rails’ bug tracker (see Josh Peek’s comment about ActiveRecord::Base.clear_active_connections!).

I rewrote the Rails Metal piece to always ensure it clears any active database connections with ActiveRecord::Base.clear_active_connections!:

class Adamantine
  def self.call(env)
    if env['PATH_INFO'] =~ ROUTE_REGEX
      location = Location.find(1)

      [200, { 'Content-Type' => 'text/html' }, [location.to_json]]
    else
      # Leave it to Rails to deal with the request.
      [404, { 'Content-Type' => 'text/html' }, ['Not Found']]
    end
  ensure
    # Release the connections back to the pool.
    ActiveRecord::Base.clear_active_connections!
  end
end

Moral of the story: Don’t forget to release your database connections if you’re using ActiveRecord in your Rails Metal. Or even better, don’t use ActiveRecord in Rails Metal – you’re aiming for raw speed anyway right?

If you belong to what I imagine must be now the microscopic population of loyal readers of my blog, please do hook up with me on FriendFeed or Twitter. I’ll follow you back if you’re a reader!

This tab tearing capability is a pretty neat feature – I know you can already do this in Safari, Opera and Galeon. It’s really well done in Safari, which I think is what Firefox is emulating. Nice to see Firefox follow suit!

If you can’t wait for Firefox 3.1, try it out in a recent Firefox nighty build (remember to use a new profile unless you are willing to risk corrupting your daily profile).

Here’s what you can do:

1. Go to about:config in Firefox.
2. Type in “session” in the “Filter” box.
3. Edit browser.sessionhistory.max_entries – this is the number of pages stored in the history of your browsing session. Basically these are pages that can be reached using your Back and Forward buttons. The default is 50 – I reduced it to 20.
4. Edit browser.sessionhistory.max_total_viewers – this is the number of pages that are stored in RAM so that they aren’t re-processed by Firefox’s rendering engine. This is what allows you to go Back to a page in Firefox and have it load almost instantaneously. The number of pages stored actually depends on the amount of RAM on your machine (see this). I reduced this to 4 (I have 2GB RAM).
5. Edit browser.sessionstore.max_tabs_undo – the number of tabs you can restore after closing them (you can do this with Ctrl/Cmd-Shift-T). The default of 10 is more than I really need, so I reduced it to 3 tabs.
6. Edit browser.sessionstore.interval – Firefox saves your session after every 10 seconds by default. I changed this to a more conservative 30000 milliseconds.

You can read more about these preferences and more at the MozillaZine Knowledge Base. If you’ve any tips on how to improve Firefox’s performance, be sure to share!

But we were attached to our ugly shade of green to leave those pages in their naked glory. Preliminary CSI work told me that some cached stylesheets generated by Rails were empty files. Why is this is happening?

stylesheet_link_tag and the :cache option

Was I overriding the stylesheets generated by Rails in different pages? Because we have a lot of cobranded sites and country sites on wego.com, I use the :cache option when using stylesheet_link_tag very often.

For example, the main wego.com site’s layout template has a stylesheet_link_tag like this (in reality there are a whole lot more stylesheets):

<%= stylesheet_link_tag 'yui/reset-fonts', 'search',  :cache => 'cache/search/listings' %>

When I need to make a new page for a cobranded site, I’ll create a new layout template with this:

<%= stylesheet_link_tag 'yui/reset-fonts', 'search', "sites/#{current_site}/cobrand", :cache => "cache/#{current_site}/search/listings" %>

Oftentimes I’d copy and paste (boo and hiss all you want!) the stylesheet_link_tag from one layout template to another and forget to update the cache path (the :cache => '/path/to/stylesheet' part). Two different stylesheet sets being cached to the same path is naturally a very stupid thing to do. So this wasn’t it, but it’s good to point this out because I have made this mistake at least 2 times!

Don’t check in generated CSS files by accident

Next, since I was using Sass, I was by now pretty sure that was it. First things first: did I check in a generated CSS file into source control (we use Git)? It’s another amateur mistake, but unsurprisingly, I’ve done this a couple of times. I think I’d wasted about an hour hunting down the reason for a style change that just wouldn’t show up. Yeah, I could have just added *.css to .gitignore, but I’m still using a mix of pure CSS and Sass templates.

The problem

In the end, I found this blog post by Ari Lerner on the CitrusByte blog about similar woes with Sass in production that set me on the path to a solution. It seems that when Rails encounters stylesheet_link_tag calls, it starts to pull together all the stylesheets and sometimes Sass is unable to generate the CSS files fast enough. Rails then throws an exception about not being able to find the CSS files and outputs an empty CSS file to the cache path.

The solution

The solution? Generate all the CSS files from Sass templates prior to restarting Rails when deploying. I added a rake task for updating all the Sass stylesheets:

namespace :sass do
  desc 'Updates stylesheets if necessary from their Sass templates.'
  task :update => :environment do
    Sass::Plugin.update_stylesheets
  end
end

Then, I created a mirror of this as a Capistrano task:

namespace :sass do
  desc 'Updates the stylesheets generated by Sass'
  task :update, :roles => :app do
    invoke_command "cd #{latest_release}; RAILS_ENV=#{rails_env} rake sass:update"
  end

  # Generate all the stylesheets manually (from their Sass templates) before each restart.
  before 'deploy:restart', 'sass:update'
end

Now, whenever I do a cap deploy, the stylesheets are generated before the Rails processes are restarted, ensuring that Rails’ stylesheet_link_tag helper is always able to find the pure CSS files when trying to merge them together and caching them to a single file.

Anyway, if you hold down the Cmd key (aka the Apple key), a menu pops up that shows the folder hierarchy of where the currently opened QuickTime movie is in your filesystem:

This works in all Mac apps that display the filename of the currently open/focused file in the titlebar.

It’s useful for me since my NADD means I try to close as many unused windows as possible to adhere to my Cmd-Tab diet – now I can close Finder windows after opening files and be sure that I can get back to them quickly. What about QuickSilver? Yup, I do use (and love) QuickSilver but I don’t let it catalog every single file!

Another nice thing about this is that I can easily reveal files in Finder in my favorite text editor (TextMate) without needing to use the project drawer:

Catch the new first edition – this one’s about the API changes since Rails 2.1.