procedural

Running Tests Against a Gem

2008-05-30T10:23:55Z

I use Hoe to build my projects. Sometimes I add a new source file and forget to update my Manifest.txt, so my gem doesn’t contain all of the necessary files. My tests pass, but then I install my gem and it doesn’t work, usually due to a failed require.

So I thought: it would be nice if I could run my tests against the gem, so I could know that it was built correctly.

Here’s how I do it:

def recreate_dir(dir)
  FileUtils.rm_r(dir) if File.exist?(dir)
  FileUtils.mkdir_p(dir)
end

def error(msg)
  $stderr.puts "ERROR -- #{msg}"
  exit 1
end

def run(command)
  puts command
  error 'command failed' unless system(command)
end

namespace :my do

  # initializes test output dir
  task :setup do
    # typical setup stuff
  end

  Spec::Rake::SpecTask.new(:run_system_specs) do |t|
    # typical RSpec stuff
  end

  desc "Runs RSpec System Tests against gem"
  task :system_specs_gem => ['my:setup'] do
    # builds gem
    Rake::Task[:gem].invoke

    # my code prettifier doesn't like this line.
    # uncomment if you want this code to actually work.
    #gem = Dir.glob("pkg/*.gem").first

    # installs gem into tmp dir
    install_dir = 'tmp'
    recreate_dir(install_dir)
    run "gem install --install-dir=#{install_dir} --no-rdoc --no-ri -y #{gem}"

    # points tests to gem and runs
    ENV['GEM_PATH'] = install_dir
    Rake::Task['my:run_system_specs'].invoke
  end

end

Notes:

Hoe gives me the :gem task for free
I put my tasks in the :my namespace so I can easily differentiate them from the ones provided by Hoe
I don’t run my tests in this fashion as I’m writing code, but I try to do it before I commit. My automated build runs the tests against the gem before it installs it on my local system.

Trac Broke in Debian After Python 2.5 Became the Default

2008-05-05T01:53:35Z

My Trac is broken

A few days ago I tried to log into my local Trac instance but saw “Internal Server Error”.

Looked in /var/log/apache2/error.log and saw:

[error] get_interpreter: no interpreter callback found.
[error] [client 192.168.1.100] python_handler:
  Can't get/create interpreter.
[error] make_obcallback: could not import mod_python.apache.\n
  ImportError: No module named mod_python.apache
[error] make_obcallback: Python path being used
"['/usr/lib/python2.4/site-packages/setuptools-0.6c3-py2.4.egg',
'/usr/lib/python2.4/site-packages/buildutils-0.1.1-py2.4.egg',
'/usr/lib/python2.4/site-packages/4Suite_XML-1.0.2-py2.4-linux-x86_64.egg',
'/usr/lib/python24.zip',
'/usr/lib/python2.4',
'/usr/lib/python2.4/plat-linux2',
'/usr/lib/python2.4/lib-tk',
'/usr/lib/python2.4/lib-dynload',
'/usr/local/lib/python2.4/site-packages',
'/usr/lib/python2.4/site-packages',
'/usr/lib/python2.4/site-packages/Numeric',
'/usr/lib/python2.4/site-packages/PIL',
'/usr/lib/python2.4/site-packages/gst-0.10',
'/var/lib/python-support/python2.4',
'/usr/lib/python2.4/site-packages/gtk-2.0',
'/var/lib/python-support/python2.4/gtk-2.0',
'/usr/lib/python2.4/site-packages/wx-2.6-gtk2-unicode']".

Ran dpkg -L libapache-mod-python and noticed that some stuff was installed in /usr/lib/python2.5/site-packages. But the log implies that Trac is using Python 2.4?

Wait, which is Debian’s default Python version?

->matth@juggernaut ~ ? type python
python is /usr/bin/python
->matth@juggernaut ~ ? ll /usr/bin/python
lrwxrwxrwx 1 root root 9 2008-05-02 06:35 /usr/bin/python -> python2.5*

Looks like the default switched to 2.5 since that morning’s apt-get upgrade.

Getting Trac running with Python 2.5

I restarted Apache, but then it complained because ClearSilver (Trac’s templating engine) wasn’t installed for Python 2.5.

easy_install ClearSilver failed (”couldn’t find setup file”). Download, configure, make, make install also failed, complained about a C# test failing.

I then read that you can easy_install Trac. I think I would prefer to manage it this way. I apt-get removed it, then easy_installed it.

Trac ran, but complained about not having permissions to write to /var/www/.python-eggs. I created this directory and blessed the permissions.

Trac ran! But my custom header wasn’t there.

The Trac 0.11 Release Notes are informative. They switched their templating engine from ClearSilver to Genshi.

Porting my templates from Clearsilver to Genshi

PortingFromClearSilverToGenshi helps.

I created site.html in $trac_root/templates. ($trac_root is nothing required by Trac, it’s just where I keep my custom Trac scripts and similar things.)

<html xmlns="http://www.w3.org/1999/xhtml"
  xmlns:py="http://genshi.edgewall.org/"
  py:strip="">

  <head py:match="head" py:attrs="select('@*')">
    ${select('*|comment()|text()')}
    <link rel="stylesheet" type="text/css"
      href="${href.chrome('site/style.css')}" />
  </head>

  <body py:match="body" py:attrs="select('@*')">
    <div id="siteheader" style="margin:1em 0 1em 0;">
      <a href="/">trac root</a>
      /
      <a href="/$project.name">$project.name</a>
      /
      <a href="/$project.name/report/3">active tickets by milestone</a>
    </div>

    ${select('*|text()')}

  </body>
</html>

Then, for each of my Trac project directories, I added this to $project/conf/trac.ini:

[inherit]
templates_dir = /x/system/trac/templates/

(my $trac_root = /x/system/trac)

Good night, and good luck.

Generating /etc/motd from ‘fortune’ at boot

2008-04-14T10:46:00Z

I created a script fortune2motd:

file=/etc/motd

# the wrapping echos are for spacing
echo > $file
/usr/games/fortune -a >> $file
echo >> $file

Then I created/etc/init.d/fortune2motd:

#! /bin/sh

N=/etc/init.d/fortune2motd

case "$1" in
  start)
        /home/system/private/bin/fortune2motd
        ;;
  stop|reload|restart|force-reload)
        echo "got 'em"
        ;;
  *)
        echo "Usage: $N {start|stop|restart|force-reload}" >&2
        exit 1
        ;;
esac

Then I configured it to run at boot using Debian’s update-rc.d command. I don’t remember the exact syntax, but I think it’s something like:

update-rc.d fortune2motd defaults 99

And now I get cute messages whenever I login to my box remotely.

Differentiating JRuby executables from CRuby

2008-04-07T22:17:15Z

Now that I’ve installed JRuby alongside CRuby, I have a hard time knowing which gem or rake I’m running.

I fix this by running this script from my .bashrc:

export JRUBY_HOME=/usr/local/jruby
export PATH=$PATH:$JRUBY_HOME/bin

if [ -d $JRUBY_HOME ]; then
  for file in $JRUBY_HOME/bin/*; do
    alias jruby_`basename $file`=$file
  done
else
  echo "$JRUBY_HOME not found"
fi

Then I can run jruby_gem or jruby_rake and know it’s legitimate.

Logging Memory Usage in Java

2008-04-07T00:42:46Z

Sometimes our system gets slow, and we want to investigate what’s going on. Checking the memory usage is one of the first steps in this process.

I wrote this command to help. It uses the following classes from java.lang.management: ManagementFactory, MemoryMXBean, and MemoryUsage.

@Command(path = CMD_PREFIX + "/memory")
public void mbeanMemory(CommandInvocation ci) {
  final MemoryMXBean memoryBean =
    ManagementFactory.getMemoryMXBean();

  final String date =
    new SimpleDateFormat("yyyy/MM/dd hh:mm:ss")
    .format(new Date());

  ci.out.println(String.format("[%s] heap:{%s}, non-heap:{%s}",
    date,
    describe(memoryBean.getHeapMemoryUsage()),
    describe(memoryBean.getNonHeapMemoryUsage())));
}

private static final String describe(MemoryUsage memory) {
  final long max = memory.getMax();
  final long used = memory.getUsed();
  final long free = max - used;

  return String.format("used:%sMB, free:%sMB, max:%sMB",
    bytesToMegs(used), bytesToMegs(free), bytesToMegs(max));
}

private static final long bytesToMegs(long bytes) {
  return bytes / (1024 * 1024);
}

The @Command annotation is a part of an OSGi Commands system created by one of my colleagues.