Sevenforge by Curtis Spencer

Using Non-Standard Library Locations with cmake

2009-10-28T00:00:00+00:00

I am currently running some builds on a shared host where I don't have root access. Unfortunately, this means that most of my 3rd party libraries are installed in ~/local instead of in standard locations like /usr/ and /usr/local

In my CMakeLists.txt, I have commands such as FIND_LIBRARY and FIND_PATH. I didn't want to change those to use the machine dependent paths, but rather control them with environment variables.

So in order to get everything working smoothly, it proved to be pretty easy.

I just added the following to my .bashrc

export CMAKE_PREFIX_PATH=~/local/
export LD_LIBRARY_PATH=~/local:$LD_LIBRARY_PATH
export LD_RUN_PATH=~/local:$LD_LIBRARY_PATH

Now cmake can find the nonstandard libraries and when I execute the program, the libraries are dynamically loaded properly.

Stripping a Bad Revision from Mercurial

2009-10-21T00:00:00+00:00

Sometimes you do a hg addremove; hg ci; hg push; when you aren't really thinking. Here is the antidote.

$hg --config extensions.hgext.mq= strip bad_revision_num

This hg command will strip the bad_revision_num and every consequent revision. If other developers have started pushing onto your bad repository, it is an even bigger headache to strip that bad revision and keep the future ones. I haven't ran into that problem just yet, but I imagine a little work with turning the revisions into patches and applying the cherry picked patches would be one option.

Adding rxvt-unicode to termcap on OSX

2009-08-31T00:00:00+00:00

I recently installed rxvt-unicode using MacPorts on OSX. However, when I was doing some terminal related tasks such as using screen I was getting errors such as "Missing termcap entry for rxvt-unicode". After digging through the rxvt-unicode documentation, the following bash command install the appropriate termcap entry.

$ infocmp rxvt-unicode > /tmp/termcap
$ tic /tmp/termcap

The Pain of Connecting to WPA Wi-Fi on the Linux Command Line

2009-07-28T00:00:00+00:00

It is sometimes useful to be able to connect to a Wi-Fi access point without a heavyweight GUI, especially when installing packages before you have a desktop environment installed or getting a lonely home server on the grid without powerline networking. I spent a while culling all the best information from various blog posts and howtos about connecting to wireless access points from the command line. I think I have come across a good mix. Once you have a GUI, it is quite a bit easier to just use wicd or Gnome's Network Manager.

I did this on Ubuntu Jaunty, but it should work on any distro with wpa_supplicant.

First you want to figure out your network hardware configuration

$ sudo lshw -C network

For example, eth1 showed up as my wireless network card's logical address. So for the future examples, eth1 is synonymous with your wireless network interface.

Next, you can get a list of the local access points using the following command:

$ sudo iwlist eth1 scan

One example result is:

eth1 Scan completed :
Cell 01 - Address: 00:19:E3:FA:38:95
ESSID:"MyEssid"
Mode:Managed
Frequency:2.462 GHz (Channel 11)
Quality:4/5  Signal level:-63 dBm  Noise level:-92 dBm
IE: IEEE 802.11i/WPA2 Version 1
    Group Cipher : TKIP
    Pairwise Ciphers (2) : CCMP TKIP
    Authentication Suites (1) : PSK
IE: WPA Version 1
    Group Cipher : TKIP
    Pairwise Ciphers (1) : TKIP
    Authentication Suites (1) : PSK
Encryption key:on
Bit Rates:1 Mb/s; 2 Mb/s; 5.5 Mb/s; 6 Mb/s; 9 Mb/s
          11 Mb/s; 12 Mb/s; 18 Mb/s; 24 Mb/s; 36 Mb/s
          48 Mb/s; 54 Mb/s

Now, we need to generate a wpa_supplicant configuration file based on the WPA access point we want to use

$ wpa_passphrase "MyEssid"
mypassword

Take the resulting output and put it into /etc/wpa_supplicant.conf The file should look something like this

# reading passphrase from stdin
network={
  ssid="MySSID"
  #psk="mypassword"
  psk=949744ef68a32f47239b005309823712728ea8238928f6649e2a1be9c43cab
  proto=WPA
  key_mgmt=WPA-PSK
}

Now, you need to start running wpa_supplicant using that configuration file as input

$ sudo wpa_supplicant -D wext -c /etc/wpa_supplicant.conf -i eth1

Once wpa_supplicant is running you need to run dhclient to get an ip address from the access point.

$ sudo dhclient eth1

Now, if you want to save this so you don't have to do it again, put the following into your /etc/network/interfaces file:

auto eth1
  iface eth1 inet static
  address 192.168.0.100
  netmask 255.255.255.0
  wireless-essid MyEssid
  pre-up wpa_supplicant -Bw -Dwext -ieth1 -c/etc/wpa_supplicant.conf
  post-down killall -q wpa_supplicant

This is something that is a lot more complex than it should be... In writing it down, I hope that the next time I need it, I should be able to reproduce it

Meet Pillowtalk, an ANSI C API for CouchDB

2009-07-13T00:00:00+00:00

For those that don't know, CouchDB is an exciting document oriented database that is built upon Erlang. It uses JSON as its data format and HTTP to communicate with clients. These two features are especially exciting because they create a platform agnostic way to pass around data and a horizontally scalable communication mechanism.

Enter Pillowtalk

We have been doing some experimenting with CouchDB over at Cruxlux, but a lot of our stack is in C. Unfortunately, there wasn't a simple way to interact with CouchDB because the C API was deprecated a while back.

Pillowtalk is a C library built upon libcurl for the HTTP communication and yajl for parsing and generating JSON. I have tagged a 0.0.1 release for anyone who wants to kick the tires over at github. Learn more about the library with an example.

Add Duration to FLV files using flvtool2

2009-02-17T00:00:00+00:00

I have been messing around with the Open Source FLV Player, and I have found that the FLV’s i was pointing it to didn’t have duration meta data. That can remedied by this simple shell command (as long as you have flvtool2 installed):

$ sudo gem install flvtool2
$ cat movie.flv | flvtool2 -U stdin movie.flv

Converting a Virtual PC .vhd to a Virtual Box .vdi

2009-02-04T00:00:00+00:00

I had some Virtual PC images in .vhd that I wanted to convert to Virtual Box .vdi files. After some tracing through a few Google searches and some experimentation of my own, the best way seems to be:

$ qemu-img convert -O raw start.vhd intermediate.bin
$ vboxmanage convertdd intermediate.bin final.vdi
$ vboxmanage modifyvdi final.vdi compact

Once you have the .vdi, I found that I had to enable APIC and IO-APIC to avoid a blue screen of death in my windows Virtual PC images on startup.

Unicode Line Separator

2009-02-01T00:00:00+00:00

U+2028

This little Unicode character gave me a run for my money today. Hard to see, it shows up as a space in every editor I tried, yet it caused JS unterminated string errors when it was used in a JS string in Firefox. Moral of the story, if you are using strings in Javascript, you gotta make sure to remove these little buggers.

How Rails Calculates the Cache-Busting Asset ID

2008-12-18T00:00:00+00:00

In asset_tag_helper.rb, you find this code:

def rails_asset_id(source)
  if asset_id = ENV["RAILS_ASSET_ID"]
    asset_id
  else
    path = File.join(ASSETS_DIR, source)

    if File.exist?(path)
      File.mtime(path).to_i.to_s
    else
      ''
    end
  end
end

Debugging bjam's configuration

2008-12-08T00:00:00+00:00

$ bjam --debug-configuration

This is useful if you think you are loading up a site-config.jam or a user-config.jam that you shouldn’t be. I was getting the following error after an upgrade to Intrepid Ibex:

error: at /usr/share/boost-build/build/project.jam:825 
error: duplicate initialization of gcc with the following parameters: 
error: version = 4.3.2

Then I was able to find that the site-config.jam file in /etc/ was initializing gcc.

Ubuntu Developer Man Pages

2008-11-25T00:00:00+00:00

$ apt-get install manpages-dev

Simple as that.

fopen error "Value too large for defined data type"

2008-11-20T00:00:00+00:00

Was working with some large data files from Wikipedia in C, and got a NULL pointer from fopen. After checking strerror, I got the “Value too large for defined data type.” error. The solution. Put the following definition at the top of your .c file and it will set you straight.

#define _FILE_OFFSET_BITS 64

Rendering rhtml files to string without a controller

2008-10-29T00:00:00+00:00

Sometimes you need to pierce the MVC veil and render a view somewhere lower level without an HTTP request, such as preemptively refreshing a cache without a client request, or outputing rhtml to a static html file. To preserve some of the “Don’t repeat yourself mantra”, it would be nice to have a way to render an .rhtml without a request/response.

av = ActionView::Base.new(Rails::Configuration.new.view_path)
av.assigns[:item] = @item
output = av.render('controller/view')

Thanks to these blog posts: Choonkeat and acts_as_renderer for the inspiration.

Tidy up XML on the Command Line

2008-10-20T00:00:00+00:00

This is a useful snippet if you ever encounter some XML using curl or wget that you want to format on the fly just for readability purposes.

$ curl http://myxml.com/myxml.xml | tidy -xml -i -w 0

Don’t know how many people need to read XML on the command line out there, but if you are doing any sort of XML parsing it may come in handy.

Fine Tuning Valgrind Callgrind Critical Areas

2008-10-09T00:00:00+00:00

Sometimes code that you want to optimize is deep inside your application and you only want to focus callgrind on that snippet. There are two steps here. The first is to include the callgrind header file, which contains a few macros to enable programmatic start/stop of valgrind instrumentation.

#include <valgrind/callgrind.h>

/* Stuff */

CALLGRIND_START_INSTRUMENTATION;
do_something();
CALLGRIND_STOP_INSTRUMENTATION;

/* More Stuff */

The next step involves calling valgrind with instrumentation turned off by default.

$ valgrind --tool=callgrind --instr-atstart=no bin/my_exec

When you close your executable, you should have a callgrind.out with just the results of the do_something() method, so you can focus your optimization efforts there.

Attaching to Child Processes using gdb attach

2008-09-30T00:00:00+00:00

I was working on a deviantly fishy segfault today that was introduced after I added daemonization to an existing service of mine. It turned out that the daemonizing code chdir’ed to the root folder, so the relative paths in the service no longer worked. However, it was hard to detect this in the child process as it silently failed.

Gdb Attach to the rescue!

Short story even shorter. Simply run your daemonizing process, then do a ps to find out the PID, then use

(gdb) attach <pid>

to connect to the process. Then you can easily see the backtrace when it ends up failing on you.

Secondly, if you are having trouble attaching fast enough, just add a sleep(10) somewhere at the beginning the child process, so you can connect fast enough.