Instead of

<mirrorOf>*</mirrorOf>

I now have

<mirrorOf>*,!clojars</mirrorOf>

That is, use my mirror definitions for anything except clojars. Before I did this, leiningen always failed to download clojure dependencies from clojars (that were not also in maven central). The reason why I do not want clojars to be mirrored is a dark and deeply kept secret.

In theory, this should be easy. Just install the necessary python packages, setup graphite and its components (whisper database and carbon I/O-server), start a web server and – presto!, off you go.

With one exception – pycairo - all of graphites dependencies can be installed using pip (or easy_install). Pycairo, on the other hand, turned out to be another story completely. Pycairo is a package of python bindings for the cairo graphics library and is used by graphite to produce the graphs that we (well at least I) have come to love.

Zeroth problem: pycairo depends on cairo which is best installed using homebrew.

First problem: pycairo cannot be installed using pip (writing this I am unsure whether I tried easy_install at all). Something about a missing setup.py file. Ok, guess I could install it from a source download then…

Second problem: Installing pycairo from source (tar-ball) is not possible because of some kind of incompatibility with Python 2.6.1 (default installation on  OSX Snow Leopard) and perhaps some kind of cross compilation issue for x86_64 cores (sigh).

Third problem: Ugrading the (default) python installation (especially beyond 2.6) on OSX may confuse yourself or parts of your system (that depend on Apples installation of python). See for example the homebrew documentation for its python 2.7 formula.

A hero emerges

Having spent the better part of my evening getting this far I was about to give up. Then it dawned on me – why am I spending all this time trying to upgrade my default python installation when I should be using virtualenv?

Indeed, with python 2.7.1 installed via homebrew (but not overriding Apples default installation) I could get a python 2.7 virtual environment up and running. In the new python 2.7 environment I could perform the pycairo installation from source with a few tweaks (see below). After that, moving on with the final installation steps for graphite (still inside the virtual environment) was a breeze.

So, next time I hope I will remember setting up a virtual environment sooner.

–

I would never have realized what I was missing or doing wrong if it had not been for these posts:

1. http://eternusuk.blogspot.com/2011/09/installing-graphite-on-osx-snow-leopard.html
2. http://stackoverflow.com/questions/6886578/how-to-install-pycairo-1-10-on-mac-osx-with-default-python 
3. http://www.niconomicon.net/blog/2011/10/09/2120.wrestling.python.snow.leopard

git log to the rescue.

‘git branch -r’ will list the remote branches tracked by your local repository. Using it in a git-svn repository will revealy all branches tracked on the subversion server including trunk. The trunk branch tracks the remote trunk and is not the same as your git master branch. Hence trunk can be compared to my working branch, where all my local commits end up. A terse and, in my opinion, clear solution to the comparison problem is the use of the dot-dot range operator:

git log trunk..

A missing branch name at one end of the range (here after the two dots) is replaced by the HEAD of the currently checked out branch. The above snippet then translates to log commits reachable from HEAD but not from trunk.

I currently use this last form a lot.

This Stack Overflow thread proved me right: http://stackoverflow.com/questions/957928/is-there-a-way-to-get-to-the-git-root-directory-in-one-command

So, adding that information to the suggestion further down the thread I created a shell alias

# Jumping to git roots
alias git-root='cd $(git rev-parse --show-toplevel)'

http://stackoverflow.com/questions/3376863/unable-to-type-braces-and-square-braces-in-emacs

Specifically I have been wanting Emacs and Eclipse to be able to use the same scripts and applications from the command line as my shell.

Here is a rather short explanation of how to make an environment variable (like PATH) available to a “.app”-application on MacOS X.

Not much later I also found a way to automatically update the “environment.plist” by way of a shell script, here (which links to the origin of the script here).

Basically, there is a properties file called environment.plist located under ~/.MacOS in my home folder. I can update the PATH property in this file (instead of as a bash environment variable) to make paths available to other applications. Sadly it appears to need a restart of MacOS to take effect. Not what I had been hoping for.

http://fixunix.com/ubuntu/543595-where-does-ubuntu-mount-samba-shares.html

Why do I want to know this? Well, I frequently use trial and error search using nautilus (or Places -> Connect to Server) to find or browse network folders in Ubuntu. Once they are connected, Ubuntu seems to secretly mount them to somewhere.
Somewhere, in this case, is the directory .gvfs in your user home, where they can be reached by any able terminal or application.

My first discovery was that there is a BIOS setting that disables USB keyboards during boot (and, hence, in the Ubuntu GRUB2 boot menu). For some inane reason the default BIOS setting was to disable my keyboard. With the keyboard disabled, there is neither any way to operate the custom boot order menu nor to choose which operating system to launch once GRUB and the dual boot menu was in place.

My second discovery was StartupManager, a graphical tool that lets you change the order and defaults of the GRUB boot list. Instead of having to remember which of the grub files to change (and not to change), what tool to run in order to regenerate lists or menus inbetween changes and having to understand the arcane and cross-referencing syntax used between the files, StartupManager does the job for me through a rudimentary GUI. Much less hassle for this task that I typically only perform once or twice per dual boot installation.
As usual, getting StartupManager is dead simple using apt:

sudo apt-get install startupmanager

After having set things up to (what I thought was) my satisfaction I discovered that Windows 7 did not like being booted after GRUB2. For some reason, I gave up after googling for a few minutes, Windows wants to repair itself on every startup after I have used Ubuntu. Annoying, especially so since repairing is the default option which is started if the computer is left alone. Instead of fighting Windows I decided to go the other way around, that is have the Windows bootloader start first and add an option to it to start Ubuntu. So, the third discovery I made was a tool for windows, very much like StartupManager for Ubuntu – EasyBCD. Amongst other thigs, EasyBCD relieves me of BCD.exe, a command line tool which looked so complicated that I did not even bother to learn it. It also automatically discovered my Ubuntu installation (and GRUB2 on it) and could create a windows boot loader menu option to launch it.

Ok, I think that covers what I learned this time. Hope I can remember it next time around.