Muddygoat.org

Prettifying command output in scripts.

2012-04-04T00:00:00+00:00

This is so simple but so nice, that I had to share. Often, I want to create a small script which gives me some information, using another program, and I want to display it in some way which looks nice.

For example:

% tmux ls
dev: 1 windows (created Wed Apr  4 14:54:56 2012) [116x34] (attached)
main: 5 windows (created Wed Apr  4 14:31:07 2012) [116x34] (attached)

Can be wrapped up in a prettifying script, like so:

echo "Running tmux sessions:"
tmux ls

Which gives us:

Running tmux sessions:
dev: 1 windows (created Wed Apr  4 14:54:56 2012) [116x34] (attached)
main: 5 windows (created Wed Apr  4 14:31:07 2012) [116x34] (attached)

But this doesn't look too nice, and if there's a lot of output from different things, might just become one big block of text.

Never fear! With sed and pipes to the rescue, we can prettify things further. Indentation is one example of this, by replacing the start of each line output by 'tmux ls' with some whitespace. Our script might become:

echo "Running tmux sessions:"
tmux ls | sed "s/^/    /"

Which would give us:

Running tmux sessions:
    dev: 1 windows (created Wed Apr  4 14:54:56 2012) [116x34] (attached)
    main: 5 windows (created Wed Apr  4 14:31:07 2012) [116x34] (attached)

Or even nicer:

echo "Running tmux sessions:"
tmux ls | sed "s/^/--> /"

Which would give us:

Running tmux sessions:
--> dev: 1 windows (created Wed Apr  4 14:54:56 2012) [116x34] (attached)
--> main: 5 windows (created Wed Apr  4 14:31:07 2012) [116x34] (attached)

Nice, and very very simple.

Keeping git submodules up to date.

2011-12-04T00:00:00+00:00

Like many others, I keep my dotfiles in a git repository, to easily track them, back them up and share them across machines. This includes my Vim configuration and add-on scripts. Even better, I use the excellent pathogen script to keep each external script in a separate git submodule. If you're not familiar with this idea, I'd suggest watching Drew Neil's excellent VimCast on how to get it set up.

I won't repeat any of that here, but what I do always seem to forget is how to update the submodules to the latest version. Remembering that git submodules are always kept in detached head state, i.e. not on any branch (which makes perfect sense from a software reliability point of view), we need to get back on the master branch before we pull in the latest changes:

$ git submodule foreach git checkout master
$ git submodule foreach git pull

Then, when you're happy that everything still works:

$ git commit -a -m "update submodules"

That's it!

How to enable support for multi-lun USB devices.

2011-10-27T00:00:00+00:00

Damn, here was a problem. My shiny new USB smart card reader and integrated memory stick wouldn't quite work properly in Arch. The smart card reader was fine, but the mass storage device wasn't showing up. It seemed that the device had multiple LUNs (logical units), and the default Arch kernel wasn't configured to see these. No bother, we can add this to /etc/modprobe.d/modprobe.conf:

options scsi_mod max_luns=4

Then be sure to rebuild your initial RAM disk:

# mkinitcpio -p linux

And upon a reboot, the USB mass storage device is found!

(I have no idea why this functionality isn't enabled by default; perhaps there's some other problem that this can cause?)

Thanks to this article on the Adaptec knowledge base for providing the know how.

Sending desktop notifications from cron jobs.

2011-08-12T00:00:00+00:00

I often use the notify-send command to create pop-up notifications on my desktop when things happen (usually when a time consuming job completes). Here's an example:

% do_stuff && notify-send "Done"

I started sticking lines like this in some cron jobs too, but for some reason it wasn't working. Then I discovered via this snippet on Command Line Fu that it's because cron doesn't run as a sub process of the desktop - hence it's not aware of your X session. This can be fixed, with a simple:

export DISPLAY=:0
export XAUTHORITY=~/.Xauthority
do_stuff && notify-send "done"

Nice!

Inconsolata and boldface.

2011-07-17T00:00:00+00:00

Inconsolata has recently become my typeface of choice for monospaced environments (e.g. Vim, mutt, the terminal), where to be honest I spend the vast majority of my computing time.

It's an exceptionally readable and legible typeface, with all those nice features you'd want in a coding font, such as slashed zeros and obvious differences between punctuation marks. Here's it in action in gVim:

Notice the bold font that's used to indicate the article's title. This is in fact not due to Inconsolata itself, since it only comes with a medium weighted font, not a boldfaced one. Instead, gVim does some clever stuff that bolds out the medium weighted font. Other programs seem able to do this too, urxvt for example. However, not all get it right. Here's a shot of konsole's attempt:

Luckily, Sven Schwyn has created a variant of Inconsolata, called Inconsolata XL, which includes a bold weighted font. It's available here.

Here's konsole using Inconsolata XL:

Nice!

Building ruby 1.8.7 with glibc 2.14

2011-07-15T00:00:00+00:00

It seems that there's a bug which means that ruby 1.8.7 and ruby enterprise edition 1.8.7 don't build with glibc 2.14:

/bin/sh ./libtool --tag=CXX --mode=compile g++ -DHAVE_CONFIG_H -I. -I. -I./src  -I./src   -DNO_TCMALLOC_SAMPLES -pthread -DNDEBUG -Wall -Wwrit
e-strings -Woverloaded-virtual -Wno-sign-compare -fno-builtin-malloc -fno-builtin-free -fno-builtin-realloc -fno-builtin-calloc -fno-builtin-c
free -fno-builtin-memalign -fno-builtin-posix_memalign -fno-builtin-valloc -fno-builtin-pvalloc  -DNO_FRAME_POINTER -g -O2 -c -o libtcmalloc_m
inimal_la-tcmalloc.lo `test -f 'src/tcmalloc.cc' || echo './'`src/tcmalloc.cc
libtool: compile:  g++ -DHAVE_CONFIG_H -I. -I. -I./src -I./src -DNO_TCMALLOC_SAMPLES -pthread -DNDEBUG -Wall -Wwrite-strings -Woverloaded-virt
ual -Wno-sign-compare -fno-builtin-malloc -fno-builtin-free -fno-builtin-realloc -fno-builtin-calloc -fno-builtin-cfree -fno-builtin-memalign 
-fno-builtin-posix_memalign -fno-builtin-valloc -fno-builtin-pvalloc -DNO_FRAME_POINTER -g -O2 -c src/tcmalloc.cc  -fPIC -DPIC -o .libs/libtcm
alloc_minimal_la-tcmalloc.o
src/tcmalloc.cc:1672:54: error: conflicting declaration ‘void* (* __memalign_hook)(size_t, size_t, const void*)’
/usr/include/malloc.h:183:39: error: ‘__memalign_hook’ has a previous declaration as ‘void* (* volatile __memalign_hook)(size_t, size_t, const
 void*)’
make: *** [libtcmalloc_minimal_la-tcmalloc.lo] Error 1

While those that understand this sort of thing sort it out, there's a workaround posted on github. Here are the instructions (for ree):

Run rvm install ree-1.8.7
Wait for it to fail.
Fix the broken files with:

% cd ~/.rvm/src/ree-1.8.7-2011.03/source/ext/dl

% rm callback.func

% touch callback.func

% ruby mkcallback.rb >> callback.func

% rm cbtable.func

% touch cbtable.func

% ruby mkcbtable.rb >> cbtable.func
Re-run rvm install ree and the ruby files will not be re-extracted, so your changes to callback.func and cbtable.func will persist.

Theming the clock in KDE.

2011-07-13T00:00:00+00:00

The default digital clock plasmoid (that's KDE's fancy term for a desktop widget, by the way) in KDE 4 so far has not been particularly configurable. You can change the font it uses, but not the font size, nor specify in much detail how the date is displayed. For example, after a bit of fiddling around, I'd managed to make some progress, but it seems to "guess" a lot of things, and get them wrong sometimes. I got it looking like this:

It has clearly not guessed the right font sizes.

But, never fear, since there are a number of alternative clock plasmoids. One particularly configurable one is adjustable clock. For some reason, the built-in "get new plasmoids" thing couldn't find this for me, but there's an Archlinux PKGBUILD in the AUR.

On the face of it (no pun intended) this looks basically the same, but due to Plasma's great styling and scripting engines, you can specify any layout you like using HTML and CSS.

I wrote this snippet:

<div style="text-align:center; white-space: nowrap; opacity:0.85; background:none; padding: 12px 0 0px 0;">
    <span style="font-size: 52px; font-family:'Gill Sans MT Pro'; text-shadow: black 0.1em 0.1em 0.2em">
        <span style="color: rgb(255, 255, 255);"><b>%!t</b></span>
        <span style="color: rgb(200, 200, 200);"> %T</span>
    </span>
</div>

This is fairly self-explanatory, I guess, but %!t provides the time sans seconds, while %T provides the long date. The full range of options are available from inside the adjustable clock's settings panel.

Here it is what it looks like:

Now, there's still a quite annoying algorithm for scaling the widget, meaning that it's the ratios between font sizes and padding etc., that you have to get right, but a bit of fiddling and it's okay.

Let's hope KDE continues to remain as configurable as it has been known for!

Removing the delay in switching Vim modes in tmux.

2011-04-12T00:00:00+00:00

I've had a real annoyance with console Vim for a while now, which I just haven't been able to figure out. Today I did.

When you're in insert mode, you hit escape to leave and go back to normal mode. However, due to the way that terminals often map escape characters, when you hit escape, the terminal can wait around for a second, just in case you wanted to press another key and complete the escape mapping. But, this means that every time I wanted to leave insert mode, I was having to wait for around a second. If I didn't, then Vim instead inserted some sort of symbol like

ä

instead of doing

Esc d d

to leave insert mode and delete the current line.

I tried a few things, including reducing Vim's timeoutlen, which appeared to have no effect.

This really interrupted my flow, and led me to use graphical Vim much more than console Vim, since the problem didn't happen there. But today I found that it wasn't happening outside of tmux at all. So, must be tmux's problem.

The solution lies in telling tmux not to wait. I stuck the following in my .tmux file:

set -sg escape-time 0

And now hitting escape exits insert mode instantly, even in tmux!

Bulldozing Java's classpath system to just make stuff work.

2011-04-01T00:00:00+00:00

If you use Java, then you'll know it has this horrible (in my view) classpath concept, which usually just means more work to get something running (again in my experience.)

I came across this problem the other day when trying to run someone else's code, which was arranged neatly into projects, packages etc., each of which built to a separate .jar file in its own subdirectory... ugh. (In case you haven't realised yet, I don't much like Java!)

Anyway, yesterday I stumbled upon this little nugget, which makes running something structured like this a breeze.

First, stick this line in a script in the root directory of the Java project:

find . -name '*.jar' | tr '[:space:]' ':' | sed -e 's/:$//g' > classpath

which finds all the .jar files in any subdirectory, lists them, formats them suitably for Java to understand and sticks them in the file "classpath".

Second, run the program like this, using the newly generated "classpath" file to tell the virtual machine where to look:

java -cp `cat classpath` example.Main

Thanks to Cefn Hoile for the tip!

pacgem: managing Ruby gems in Archlinux with ease.

2011-03-23T00:00:00+00:00

Ruby is great. It's also designed to be platform independent and highly modular (i.e. through its gems system). The need to manage the gems and dependencies between them led to the development of its own package manager, rubygems, which works on any platform that can run ruby. However, here's the problem: what if your system already has a very good package manager (i.e. pacman)? Well, then you probably don't want to allow other things to start messing with what's in your file system.

One approach to dealing with this is to create a PKGBUILD for each and every gem (there are a few already in the AUR), but there are far more gems than can feasibly done by hand, and also unless an AUR helper which handles dependencies is used, it can be quite tiresome installing them all one by one. This solution is really not scalable, given the number of ruby gems that exist (22,366 according to rubygems.org).

Enter pacgem! This was created by minad and announced here on the Archlinux BBS.

It seems to me that pacgem makes managing ruby gems in Archlinux an absolute breeze. Basically, it creates a custom PKGBUILD for the specified gem, on the fly, handling dependencies between gems automatically.

Observe:

% pacgem nesta

==> Saving package files in /home/pete/arch/builds/pacgem/ruby-bundler
==> Resolving gems...
  -> (New) nesta-0.9.4: Ruby CMS, written in Sinatra
  -> (New) haml-3.0.25: An elegant, structured XHTML/XML templating engine. Comes with Sass, a similar CSS templating engine.
  -> (New) maruku-0.6.0: Maruku is a Markdown-superset interpreter written in Ruby.
  -> (New) syntax-1.0.0: Syntax is Ruby library for performing simple syntax highlighting.
  -> (New) RedCloth-4.2.7: RedCloth-4.2.7
  -> (New) sinatra-1.1.2: Classy web-development dressed in a DSL
  -> (New) rack-1.2.2: a modular Ruby webserver interface
  -> (New) tilt-1.2.2: Generic interface to multiple Ruby template engines
  -> (New) shotgun-0.9: reloading rack development server
==> Making package: ruby-haml 3.0.25-1 (Tue Mar 22 19:23:59 GMT 2011)
==> WARNING: Skipping dependency checks.
==> Retrieving Sources...
  -> Found haml-3.0.25.gem
==> Validating source files with sha256sums...
    haml-3.0.25.gem ... Passed
==> Extracting Sources...
==> Entering fakeroot environment...
==> Starting build()...
==> Installing gem...
Successfully installed haml-3.0.25
1 gem installed
==> Fixing gem installation...
==> Tidying install...
  -> Purging other files...
  -> Compressing man and info pages...
  -> Stripping unneeded symbols from binaries and libraries...
  -> Removing empty directories...
==> Creating package...
  -> Generating .PKGINFO file...
  -> Compressing package...
==> Leaving fakeroot environment.
==> Finished making: ruby-haml 3.0.25-1 (Tue Mar 22 19:24:03 GMT 2011)
==> Checking ruby-haml-3.0.25-1-any.pkg.tar.xz with namcap...

A short while and several pages of text later, the nesta gem and all the other gems it depends on are installed!

And, crucially, all the files are still under the management of pacman, each gem with its own package. E.g.

% pacman -Qo /usr/lib/ruby/gems/1.9.1/gems/nesta-0.9.4/Gemfile

/usr/lib/ruby/gems/1.9.1/gems/nesta-0.9.4/Gemfile is owned by ruby-nesta 0.9.4-1

% pacman -Qo /usr/lib/ruby/gems/1.9.1/gems/haml-3.0.25/Rakefile

/usr/lib/ruby/gems/1.9.1/gems/haml-3.0.25/Rakefile is owned by ruby-haml 3.0.25-1

Nice! Thanks minad!