Screen scrollback revisited

In this post, I comment on entering copy mode in GNU screen in order to scroll back with the keyboard. But what if you just want to use the scroll wheel? I think what you're looking for is this line in your configuration file, ~/.screenrc:

# scrolling:
termcapinfo xterm ti@:te@

I also found this cryptic code in my configuration file; I can't tell if it's relevant or not:

# Scroll up
bindkey -d "^[[5S" eval copy "stuff 5\025"
bindkey -m "^[[5S" stuff 5\025

# Scroll down
bindkey -d "^[[5T" eval copy "stuff 5\004"
bindkey -m "^[[5T" stuff 5\004

# Scroll up more
bindkey -d "^[[25S" eval copy "stuff \025"
bindkey -m "^[[25S" stuff \025

# Scroll down more
bindkey -d "^[[25T" eval copy "stuff \004"
bindkey -m "^[[25T" stuff \004

Happy screening!

Correct slide alignment with Beamer

Sometimes -- and I'm not sure exactly when or why -- I find that my slides pick up a vertical offset. Every slide except the first one is shifted downwards by a few pixels. This is frustrating if you have a background image that you would like to align exactly the same way on every slide. The fix, found here, turns out to be:

• use a newer version of PGF (I think 2.0 or greater), or
• insert the following in the preamble:
  \makeatletter \let\Hy@FixNotFirstPage=\relax \makeatother

The issue apparently has something to do with a strange interaction between the hyperref and pgf packages.

Testing bunzip2 files quickly using hexdump

My work sometimes involves dealing with large amounts of data stored in .tar.bz2 files. Recently, we found that some of the files were corrupted. The files are large enough that a successful "bunzip2 -t" takes prohibitively long. As a shortcut, we can just check for the bzip magic numbers at the beginning of the file. The bzip2 format requires that each file starts with the string "BZh". Our files are stored in batches by calendar date (MMDD). I generate a log of the files' magic numbers like so:

(for d in ????;
do
  for f in $d/*bz2;
  do
    echo $f `head -1 $f | hexdump -c -n3 | grep '0000000' `;
  done;
done) > magic_numbers.txt

And then I can find the bad files like so:

grep -v 'B Z h' magic_numbers.txt

hexdump is a nice tool for manipulating binary files on the command line. This method of testing the files runs relatively fast.

Enabling hibernate on Ubuntu 12.04 LTS

I recently decided I would prefer to have the option of using hibernate on my laptop running Ubuntu 12.04 LTS. Not only does hibernate need to be enabled manually on any system running that OS; I also had insufficient swap space, so I had to resize my partitions.

Now hibernate would work, but resume would not! It turns out the reason is that I elected to encrypt my $HOME folder during installation. This requires encrypted swap; when you reboot, the swap is encrypted with a fresh, random key each time.

By insisting on using the same key each time, which requires you to enter a password during boot, you enable the boot system to detect and reload the hibernate image. If you forget the password, you boot with no swap partition.

Detailed instructions are here. It worked for me!

Formatting terminal output in Python

From Nadia Alramli's Blog: Terminal Controller for Python. This page provides a simple module for prenting formatted text to the terminal using low-level commands in the curses module.

JACK + Ardour + headphones on a laptop

I use JACK + Ardour for music recording and mixing. For listening, I typically just use a laptop's headphone jack. This can lead to strange and confusing behavior. Well, no more: at least on my laptop, this all can be avoided by connecting Ardour's master out to system:playback_3 and system:playback_4 in addition to system:playback_1 and system:playback_2. This way, the master out goes to all the software channels, and somewhere, Ubuntu decides whether the headphones are plugged in and only sends the output to the correct place.

Saving executable permissions with Subversion

After getting angry at Subversion for failing to save "u+x" permissions on my version-controlled scripts, I finally Google'd for the solution. Turns out, it's simple. Just as you use "svn move" rather than "mv" to move a versioned file, you use "svn propset" to make a script executable:

$ svn propset svn:executable ON [filename]
$ svn commit

Thanks to John M, who says he referred to this page (see also these details).