Blog - /Tech

So you've recently converted a darcs repo into a git repo. Good choice! Unfortunately, the conversion process made it so that people who used a short login name rather than a full name when committing now have the email address for their commits set to your own email address. Not good. Here's how to fix that, by changing the email address in their commits to be just their short login name.

git checkout master
git checkout -b rebased
git-filter-branch --env-filter \
  'if test "$GIT_AUTHOR_NAME" = $(echo $GIT_AUTHOR_NAME | cut -f 1 -d" "); \
  then export GIT_AUTHOR_EMAIL=$GIT_AUTHOR_NAME; fi' HEAD

If you like what you see, then it suffices to do the following to make the master branch point to the new commit history.

git checkout master
git reset --hard rebased
git branch -D rebased

For a while I thought I was using kqemu, but in fact was not. Here's how to fix that.

1. sudo module-assistant a-i kqemu
2. Edit /etc/modules and add the line:

   kqemu

3. modprobe kqemu
4. Make a new file called /etc/udev/rules.d/60-kqemu.rules. It should contain the line:

   KERNEL=="kqemu", NAME="%k", MODE="0660", GROUP="mwolson"

   Change the group to something appropriate for your setup.

5. Run udevtrigger to make the previous step take effect.
6. Run qemu, passing it the --kernel-kqemu option.

When I switched to Pulseaudio recently, I noticed that MusicPD started consuming way more CPU time than it used to. The culprit is the following lines in ~/.mpdconf.

mixer_type                      "alsa"
mixer_device                    "default"
mixer_control                   "Master"

The fix is to use the software mixer instead:

mixer_type                      "software"

Delving further into what makes Python difficult for me to debug, I think part of the problem might be the backtraces. Emacs Lisp and Perl backtraces display the affected functions and files from the inside out, while Python takes the unnatural step of reversing the order.

Another specific problem that I have with some Python programs is the sheer number of small classes that they are composed of, often with many classes per file. Better is to have one class per file, with that class having more than just 3 or 4 methods. If it has less, you are abstracting too much! (Aside: This is not to say that Java gets it right. Hell no. The compiler should not care what you name the file that happens to contain your class.)

As for replacing existing programs: I don't really care what language the program is written in, as long as it works. I enjoy Miro now, but when it was broken in the alpha days for Ubuntu Gutsy, it was a huge pain to try to figure out what was going wrong. Things that I enjoy extending and hacking on, such as blogging software (PyBlosxom), are prime candidates for replacement. In the case of PyBlosxom, I am going to try out ikiwiki. I have even done some initial work on getting Emacs Muse to work as a potential extension for it, so that I can push .muse files to my webserver using git, and have them published automatically once there. I still need to buckle down and write the Perl glue to connect to this Emacs Muse "compiler", though.

Dear Python,

I love some of your applications, but I hate debugging them. You enable programmers to write shitty and ill-advised objects and your advocates call it "magic". The magic is gone from our relationship — it's time to call it quits. Any Python program that I've ever wanted to extend or fix will now be replaced with something that is written in a saner language. Goodbye, Pyblosxom. Good times, MoinMoin. Good riddance, Bazaar — being a GNU project can't fix what ails you. Tailor, you are a horrible and poorly-documented siren of version control interoperability, and I have wasted enough time on you.

I'm on twitter now: http://twitter.com/mwolson. I'd like to use it as a way to drive facebook status updates. That means posts will be in third-person present tense. I've added Twitter updates to my Website: About Me page. I like the look of having the list of updates be inside of a single-item definition list.

I am pleased to announce the release of Remember 2.0.

Remember is an Emacs mode for quickly remembering data. It uses whatever back-end is appropriate to record and correlate the data, but its main intention is to allow you to express as little structure as possible up front.

Release info

Tarball

http://download.gna.org/remember-el/remember-2.0.tar.gz

Zip file

http://download.gna.org/remember-el/remember-2.0.zip

Commit ID

79f69e6188e4e084c9761b24ecb54b5cac9f41e3

Instructions follow for those who track the source code using git. If you wish to track the development of Remember, visit <http://www.emacswiki.org/cgi-bin/wiki/MuseDevelopment> for instructions and substitute "remember" whenever you see "muse" :^) .

This release has the signed git tag "v2.0". The tag points at a commit object with the Commit ID mentioned above. To verify the tag, do:

git fetch origin
git tag -v v2.0

To check out the release, do:

git checkout v2.0

To switch back to the master branch, do:

git checkout master

News Items

Remember is now a part of Emacs 23, and will be included with that release of Emacs when it comes out.

The NEWS items for this release are available here.