So, london.pm technical meeting last night. You can read about what happened in acme's journal entry. Fun fun fun.
I gave my talk on Building Tests with Test::Builder, which also included some discussion of my Test::Builder::Tester to test the testing tool you were creating. It went okay, as well as can be expected. I think it was a little rushed, but it's quite a complicated subject matter to cover in a short talk. I think I should write it up as an article and submit it to somewhere.
Everything that could go wrong before this talk did. On Tuesday I was upgrading my laptop with Ximian, which included upgrades for libc and all. Which didn't work, and left me without libc. Which meant that non of my programs, like, say the login prompt or anything from init at all, worked. Thanks Ximian!
In the end I ended up booting into my Win98 partition and reading off my home directory with Explore2fs, since my last backup was two days old. This works quite well, but means that I lost all my permissions for the files. Darn. (I sort of fixed this later with a recursive Perl script that guessed what the permissions should be depending on the path and filename.)
blech and hitherto were kind enough to sit me down with a net connection and talk me though bootstrapping a Debian from the windows partition (I have no internal floppy or CD drive, and you can't really boot from the external USB ones.) So by 3am the same day (well, the morning after) I had a system that was apt-geted (or dselected) enough stuff to be useful. Some things need tweaking (for example, the wireless networking needs frobbing some more) but basically with a few more prompts from richardc I had MagicPoint working properly so I could work on my slides
This left me with having to complete my slides the night before the talk, which isn't ideal, but Tuesday night had been lost to the reinstall gods. Which would have been fine, if I hadn't got an email earlier that day telling me of a bug in Test::Builder::Tester, caused by an incompatability with the latest version of Test::Builder. Ooops. I'm now giving a talk on a module that no longer works.
Of course the bug would break some of the hardest to debug sections of tests ever. To test that Test::Builder::Tester produces the correct error messages when the Test::Builder module that it is testing produces incorrect error messages (i.e. so we get the right debug messages showing that the test we were testing didn't produce the right debug messages itself) we need to do some nasty stuff. Basically we reimplement a cut down version of Test::Builder::Tester in the tests to test itself. The trouble with the cut down version is that it doesn't have all the nicities that Test::Builder::Tester has - in particular no easy to debug colour highlight mode - so I end up staring at two screens of text for half the evening looking for the one char difference in the test output that is causing the tests to fail...
So that took up a lot of my evening. On the other hand, when I got a CPAN Testers message saying my module was broken the next day I was able to email back straight away and say "please retest the next version" ;-)
<Trelane> my fun todo?
<dispy>
i guess your fun todo is install slash on intranet server, gtk wrapper
for sudo, Test::GD, GD::TextImage, add things to grubstreet, read the
jabber book or patch T::P::XML::DOM documentation, write
documentation for TT and XML, write an rsync backup scripts or makepm script
or tidy up Test::GetSet and CPAN it
