Bad coding day

time to read 4 min | 671 words

I released NQA yesterday after putting quite some effort behind it. As soon as I released it, I seemed to have reached a karma high point and needed some nose rubbing to get me down to earth again. So todya I had:

  • Subversion' failing on me and thinking it commited successfully while the server roll back the transfer. To be honest, I think that this is actually an issue with Berlios.de recovering from the (yet unexplained) and removing my latest commit. (I'm talking about five minutes difference here, I just finished committing a bunch of files, made a minor change and tried to commit again, and was blocked.) This issue caused me to lose quite a bit of work; incidently I'd to refix that bug I'd fixed in the flash movie. It took me some time to realize that I did infact lose some data, so that was particulary nasty.
  • My current bug tracking system is strongly Unix based, this mean such things as somewhat limited support for IIS (the developers don't have servers to run tests on; and are quite uninterested in trying to do so :-( ). On of the results of that is that the mails from the system reach with Unix's line endings which are nearly unreadable when rendered in outlook. It should've been a simple change, but I made some mistake and took the system down. This time there wasn't any data loss, but a lot of productivity loss.
  • Next we'd TestDriven.Net decide that it's not going to play well with the debugger. After installing, uninstalling, checking the registery, reinstalling vs.net I checked the developer's blog and he had a more recent version that fixes this issue. Argh!
  • After intalling the new version, I can't test at all. Double Argh! The resolution was not to install to all the users in the machine but just to myself.
  • I spend half a day tracing a particulary annoying bug before realizing that I only ever tested a certain code path if I create the document from scratch, and not if I load if from disk. I update this and break the part that creates from scratch, I try to fix that and breaks the part that is responsible of syncing the model and the UI. I finally get what the real problem is and then I realize that the original code implemented this solution already, and I just had to add an if to fix the real problem. (Didn't really touch that particular code in 3 months, and it's a pretty complicated one.)
  • The subersion > Bug System integration is broken, and I've no strength left to even try to fix that. (nothing like trying to diagnoze a problem with a language you don't know (PHP) on a system you are not familiar with (unix) on a remote session using ssh with 0.4 seconds delay, which is just enough to be annoying.)
  • I'm writing this post and I notice that I made another GUI regression. I really wish I had a way to test GUI in a maintainable way.

Murphy, do me a favor, there are a lot of other people you can bug, what me? This day has been full of frustrations.

Oh, I got about 14 bug reports for NQA (all but 3 are resolved and in the repository and on the site).

The only open bugs right now relate to running on .Net 2.0 (untested as of yet) and using lazy collections (it should work).