A small test case revealed weak points in my code. Fixing each point was making the code worse and worse. Fortunately, I got insight. Why not to use a fake node? I did it, and code became simpler, hundred lines smaller, and I automatically fixed a weak point for which I had had no idea for a workaround.
Archive for May, 2005
A small firm has a Subversion server located in a local network. The network is secured by a firewall. The task: give access to the Subversion server from the outside, both to Linux and Windows users. The solution should be as secure as possible.
I’ve spent 3 days on getting simple code to work. In exchange I’ve got more experience with xsltproc internals. But I’m not sure it’s a fair trade.
A confirmation letter is received: “… for the Summer School on Generative and Transformational Techniques in Software Engineering (GTTSE 2005) has been completed. As a result, your participation in the summer school is now confirmed.”
Finally, SourceForge.net repaired statistics for the projects. It was a dream since the middle of January.
The Scheme function “x:apply-templates” should work like the XSLT function “apply-templates”. The very basic test case is passed today.
Due to historical reasons, my Linux box at work uses Promise RAID. Recently (month ago) I added a new HDD (AFAIK, attached to ICH5). Unfortunately, the HDD didn’t work.
Shame on me. The TeXML processor had a bug in the core functionality.
Yesterday I updated libxml2 and xsltproc from CVS and got a coredump. I am very scared because I can’t produce a test case and I think that the problem is on libxml’s side.
I found a non-obvious issue in the Namespace Recommendation and fixed my converter. Then I decided to look at other popular de-facto stanadrd tools. They are buggy. So I don’t understand why I work so hard to make my program as correct as possible.
XML namespaces are an invention from the evil. Two harmlessly looking messages in one of the scripts evolved to two hard debug sessions. Fortunately, I managed to fix the both bugs.