On Sat, Nov 20, 2004 at 01:30:01PM +0000, Karel Kulhavy wrote: > Try to not make the bugs in PCB in the first line. Then you are going to > have small number of bugreports and less complicated bugreporting system > will suffice. > > How to do it: > * Have every interface between things specified > * Keep specifications up to date > * After writing a piece of code, read it once more over from top to > bottom like a novel is being read.
> Having doing lately almost nothing than filing various segfaults into the > bug database from work with PCB makes me thing the PCB is being > developped in the style "quickly write something made with a hot needle > and then if someone finds some bugs, repair them". PCB is 14 years old and the original author is no longer involved. A number of bugs I'm sure have been around for ages and a number of them are new too. One if the issues is that it's often times the case that a bug doesn't show up until the program is used by different users who do different things to the tool. In an ideal world, the entire thing would have been written following a test driven design approach with every small module having a full testsuite which is run automatically everynight. However, this was written in the real world by volunteers with limited time and it is difficult at best to retrofit the code for automated testing. We are not committing code which has not been tested, but the reality is that until someone steps up to the plate with an exhaustive automated testsuite, the tests will never be as complete as real users using the tool for hours. > Bugs are not normal, as well as stealing or killing people is not normal. > Bugs are evil. The fact that things like Bugzilla exist doesn't make bugs > any more normal, as well as existence of punishment doesn't make stealing > or killing any more normal. by your argument, we should evidently get rid of bug tracking systems as well as a legal system for dealing with criminals?? --
