I just spotted the following marketing blurt through testdriven.com:
[name of a build server product] helps software organizations reduce high risks of failures of projects caused by broken code base by delivering uninterrupted daily builds.
Yes! Finally! All of our software projects will be raging successes now that we have technology that builds our software automatically!
If you didn't spot the sarcasm...
It's not the tools that make continuous integration happen. It's the people. Even the most advanced build server can't do shit if the people don't integrate their changes continuously. Even the most advanced build server can't spit out a good build if the people have screwed it up.
This is not a bash against Parabuild. It could be a terrific build server. It's just that I start hearing these voices in my head when someone even remotely suggests that there's a new tool that somehow magically fixes the stupid humans.
Robert, thankyou for the clarification. However, when starting to put up the CruiseControl run, we did already have the Ant script doing what we wanted: CVS checkout, clean build with Clover instrumentation, test run with Clover, and generating the Clover reports.
The problem was that when setting CruiseControl to run this in the build loop, the build kept crashing in OutOfMemoryErrors. As I said, I don't mean to imply that this would be CruiseControl's fault; nor am I interested in finding the guilty one, but rather in getting Lahtinen and the machine gun back, as Hietanen very well puts it in a Finnish classic.
When tuning the VM and various Ant and CC parameters, I found out that we were not alone. Don't get me wrong, I think that CruiseControl is a great tool, but I like it better served with only moderate hype ;) It can easily be more work to set it up in a big project to do non-trivial things (such as the Clover runs) than what is being done in tutorials.







