[Qgis-developer] Re: Rethinking the testing and release procedure of QGIS

[Donkagen]

Software testing is probably one of those areas that separates the men from
the boys. It is great fun to add new features and functionality, but
requires sustained hard work to ensure that the end result "just works". 

I'm currently writing a tutorial for individuals with no GIS experience to
start using QGIS for real tasks. As part of this I've started adding
warnings about known problems, incomplete features and functionality that
(to me) is non-intuitive... and I'm starting to feel defensive about a
product that I love, but isn't demonstrating its inner qualities. 

Unit testing is very important, but system testing in many environments is
also necessary to get statistical confidence in stability. 

Last month, Julien Malik mentioned about Sikuli ( http://sikuli.org/
http://sikuli.org/ ). I've had a preliminary look at the documentation and
it may have potential. I will not have time to try it for a few weeks, but
the possibility of developing scripts to exercise QGIS without any knowledge
of, or access to, the code looks useful. If I get chance to develop any
tests, I'll give an update.
In the UK, I'm trying to encourage the uptake of QGIS at the community level
and, hopefully with a demonstrably growing user base, to get some government
funding. To me, the two biggest "features" that need development are
software quality and documentation (probability in that order), and that is
where I personally would recommend funding. However, any such money remains
a long-term hope.

Quality is not an add-on, but a fundamental necessity to stay in business.
If anyone doubts this, try asking the ghosts of the British motorcycle and
automotive industries.


--
View this message in context: http://osgeo-org.1803224.n2.nabble.com/Rethinking-the-testing-and-release-procedure-of-QGIS-tp6534703p6559264.html
Sent from the qgis-developer mailing list archive at Nabble.com.