Regression tests?

[Dan Kegel]

Bill Gribble wrote:
> 
> On Tue, Jan 30, 2001 at 03:56:54PM -0800, Dan Kegel wrote:
> > i.e. can I script the import of two qif files and check to make sure
> > the result is correct?
> 
> The big problem with testing qif import right now is that there's no
> totally scriptable way to do it ATM.  It wouldn't be hard to do,
> though, if you assume some reasonable defaults (or have parameters)
> for currency etc and let previous account mappings be used.
> 
> I can help with implementing a scripting interface for qif import if
> you are interested.  It's been on my to-do list for a while.

I guess my question is broader than that.  Having seen
various crash reports on this list, I'm wondering what
can be done to catch bugs before end-users do.

The project I'm working on now (a keyserver) has unit tests
for most classes; we generally write the unit test
after designing and before implementing the class.
It also has lots of integration tests, all unattended.
All the tests are run from our build script whenever we build,
and also are run unattended on two operating systems twice a day.
We catch a lot of bugs that way before the code even hits CVS.
We also have stress tests, though we don't run those daily yet.
The resulting server has been pretty reliable.

Could something similar be done for gnucash?  Would it help
make crash bugs less likely?

- Dan