Re: test driven development for GNUstep

[Lars Sonchocky-Helldorf]

discuss-gnustep-bounces+lars.sonchocky-helldorf=bbdo-interone.de@gnu.org 
wrote on 02.02.2004 17:27:05:

> 
> On 2 Feb 2004, at 15:06, Lars Sonchocky-Helldorf wrote:
> 
> > I think it would be good to employ the test driven development 
> > methodology
> > for developing GNUstep.
> > Test driven development generally ensures a high qualtity of the 
source
> > across several platforms (because everybody is able to run the 
relevant
> > test on her/his platform of choice and send the test failures back). 
> > For
> > instance the GCC project employs test driven development in the form 
of
> > DejaGnu: If you build GCC you can optionally run 'make -k check' to 
> > run a
> > test suite for GCC and send the results back to the GCC developers.
> >
> > Now there are several ways to do test driven development, on the ObjC 
> > side
> > of this world for instance DejaGnu 
> > (http://www.gnu.org/software/dejagnu/)
> > and ocunit (http://sente.epfl.ch/software/ocunit/). Anybody here knows
> > which would be better suited for GNUstep library development? I hope 
> > this
> > will initiate a fruitfull discussion which results in the use of test
> > driven development for GNUstep (in whatever form).
> 
> I wrote GNUstep-Guile/Greg as a replacement for DejaGnu because I found
> it so horrible to use for testing the GNUstep libraries ...  I haven't 
> used ocunit.
> 
> The only real problem with the existing testsuite is that it uses guile 
> ... which
> is better than the tcl stuff used by dejagnu, but suffers from being a 
> language
> which is unfamiliar to everyone working on GNUstep (and I include 
> myself).

at least dejagnu is widely used and installing it should be a no brainer.

> 
> I think if you want all developers to be able to trivially run the 
> testsuite as part
> of the development process, you need to write the test stuff using only 
> make,
> and shell scripts and ObjC code,

I think ocunit fulfils this those requirements, although I am unsure if it 
is a good idea to use a tool for testing a library which is based on the 
same library it should test. Personally I am more a *unit person than a 
dejagnu person but since I don't know if ocunit would work here I've taken 
dejagnu into consideration.

> and I think you need to have it all copyrighted
> by the FSF so that it can be distributed as part of the GNUstep core.

I don't see any problem with this. In fact I think it would be wrong to 
distribute the testing code separately since this would only set the bar 
for testing higher.

greetings, Lars