On Thursday 27 February 2003 23:13, shaheed wrote:
> This is what I did with the first generation MS Word filter:
> 
> 1. I had a set of regression test documents, as you do
> 
> 2. I wrote a script which invoked kword on the file. Unfortunately, I did 
not 
> know about koconvertoer, so I sat there doing a save and then exit from 
kword 
> (I did try scripting that using DCOP, but had a lot of trouble with that)

I think koconverter didn't exist when you developed your filter.
 
> 3. The script then extracted the top level xml from the .kwd, and compared 
it 
> against a reference copy
> 
> 
> Then , when I made a change, I checked that no unexpected changes occurred 
in 
> any of the XMLs. The changes I expected, I checked, and then updated the 
> regression copy. It used to take 2 hours to run the suite.

Okay, there are two things I don't like about that approach:
- It loses some of the data (because KWord doesn't use the whole CHP and so 
on). This doesn't trigger bugs in KWord, but might still be a regression in 
wv2.
- It's not trivial to update the reference documents when you implement a new 
feature (or KWord changes its file format).

Of course it's better than no testing at all (or visual testing of 10 
documents instead of the whole test set of 150 or so).

Ciao,
Werner
____________________________________
koffice mailing list
koffice@mail.kde.org
To unsubscribe please visit:
http://mail.kde.org/mailman/listinfo/koffice