[Kde-bindings] Incorrect number of ancestors in kalyptus ??

[Richard Dale]

Hi Marcus

On Thursday 30 October 2003 09:57, Engels, Marcus wrote:
> > Hello,
> >
> > I am trying to generate my own description of the Qt classes with
> > kalyptus from the kdebindings package. When creating the list of
> > Ancestors for a class in files like kalyptusCxxToC.pm in subroutine
> > writeClassDoc the number of Ancestors found has been > 0 in older
> > versions of kalyptus if there is more than one inherited class in the *.h
> > file. I think in the newer version this number has changed to 0 if the
> > class inherits one other class, 1 if it inherits 2 etc.
> >
> > Can anybody tell me if this change in counting the Ancestors is a bug (
> > writeClassDoc still checks for $#ancestors > 0 which might fails if a
> > class has only one ancestor) or if the change is correct for the newer
> > versions of kalyptus?
I'm not sure exactly what language or format you want to generate. But I 
wouldn't start with the C bindings generation as an example. I was intending 
to rewrite it - it doesn't do a lot and it wouldn't take very long at all. 
The output really shouldn't need any hand fixing up, and I would like to get 
all autogenerated code out out the cvs, and generate it as part of the 
kdebindings configure step. But I don't remember changing the C bindings 
generation at all recently, so I wonder if the $#ancestors change is 
something someone else has done.

The java one in the HEAD is probably best (or the Smoke code generation 
option, but that might be harder to follow). The C code generation is really 
badly written - it parses its own arguments to methods, but now that is built 
into to the kalyptus front end. Also kalyptus also has a proper symbol table 
which fully resolves all the types at code generation time, not the way the C 
bindings do it as a bit of a hack where all the types are hard coded 
statically into a table in kalyptusDataDict.pm.

-- Richard