C++ tools on the EBN

Robert Scott lists at humanleg.org.uk
Mon Jul 9 20:30:48 CEST 2007


On Sunday 08 July 2007 20:54, Adriaan de Groot wrote:
> We have had r++ (which is also included in Qt 4.3) in kdesdk/ebn for a long
> time now. It's quite stalled, and Robert's take on it is that it is
> difficult to work with. On the other hand we don't have a buildable
> PUMA-based parsing tool in there at all. I'd like to suggest we add the
> following to
> kdesdk/ebn/:

I don't know quite where my PUMA tools are going (if anywhere) - I don't want 
to end up pushing people into a bad decision just because I've limited my 
options to PUMA and rpp. I've been looking at some other possibilities here.

That being said, I would of course welcome people taking a look at the 
greenbelt sources in an environment where people can easily submit patches 
and criticism. For all I know, it'll be exactly what people want and it'll 
take off. Or people will tell me it sucks. Either way, it's better than it 
just sitting on my disk for the next couple of years.

> I imagine that importing puma might cause some squeals -- I forget how big
> it is exactly. Comments?

A modified / sanitised / pre-woven / clean-compiling copy of puma in svn would 
make things easier. But it might be a bit much for something which might get 
dumped.

On the 'other possibilities' front, something I've been looking at is LLVM. 
For those who don't know, LLVM is really a highly modular code generation 
framework which can do all kinds of compilation, jit, runtime optimisation 
etc.

The biggest plus of using LLVM is it would probably allow us to do tests at 
many levels, all using one framework. Tests could be done at the token level, 
the AST level and, if someone's feeling really clever, the intermediate 
representation level. All done through LLVM. And LLVM is maintained. 
Celebrate.

But LLVM is very complex and I don't understand it yet.


robert.


More information about the kde-quality mailing list