New parser branch (Was: Dumping the source DOM?)

[Vladimir Prus]

On Tuesday 12 July 2005 20:42, Roberto Raggi wrote:
> On Thursday 30 June 2005 15:47, Vladimir Prus wrote:
> > ghost at zigzag:~/build/kdevelop4-parser$ r++ lexer.cpp
> > Segmentation fault
> > ghost at zigzag:~/build/kdevelop4-parser$ r++ ast.cpp
> > Segmentation fault
> > ghost at zigzag:~/build/kdevelop4-parser$ r++ tokens.h
> > Segmentation fault
> >
> > Am I doing something wrong?
>
> strange! it seems to me you're using the wrong ``r++''(maybe the one
> distribuited with kdevelo3!!?). Please, can you try with "./r++ -dump
> file.cpp" (NOTE the ./ prefix)

They key has the "hash -r" invocation. With it, segfault is gone. I'd really 
suggest putting together README.txt that will say about "hash -r".

> anyway, gccxml is *NOT* a solution for KDevelop. We need to parse the
> code(actually the *incorrect/unfinished* code) in realtime. 

Well, at least Boost headers are correct and finished ;-)

> r++ is good 
> enough for KDevelop. Actually it is pretty good. It can parse Qt, KDE and
> most of boost.

I can take your word for that, but it would be much desirable to have a set of 
tests. At least I always try to create as many tests as possible, and since 
r++ is standalone, this should not be hard. 

Speaking of Boost, when I use the CVS version of it, and try to process a file 
containing just:

  #include <boost/program_options.hpp>

I get:

$ ./r++ -I/home/ghost/Work/boost-rc po.cpp
po.cpp:3: unexpected token 'token:0 -  '

While "g++" has no problems. What should I do?

- Volodya