[Opencxx-users] preprocessing and the parse tree

[Stefan S. <se...@sy...>]

hi there,

I'm currently looking into ways to enhance the ways
the parser can collaborate with the preprocessor.

I'm doing this in the context of my synopsis project,
where I generate cross referenced source code which
is easy to navigate and introspect (see
http://synopsis.fresco.org/docs/Manual/occ/index.html
for an example).

One of the challenges there is that the exact locations
in the file that should be linked isn't available
to the parser, as macro expansion has already taken
place.

I thus use a special preprocessor that provides me
with the information about macro definitions and
macro calls, so I can 'reverse apply' them to the
data stream.

While preprocessing I create a set of 'SourceFile'
objects that contain initially just these macro
definitions plus links to each other reflecting
the file inclusion dependencies.

Later more stuff is put in when the C and C++
parsers are run.

The problem I'm currently facing is related to
the lookup of the SourceFile objects during the parsing.
The parser (opencxx) can be queried for the 'current
file and line number' associated with a ptree node.

However, that's not quite enough:
As a file could be included more than once, and worse,
each time with a different set of macros defined [1],
I need to track exactly which inclusion this file
comes from. Just think about precompiled headers
for a moment and all the challenges this implies...

Has anybody already thought about this problem, in
particular in the context of opencxx ?

As I said, while my specific interest is synopsis
right now, I'd really prefer to work on opencxx
directly, as this work may be useful to others, too.

Best regards,
		Stefan

[1] My short term goal is to generate a full and
     correct reference manual for the boost project
     (http://www.boost.org), and if you'v already
     used boost before, you know what kind of challenge
     that means :-)