Re: GCC/UML/XMI extraction (was gcc front end )

[aschwarz1309 at att dot net]

It sounds as if what is wanted is the ability to to 'end-
around' processing, that is, generating code from UML 
and then generating UML from code.

The generation of code from UML shouldn't be difficult. 
The 'end-around' part can be hard and commercial vendors 
have salesman that sell you the world, with product 
documentation which (sometimes) tells you the caveats. 

In any case, you do not need a compiler to generate code 
from UML. Assuming that the user put's correct 'code'  
and 'data' into the classes, then all linkages are 
strictly ordered with known semantics and 'code 
generation' consists of outputting the code, data, and 
class declarations according to user's intent and  
langage defined semantics.

If either end-around capability or syntax checking of 
user inputs is required, then a compiler front-end is 
required. However, 'you own the turf'. You can define 
acceptable inputs and unacceptable inputs so that issues 
raised can be circumvented. You define the environment 
needed to be successful and then define the result if 
input conditions are violated. Otherwise, there is 
bedlam. For example:

C++:    class 
        # ifdef FOO
              bar
        # else
              BAR
        # endif
              { ... };

        class derived :
        # ifdef FOO
              bar
        # else
              BAR
        # endif
              { ... };

Should work but what do you do with it?

Legacy code is another matter. But, if your 'rules' are 
violated then legacy code will only be partially 
understood. A problem, but then legacy code is always a 
problem. 

You may want to involve yourself with one of the ongoing 
UML projects and leverage your interests onto work 
already started.

art