Thanks for the reply. On Thu, 2006-04-20 at 12:36 -0400, Miguel de Icaza wrote: > > Having it that way might make things more complicated then they have to > > be. I'm not sure how much of the current code comes from mcs, but it > > would make sense to just make mbas parse VB.NET code and then generate > > the appropriate datastructures for the classes in gmcs. For example, I > > don't think mbas should contain a 'Statement' class but should rather > > use the one from gmcs directly (rather then import it). That should also > > allow sharing of optimalizations. > > There are several stages. As you point out parsing is likely going to > be different, and it will produce the internal AST consumed by the > "resolution" stage of the compiler. > the > Now, there are two issues: > > * The semantics of the same constructs might differ across > compilers. In these cases, you would either create a new > kind of statement that would have the proper semantics. What levels are these exactly, are they documented somewhere? I'm afraid my knowledge of CIL internals is rather limited at the moment.
> * The error codes will likely be different at various stages, > and you want to use the same error codes that MS produces > (readily available documentation, easy to track what is > missing). Right, that makes sense. Cheers, Jelmer -- Jelmer Vernooij <[EMAIL PROTECTED]> - http://samba.org/~jelmer/
signature.asc
Description: This is a digitally signed message part
_______________________________________________ Mono-vb mailing list [email protected] http://lists.ximian.com/mailman/listinfo/mono-vb
