Re: [Introspector-developers] current status

[James M. D. <mdu...@ya...>]

--- "cr...@is..." <cr...@is...> wrote:
> On Sat, 21 Dec 2002, James Michael DuPont wrote:
> 
> > hey craig,
> > thanks for writing. Come and visit us on the #introspector
> > irc.openprojects.net for a chat.
> 
> Email is my preferred communication medium, but I'll consider getting
> an
> IRC client (folks on the schmoo and a couple other lists wanted to
> chat
> via IRC as well; I guess I'm just not the chatty type :-/).

:) thats ok. 

> 
> Note that I'm going to be celebrating Yule/Winter Solstice for most
> of
> this weekend.  After that I should have some time over the holidays.

sure, i will also be offline next week. Happy holidays!


> 
> > > Hey -
> > >
> > > I've been looking at this and gccxml. It's a little hard for me
> to
> > > figure out what the current status of this project is. Will this
> work
> > > on win32?
> 
> > yes it will, with cygwin, i have been testing with the 3.2, but the
> > berkley db support of redland is not dont on win32.
> 
> I don't know much about redland or RDF (tho I know a bit about XML
> Schema).
The redland is a application framework, We replaced all the printf with
it.

> Can we use mysql rather than berkely db, or is redland
> itself not
> ported to win32? 
I have ported it without any real problems, have also tested the patch
on cygwin, and debian

> I did not see cygwin listed on redland's site.
> 
> Do you currently serialize the parse tree in binary format, or do you
> only
> write it out to xml/rdf? 
Redland supports a binary format, using bdb, and also other storages. 
There is also the tree serializer branch of the gcc.

> Does it make sense to use binary format? 
> Was
> part
> of the vision to allow arbitrary modification of the parse tree
> during
> compilation, and if so, would that require converting the xml/rdf
> back to
> binary format for gcc to process? 
good question. Yes we want to support that, a two way interface between
redland and the gcc. I figure we can store a native pointer to the node
int the redland object, and then have an application layer that can
call methods on a redland node that is a proxy to a gcc. 

As far as reflection, there are some reflection libs for c++,
eg http://www.garret.ru/~knizhnik/cppreflection/docs/reflect.html
but they still need data input. I figure we need to add in some
built-in functions into gcc that can give you access to the trees. 
like the typeof operator, or a class lib. 
You might want to take a look at an older effort of mine to wrap the
trees in a c++ wrapper, it just might be what you want , we can make a
redland feed into that : 
http://prdownloads.sourceforge.net/introspector/itrspct-clunky-old.tgz


>Sorry if these are naive/obvious
> questions...
very good questions. We are testing the rdf using the perl interface
now, outside of the compiler. 
> 
> > > I started to look into this option (a couple years ago now), RMS
> sent
> > > me a
> > > 'cease and desist letter', which kinda took the wind out of my
> sails.
> > well i have had lots of interactions with RMS, he is just trying to
> > play politics. Most of the gcc guys dont like the idea of making
> > interfaces, but they will like it when we are done.
> >
> > > I know this has been a long-standing debate, but I never got any
> clear
> > > answers except for 'RMS don't like it, and I'm tired of arguing
> with
> > > him about it', and RMS replying 'I don't have time to go into
> details,
> > > I just don't like it'. Oh well. It finally appears that there's
> enough
> > > momentum for this to finally get somewhere. :-)
> 
> > yes, please do help. get yourself a Sourceforge account and mail it
> to
> > me.
> 
> cperras is my sourceforge account. I just read from Redland's site
> that
> sourceforge has a compiler farm. That's really cool. I'll have to
> check
> that out.

adding you now. welcome aboard!

> 
> > > I've looked into several related projects, including OpenC++,
> gccxml,
> > > and
> >
> > > PDT (http://www.cs.uoregon.edu/research/paracomp/pdtoolkit/), and
> > I might have seen this, but it is not free.
> 
> Yes, the license seemed to indicate it was, but the EDG front-end is
> not
> freely redistributable.

We have a project member that uses the EDG front end, djbarker,
maybe he can tell us some of his experince with EDG?

> > > Root
> > > (http://root.cern.ch/).
> > that is new
> 
> It's pretty interesting. Like most of these projects, however,
> there's a
> number of restrictions placed on the C++ code.
> 
> > >  I'd be happy to try to help out on this project. Some things I
> have
> > > potentially relevant knowledge of include:
> > >
> > > c, c++, gccxml, openc++ (tho it's been a little while), pdt, xml,
> > > mysql/odbc, boost. I'm not very skilled with Perl or Bison/Yacc.
> My
> > > main linux dev box has been having hardware problems, and also
> for a
> > > few other reasons, win32 is my main dev platform right now.
> >
> > ok, then you might want to help with the testing of the debian
> packaging
> > of the gui elements for mingw32/cygwin. Mike Garnsey and I have
> been
> > working on porting gtk to windows via the deb packaging system for
> the
> > gui.
> 
> I really want to focus on C++ reflection right now, and get that
> working
> on win32 if it's not already. 
that is not working yet.

>I'm not sure how the gui will help with
> that.

for visualization, for uml diagramming

> OTOH, I've been looking into porting gtk/gtkmm to win32 off and
> on.

ok, i am just finishing up the win32 port of gtk, we have %90 done

> If I have extra time, I can help with this effort. What kind of
> testing do
> you need?
> 
> > > One thing I noticed is that, for some reason, gcc doesn't keep
> track
> > > of the column number while it's parsing (well, the tree node
> doesn't
> > > keep track anyway). I've considered amending this; I don't know
> if
> > > this would be useful to introspector tho.
> >
> > you can turn that on in bison
> 
> Does it require modification of the parse tree to actually store the
> column? Otherhwise, unless there's a parse error, I think it's just
> ignored.

There is a bison flag for that, lemme google
http://www.informatik.uni-hamburg.de/RZ/software/gnu/gcc/bison_7.html#SEC73

Best regards,
mike

=====
James Michael DuPont
http://introspector.sourceforge.net/

__________________________________________________
Do you Yahoo!?
Yahoo! Mail Plus - Powerful. Affordable. Sign up now.
http://mailplus.yahoo.com