Re: [orbitcpp-list] Comments about o2cpp
Status: Beta
Brought to you by:
philipd
From: Phil d. <ph...@us...> - 2000-02-03 10:52:18
|
Andreas Kloeckner wrote: > > Phil dawes wrote: > > > Hi Phil, > > > ... marshalling techniques... > > Got your point. Cool. For some reason, I was stuck with thinking that > stubs needed some virtual methods, thus must be created by their genuine > constructor, thus requiring stub factory hashing... Yep. Of course the lack of virtual functions means you can't do straight C++ to C++ short circuiting (since the client is statically bound to the marshalling stubs). However the C ORBit runtime has the ability to shortcircuit at the level below if both client and server are in the same address space, so the call never gets marshalled to an actual giop stream. If the speed hit of marshalling exceptions and that extra layer of indirection becomes a problem (I seriously doubt it, but maybe...) then we could add a vtable interface (common client/server abstract base class) on top at a later date. > > Good. Can you recommend a good C++ indent program? (I've seen a couple > > on freshmeat so I might investigate...) > > nope. i'll also see what i can find. > > > > As you may have noticed, the compiler also does some indentation of its > > > output, and of course the output is indented my way. we might have to > > > look into that, or remove indentation entirely. > > > > With the python compiler I didn't bother with indenting, and assumed > > that I could probably just pass the file through indent or something > > after compilation. > > it all depends on whether we can find some good indentation tool. if so, > indentation by the compiler is just useless code clutter, if not, > well... > I'm tempted to leave indentation out. I never had problems with orbit-idl2cpp because I just got emacs to indent the stubs if I wanted to look at what was going on. All this talk of indentation has got me interested in the possibility of writing a CVS extension to do indentation for the user 'on the fly' (during checkouts, updates, diffs etc..)... hmmmm... Certainly would solve a few problems in the free software world. Phil. |