Re: [PATCH 0/6] Convert gimple to a C++ class hierarchy

[Jan Hubicka <hubicka at ucw dot cz>]

> On Mon, 2013-09-02 at 14:35 +0200, Martin Jambor wrote:
> > Hi,
> > 
> > On Fri, Aug 30, 2013 at 03:21:22PM -0400, David Malcolm wrote:
> > > Apart from the GTY aspect, how do people feel about the patch series?
> > > FWIW I have vague thoughts about doing something similar for tree -
> > > doing so *might* give an easier route to the type vs expression
> > > separation that Andrew spoke about at the Cauldron rearchitecture BoF.
> > > (I started looking at doing a similar C++-ification of rtx, but...
> > > gahhhhh)
> > > 
> > 
> > I like it but before you start looking at the biger things, could you
> > perhpas proceed with the symtab?  It has much fewer classes, will
> > probably affect private development of fewer people, the accessor
> > macros/functions of symtab are less developed so it will immediately
> > really make code nicer, Honza has approved it and I'm really looking
> > forward to it.  Also, perhaps it will show us at much saller scale
> > potential problems with the general scheme.
> 
> Sorry about the delay.  I wasn't aware that it had been approved; there
> seemed to be a lot of caveats and objections in that thread.  On
> re-reading, http://gcc.gnu.org/ml/gcc-patches/2013-08/msg01147.html
> could be seen as approval, but I guess I was making a conservative
> reading of that post.  I hope to refresh the patches and
> reboostrap/repost them at some point this week.

I think the patch was generally in right direction.  I would welcome
if you got rid of symtab_node_base (and made symtab_node the basetype)
and possibly also did the suggested grand renaming.  But the second
can be probably done incrementally.
> 
> > I'm only writing this because the development there seems a bit
> > stalled and it it a shame.  Of course, you ay want to simplify the
> > manual markings first.  I'd perfectly understand that.
> 
> I've been poking at gengtype (and running benchmarks; see other post),
> which would affect the symtab patch, though it's something of a
> quagmire...

Making gengtype to generate ggc_mark for each type would make hand writting
easier - you can use C++ overloading instead of trying to guess the funny
names gengtype uses right now.
But that is independent of this change.  I am slowly getting used to the
world of hand written gengtype markings.

Honza