Re: [PATCH] Vtable pointer verification, runtime library changes (patch 3 of 3)

[Benjamin De Kosnik <bkoz at redhat dot com>]

> You're welcome.  Does this mean that once GCC 4.9 opens up, these
> changes are approved (with the Makefile.am change mentioned above)?

I don't see anything wrong with your approach or design from my end, but
I haven't gotten testresults yet. I hand-applied your patches (went
fine) but I got a build error when I checked back later:

In file included from /home/bkoz/src/gcc.git-trunk/gcc/tree-flow.h:28:0,
                 from /home/bkoz/src/gcc.git-trunk/gcc/tree-vtable-verify.c:145:
/home/bkoz/src/gcc.git-trunk/gcc/gimple.h: In function âunsigned int
vtable_verify_main()â:
/home/bkoz/src/gcc.git-trunk/gcc/gimple.h:1146:20: error: âdef_stmtâ
may be used uninitialized in this function [-Werror=maybe-uninitialized]
   return g->gsbase.code;
                    ^
/home/bkoz/src/gcc.git-trunk/gcc/tree-vtable-verify.c:754:10: note:
âdef_stmtâ was declared here
   gimple def_stmt;
          ^
cc1plus: all warnings being treated as errors


I'll probably just start from scratch on your git branch? Or try again
when I'm paying full attention later in the week when everybody is at
FOSDEM.

There are some things I'm curious about, and want to examine further.
But don't expect trouble from me. I think this looks pretty good.

> > Are you planning on submitting some internals docs, or some kind of
> > end-user description of this feature, perhaps with a NEWS update? I
> > think some kind of overview for the internals docs is entirely
> > appropriate.
> >
> 
> I haven't given it a lot of thought.  This is the first time I've ever
> added such a major new feature, so I'm unsure about the right way to
> document it, and I'd appreciate guidance on that.  Exactly what do you
> think I ought to put where?

Here's a list of things to think about:

1) NEWS, see Gerald P. This is definitely NEWS-worthy.
2) http://gcc.gnu.org/gcc-4.9/changes.html, C++ entry?
3) maybe recast the cauldron talk into something for gccint.texi? or is
this being documented as part of the C++ Runtime ABI? Hmmm.

-benjamin