Re: C++ header compatibility patches

["J. Johnston" <jjohnstn at redhat dot com>]

Benjamin Kosnik wrote:
> 
> > AFAICT you have hard-wired "on" the namespace code instead of "off".
> > Am I missing something?
> 
> You are correct. It should be hard-wired to "off" before check-in. (#if
> 0 should be #if 1, or whatever.)

I imagine you might want to key this off a new define.  In the future,
we could trigger this define via configuration (either config.h or
via a dynamic header file).

> 
> > There are a number of header files that use the extern "C" mechanism
> > that you have chosen not to modify.  Is this because you are only
> > changing ANSI C header files?
> 
> Yes. I'm trying to do the smallest possible change.
>

Ok.
 
> > For example, you have not changed any of
> > the libc/sys include files and there are some header files like
> > malloc.h which you do not change in the main include directory. As
> > well, you have moved some of the _BEGIN_STD_C statements from the
> > location their corresponding extern "C" statements were.
> 
> extern "C" { nests, but
> 
> namespace std { extern "C" {
> 
> does not. I tried to make the movement as small as possible.
> 
> >  If my
> > assumption is correct and you only care about ANSI C variables and
> > prototypes, that would explain it.  Would there be a reason to still
> > have the extern "C" statements covering the non-ANSI bits?
> 
> Yes. For C++ people who use the "C" headers. (Some, like pthreads.h, are
> almost always used, for instance.)
> 

If you change the patch with regards to the setting of _BEGIN_STD_C and _END_STD_C
and verify that newlib builds with the patch, you may apply it or resubmit it here and I
will apply it for you.

-- Jeff J.