Re: Porting Questions

["David A. Greene" <greened at eecs dot umich dot edu>]

Phil Edwards wrote:


> For C headers:  Gaby is correct; forget about c_shadow.  The different
> schemes are different approximations of an ideal <c...> header
> implementation, given only a random underlying C library.[*]  The three
> current choices are:
> 
>  - 'c' which just passes things along unchanged (<cFOO> becomes <FOO.h>)
>  - 'c_std' which passes things along unchanged (except that macros are
>    undefined so that functions take their place) and also places the
>    global symbols into std::.
>  - 'c_shadow' which doesn't allow anything to pass through into global
>    scope, and only places into std:: the symbols which the standard says
>    should be there.  This is the ideal.  Also, it doesn't work.  (The
>    current implementation, that is, not the concept.)

So, if I've got you correct, c_shadow is the standard-compliant
version, c_std is a pretty good approximation and c is something
that Just Works.  Reasonable.  :)

> Remedies for c_shadow as well as complete replacements have all been
> proposed; it's a difficult problem that's going to entail a lot of work.

I don't doubt it.

> For --enable-cstdio[=LIB], I suppose that
> http://gcc.gnu.org/onlinedocs/libstdc++/configopts.html is a bit scarce
> on the details.  I'll write a little note and link to it from that page.
> The onlinedocs get updated every 24 hours, so look at the bottom of the
> page, and if it's still revision 1.10, then it hasn't updated yet.

Great.  Thanks, Phil!

> [*]  This situation seems designed to snatch defeat from the jaws of victory,
> but since we don't also ship a C library alongside the C++ library, it's
> all that we can do.  There's lots of good discussion in the archives that
> bears reading.

I'll take a look.  Thanks again for your guidance.

-Dave