Mail Archives: djgpp-workers/2001/02/09/13:05:23
Laurynas Biveinis wrote:
> Current CVS has libstdc++-v3/include/c_std/bits directory, full of
> namespace std wrappers for standard headers.
These don't really do what is required. On the other hand, in
libstdc++-v3/include/c_shadow and libstdc++-v3/include/c_shadow/bits
there are some quite impressive wrappers. As far as I can see, they
are going to break anything that uses non-standard functions from the
standard headers, but I could probably live with that.
> > But I don't see how the problem can be solved satisfactorily from
> > the GCC side, whereas I have *already* solved it (at least to my
> > satisfaction) from the DJGPP side.
>
> I'm not saying your work is obsolete; what I'm saying is that your
> solution should work together with libstdc++ V3.
I don't care whether or not my work is used, as long as the C library
somehow handles namespace std correctly. My solution was designed not
to rely on anything from libstdc++ - only trivial <c*> headers are
required, and these can easily be supplied if the libstdc++ ones try
to be too clever. In that sense, they should work with libstdc++ V3.
But the libstdc++ V3 wrappers assume that the .h headers are not using
namespace std, so they won't work with my patches. These wrappers may
not be acceptable as a default if they break a lot of existing code,
but it wouldn't be difficult to modify my patches in order to allow
them to be used if desired. I'll post a separate message outlining a
method for doing this.
Stephen
- Raw text -