"Andrew Bachmann" <shatty@xxxxxxxxxxxxx> wrote: > On Thu, 31 Jul 2003 15:37:35 +0200 CEST "Ingo Weinhold" < > bonefish@xxxxxxxxxxxxxxx> wrote: > > We can as well build the `normal' version and write the <iconv.h> > > functions as simple wrappers for the GNU ones. > I'm okay with following this one too, especially since I got it > working > "manually". Should I check libiconv into our cvs somehow? Perhaps > in current/src/libs/iconv? The build system is configure/makefile, > should I convert it to jam? Please not yet! As I said, we should have a deeper look at libroots needs first, and decide where to put it when we did that. I don't think there is a point in putting it in now, only to have it somewhere when it's going to be removed or moved soon (you remember, stupid/simple CVS cannot move files around). > Also, are you proposing that we write iconv functions in the > header file that call libiconv functions? I think it might be more > straightforward to just build it in plugin mode. I agree here. And still, the glibc iconv version directly exports the right names. > The only other nit I have left is how to address the prototype issue. > > I can add the gnu iconv.h to headers/libs/iconv and build only > libiconv using it. (gnu iconv.h has const and other messiness) > If the other apps use the public posix/iconv.h header > then the function will still presumably link. (right?) > > As far as I can tell the posix standard is to not use const: > http://www.opengroup.org/onlinepubs/007904975/toc.htm OTOH if both, Linux and the BSD world is using "const", I see little or no problems in using "const" as well. But of course, in order to be fully Posix compatible, we would have to ditch it. Adios... Axel.