Re: [PATCH] BeOS Patch #3

[Owen Taylor <otaylor redhat com>]

"David Reid" <dreid jetnet co uk> writes:

> Guys,
> 
> Here's the third patch for BeOS.  I hope I've answered the problems that
> were outlined by Tim before the New Year.  As always let me know if this
> isn't OK and why it's not OK...

> Index: configure.in
> ===================================================================
> RCS file: /cvs/gnome/glib/configure.in,v
> retrieving revision 1.164
> diff -u -r1.164 configure.in
> --- configure.in	2000/12/22 13:44:22	1.164
> +++ configure.in	2001/01/02 02:25:40
> @@ -78,16 +78,19 @@
>  
>  AC_CANONICAL_HOST
>  
> +G_BEOS_NEEDED_LDFLAGS=
>  AC_MSG_CHECKING(for the BeOS)
>  case $host in
>    *-*-beos*)
>      glib_native_beos="yes"
> +    G_BEOS_NEEDED_LDFLAGS=../libglib-$GLIB_MAJOR_VERSION.$GLIB_MINOR_VERSION.la
>      ;;

I don't want to apply this change to CVS in this form. It doesn't make
any sense without a version of libtool newer than what we include
(such as a CVS snapshot), and if you do have such a newer libtool,
then it doesn't make any sense to make this conditional on BeOS.

Can you just keep this (in unconditional form) as a local change for
the moment? If necessary, we could conceivably switch to using CVS
libtool for GTK+, but as you might imagine, that makes me a little
nervous.

>          AC_DEFINE(HAVE_SYS_SELECT_H,1,[found fd_set in sys/select.h])
>          AC_MSG_RESULT([yes, found in sys/select.h])
>      else
> -	AC_DEFINE(NO_FD_SET,1,[didn't find fd_set])
> -	AC_MSG_RESULT(no)
> +        AC_HEADER_EGREP(fd_set, sys/socket.h, gtk_ok=yes)
> +        if test $gtk_ok = yes; then
> +            # This is for BeOS
> +            AC_DEFINE(FD_SET_IN_SYS_SOCKET_H,1,[found fd_set in sys/socket.h])
> +            AC_MSG_RESULT([yes, found in sys/socket.h])
> +        else
> +            AC_DEFINE(NO_FD_SET,1,[didn't find fd_set])
> +            AC_MSG_RESULT(no)
> +        fi
>      fi
>  fi
> +AC_SEARCH_LIBS(select, socket bind)

As I think I said earlier, we don't use LIBS, and certainly this can't
be right since these libraries, if needed, should be included in
the output of glib-config / pkg-config...
  
> Index: gmodule/gmodule-beos.c
> ===================================================================
> RCS file: /cvs/gnome/glib/gmodule/gmodule-beos.c,v
> retrieving revision 1.3
> diff -u -r1.3 gmodule-beos.c
> --- gmodule/gmodule-beos.c	2000/07/26 11:02:01	1.3
> +++ gmodule/gmodule-beos.c	2001/01/02 02:26:08
> @@ -24,10 +24,12 @@
>   * MT safe
>   */
>  
> -#include <be/kernel/image.h> /* image (aka DSO) handling functions... */
> +#undef MIN
> +#undef MAX
> +#include <kernel/image.h> /* image (aka DSO) handling functions... */

If you include system headers before glib/gtk+ headers, you can
avoid having to #undef MIN/MIX.
  
>  /*
> - * The BeOS doesn't use the same symantics as Unix's dlopen....
> + * BeOS doesn't use the same symantics as Unix's dlopen....

That's 'semantics'.

Except for the above comments, this looks OK.

Regards,
                                        Owen