Re: Problem compiling from cvs

[Hans Breuer <hans breuer org>]

On 05.01.2007 13:58, Lars Clausen wrote:
> On Wed, 2007-01-03 at 14:20 +0100, Hans Breuer wrote:
[...]
> > Unfortuantely I don't know what problem was supposed to be fixed by that.
> > The issue already is reported as
> > http://bugzilla.gnome.org/show_bug.cgi?id=385701

> Lode wrote:
See: http://mail.gnome.org/archives/dia-list/2005-April/msg00115.html

> > The use of Makefile variable DATADIR collides with a datatype in
> > w32api:
> >
> > /usr/include/w32api/objidl.h:58
> > typedef OLECHAR **SNB;
> > typedef enum tagDATADIR {
> >         DATADIR_GET=1,
> >         DATADIR_SET
> > } DATADIR;
> >
> >
> > Could DIA be adapted to use DIA_DATADIR (like DIA_SHEETDIR)?
> > (And while you're at it, LIBDIR -> DIA_LIBDIR)
>
> That's why.
Sounds like the GNOME build got broken for some obscure build issue with
windows headers. To me this looks like the wrong fix (sorry for not
noticing earlier)

- on win32 builds DATADIR should not be defined at all, cause all pathes
  need to be deduced at runtime (see lib/dia_dir.c
- the 'official' (msvc) win32 build does not define DATADIR
- IIRC the inclusion of objidl.h can be avoided by defining
  WIN32_LEAN_AND_MEAN before inclusion of windows (this is not yet done
  consitently)

Without being able to compile Dia on win32 with mingw I'm not sure this
will completely fix the issue. If not either some auto-foo would be needed
or a simple #undef DATADIR before the offending windows header (at least
app/dia-win32-remote.c ?)

Thanks,
        Hans

-------- Hans "at" Breuer "dot" Org -----------
Tell me what you need, and I'll tell you how to
get along without it.                -- Dilbert