On Fri, Nov 15, 2013 at 1:08 PM, Stephen J. Turnbull <stephen(a)xemacs.org>wrote:
> I've flipped the configure switch to pick GTK3 first, or fallback to
> GTK2. I hope nobody wants to use GTK 2 even if they have 3..
Let me try that ...
Urk. It looks like HAVE_GTK2 and HAVE_GTK3 are supposed to be mutually
exclusive, but configure.ac seems to always define HAVE_GTK2. Fixing ...
Yes, commenting out the second AC_DEFINE(HAVE_GTK2) fixes the build.
Unfortunately, it doesn't fix the crash in g_type_fundamental(). I
bet Stallman added an assert(os_type != DARWIN). :-P
I'll probably get a chance to try this on my ancient Mac soon. Maybe
crash in the same place and I can figure out a solution. The mac isn't
really a priority for Gtk, right?
To avoid this kind of duplicated but slightly wrong Issueode , you
could #define GTK_VERSION (or perhaps GTK_MAJOR_VERSION) to 2 or 3 as
discovered, and use #if (GTK_VERSION == 2) etc as appropriate.
Alternatively use #if defined(HAVE_GTK3) ... #elif defined(HAVE_GTK3).
I didn't want to use #elif so that the HAVE_GTK2 chunks could be ripped out
I anticipated bigger changes between 2 and 3. Alternatively, we could use
GTK_CHECK_VERSION and keep config.h out of it.
The clang 3.4 compiler issues a lot of warnings about unused and
uninitialized variables and functions. Are you aware of these issues?
Yes. Some code is disabled, and doesn't call those functions yet. Now
that Gtk 3 is mostly working, I'll be getting to that.
Aha, the test quite breaks because the fontconfig functions (fc-*)
aren't built. Before we try to fix that ... as I wrote earlier, I
think fontconfig should be our primary interface for describing fonts,
with translations from other font descriptors (eg, legacy XLFD and
"short names") written in Lisp. What's your current thinking about
I haven't thought forward yet. I'm happy to get a single font to load and
Pango works with fontconfig, so I'd be happy to use it.
XEmacs-Beta mailing list