patch for fonts on Mac OS X

[Mojca M. <moj...@gm...>]

Hello,

browsing through my MacPorts (actually, I was unsuccessfully trying to
figure out how to convince Octave to use a different gnuplot binary,
so that I wouldn't have to use X11 for plotting), I figured out that
MacPorts keep patching gnuplot with the following:

http://trac.macports.org/changeset/76048

--- src/variable.c.orig	2010-10-18 15:41:53.000000000 +0200
+++ src/variable.c	2010-10-18 15:42:11.000000000 +0200
@@ -244,13 +244,13 @@
 };
 #endif

-#if defined(_Macintosh) && !defined(FONTPATHSET)
+#if defined(__APPLE__) && !defined(FONTPATHSET)
 #  define FONTPATHSET
 static const struct path_table fontpath_tbl[] =
 {
-    { "/System/Library/Fonts!" },
-    { "/Library/Fonts!" },
-    { "$(HOME)/Library/Fonts!" },
+    { "/System/Library/Fonts" },
+    { "/Library/Fonts" },
+    { "$(HOME)/Library/Fonts" },
     { NULL }
 };
 #endif

I'm not sure how this influences terminals (which terminals are
influenced). If I try Zapfino font, I get mixed results:
- AquaTerm: works properly
- X11: font is change to some random font (but that is almost ok,
there's a chance that it is not easy to configure X11 to use that
font)
- wxt: (process:9969): Pango-WARNING **: couldn't load font "Zapfino
Not-Rotated 200", falling back to "Sans Not-Rotated 200", expect ugly
output.
  This should probably not have happened. The font should work. (But
I'm not sure which libraries are trying to find fonts.) The same
behaviour is observed even if I apply the patch.
- png:
> set term png
Terminal type set to 'png'
Could not find/open font when opening font "arial", using internal
non-scalable font
  (which means that there is something wrong with fonts, but I
remember asking for a patch of gd library, but the library is probably
still at 36 RC1 and even though patched, the patch has never been
released together with a library; maybe I could ask MacPorts team to
patch their library in the same way to see if anything gets improved)
- qt: works, but look awful since the font is not antialiased (that
might be an easy patch - just a boolean value to set somewhere)

About the patch: as already discussed, "defined(_Macintosh)" has zero
influence. (See GP archives "Identifying system with macro":
http://thread.gmane.org/gmane.comp.graphics.gnuplot.devel/9978,
http://thread.gmane.org/gmane.comp.graphics.gnuplot.devel/9988). There
is no _Macintosh defined anywhere.

The three folders are relevant for modern OS, but honestly I don't
know how that code influences any terminal to be able to test if
patching the code actually makes any difference.

If nothing else it would be great in MacPorts wouldn't need the above
mentioned patch any more.



About the remaining "_Macintosh" entries:

1.) in term.h:

/* Apple Macintosh */
#ifdef _Macintosh
# include "mac.trm"
#endif

1.a) Can somebody please explain me where any of the functions like
MAC_text are defined? It seems to me as if term/mac.trm does nothing.
I must be blind, but I don't find any function implemented anywhere.

1.b) Obviously that doesn't hurt anyone since _Macintosh isn't defined
anywhere but on ancient machines. But what's the point in keeping it
if mac.trm is empty anyway? (I suspect that there was some other
mac-specific source somewhere that got lost in years.)

2.) in plot.c:

#if defined(_Windows) || defined(_Macintosh)
int
gnu_main(int argc, char **argv)
#else
int
main(int argc, char **argv)
#endif

I don't understand what this does, but gnu_main is probably not needed
on Apple (modern Mac OS X), so "|| defined(_Macintosh)" may probably
go away if there is no code to support the old macintosh anyway.

3.) in command.c:

    if (sleep_time < 0) {
#if defined(_Windows)
...
#elif defined(_Macintosh)
  if (strcmp(term->name, "macintosh") == 0 && sleep_time < 0)
      Pause( (int)sleep_time );

This also seems to be related to the old _Macintosh. But if there is
no code for the old mac, there is little point in keeping this one as
well.

Mojca