Fontset problem

Fontset problem

[Andreas Schwab]

I'd like to define a fontset that uses courier as the main font, but
substitutes fixed for the three mule-unicode coding systems (because they
are much more complete).  Thus I defined the following fontset:

Emacs.Fontset-0: -*-courier-medium-r-normal--*-140-*-*-*-*-fontset-startup,\
		 mule-unicode-2500-33ff:-*-fixed-medium-r-normal--*-140-*-*-*-*-iso10646-1,\
		 mule-unicode-e000-ffff:-*-fixed-medium-r-normal--*-140-*-*-*-*-iso10646-1,\
		 mule-unicode-0100-24ff:-*-fixed-medium-r-normal--*-140-*-*-*-*-iso10646-1

The problem is that characters in the mule-unicode coding systems are
always using medium-r-normal style, even if they are supposed to be, say,
bold.  On the other hand, if I define the fontset in this way:

Emacs.Fontset-0: -*-courier-medium-r-normal--*-140-*-*-*-*-fontset-startup,\
		 mule-unicode-2500-33ff:-*-fixed-*-*-*--*-140-*-*-*-*-iso10646-1,\
		 mule-unicode-e000-ffff:-*-fixed-*-*-*--*-140-*-*-*-*-iso10646-1,\
		 mule-unicode-0100-24ff:-*-fixed-*-*-*--*-140-*-*-*-*-iso10646-1

Emacs uses the first font that matches
-*-fixed-*-*-*--*-140-*-*-*-*-iso10646-1, which may not be the
medium-r-normal style.  What is the right way to define the fontset so
that Emacs uses the right style?

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux AG, Deutschherrnstr. 15-19, D-90429 Nürnberg
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: Fontset problem

[Kenichi Handa]

In article <jer8gb192m.fsf@sykes.suse.de>, Andreas Schwab <schwab@suse.de> writes:
> I'd like to define a fontset that uses courier as the main font, but
> substitutes fixed for the three mule-unicode coding systems (because they
> are much more complete).  Thus I defined the following fontset:

> Emacs.Fontset-0: -*-courier-medium-r-normal--*-140-*-*-*-*-fontset-startup,\
> 		 mule-unicode-2500-33ff:-*-fixed-medium-r-normal--*-140-*-*-*-*-iso10646-1,\
> 		 mule-unicode-e000-ffff:-*-fixed-medium-r-normal--*-140-*-*-*-*-iso10646-1,\
> 		 mule-unicode-0100-24ff:-*-fixed-medium-r-normal--*-140-*-*-*-*-iso10646-1

> The problem is that characters in the mule-unicode coding systems are
> always using medium-r-normal style, even if they are supposed to be, say,
> bold.  On the other hand, if I define the fontset in this way:

> Emacs.Fontset-0: -*-courier-medium-r-normal--*-140-*-*-*-*-fontset-startup,\
> 		 mule-unicode-2500-33ff:-*-fixed-*-*-*--*-140-*-*-*-*-iso10646-1,\
> 		 mule-unicode-e000-ffff:-*-fixed-*-*-*--*-140-*-*-*-*-iso10646-1,\
> 		 mule-unicode-0100-24ff:-*-fixed-*-*-*--*-140-*-*-*-*-iso10646-1

> Emacs uses the first font that matches
> -*-fixed-*-*-*--*-140-*-*-*-*-iso10646-1, which may not be the
> medium-r-normal style.  What is the right way to define the fontset so
> that Emacs uses the right style?

If you specify an XLFD field other than FOUNDRY, FAMILY,
CHARSET_REGISTRY, CHARSET_ENCODING in each font name, Emacs
tries to use the specified font name as is (i.e. the same
font X server will find from that name).  In your case, you
specify ADSTYLE as NULL and POINT_SIZE as 140.  If you put
`*' in those fields, Emacs's normal font selection mechanism
should work also for those charsets.

So, please try this:

Emacs.Fontset-0: -*-courier-medium-r-normal--*-140-*-*-*-*-fontset-startup,\
	mule-unicode-2500-33ff:-*-fixed-*-*-*-*-*-*-*-*-*-*-iso10646-1,\
	mule-unicode-e000-ffff:-*-fixed-*-*-*-*-*-*-*-*-*-*-iso10646-1,\
	mule-unicode-0100-24ff:-*-fixed-*-*-*-*-*-*-*-*-*-*-iso10646-1

---
Ken'ichi HANDA
handa@etl.go.jp

Re: Fontset problem

[Andreas Schwab]

Kenichi Handa <handa@etl.go.jp> writes:

|> So, please try this:
|> 
|> Emacs.Fontset-0: -*-courier-medium-r-normal--*-140-*-*-*-*-fontset-startup,\
|> 	mule-unicode-2500-33ff:-*-fixed-*-*-*-*-*-*-*-*-*-*-iso10646-1,\
|> 	mule-unicode-e000-ffff:-*-fixed-*-*-*-*-*-*-*-*-*-*-iso10646-1,\
|> 	mule-unicode-0100-24ff:-*-fixed-*-*-*-*-*-*-*-*-*-*-iso10646-1

That does not work either, now Emacs is using courier instead of fixed for
the mule-unicode charsets.  I've also tried -*-fixed-*-iso10646-1, but
that gave me the original problem back, which is kind of strange, because
describe-fontset is displaying this pattern when I specify the pattern
fully spelled out as above.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux AG, Deutschherrnstr. 15-19, D-90429 Nürnberg
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: Fontset problem

[Kenichi Handa]

Sorry for the late response.

In article <je4rd3qllv.fsf@sykes.suse.de>, Andreas Schwab <schwab@suse.de> writes:

> Kenichi Handa <handa@etl.go.jp> writes:
> |> So, please try this:
> |> 
> |> Emacs.Fontset-0: -*-courier-medium-r-normal--*-140-*-*-*-*-fontset-startup,\
> |> 	mule-unicode-2500-33ff:-*-fixed-*-*-*-*-*-*-*-*-*-*-iso10646-1,\
> |> 	mule-unicode-e000-ffff:-*-fixed-*-*-*-*-*-*-*-*-*-*-iso10646-1,\
> |> 	mule-unicode-0100-24ff:-*-fixed-*-*-*-*-*-*-*-*-*-*-iso10646-1

> That does not work either, now Emacs is using courier instead of fixed for
> the mule-unicode charsets.

Ah, I see.  You actually have courier iso10646-1 fonts.

Hmmm, currently, Emacs tries at first the family specified
by face, and if you have that font, Emacs uses it.

Perhaps, for finding a font or non-ASCII, non-Latin-1 chars,
Emacs should prefer the family specified in the fontset to
the family specified in the face.

I've just installed this change.  Please try it.  At least
it works for me with the above resource on the machine that
has courier iso10646-1 fonts.  Other people please check if
it doesn't break your font/fontset setting.  Font selection
routine is one of very delicate parts.

---
Ken'ichi HANDA
handa@m17n.org

2002-09-26  Kenichi Handa  <handa@etl.go.jp>

	* xfaces.c (try_font_list): New arg PREFER_FACE_FAMILY.  If it is
	nonzero, try face's family at first.  Otherwise try FAMILY at
	first.
	(choose_face_font): If C is a single byte char or latin-1, call
	try_font_list with PREFER_FACE_FAMILY 1.

Index: xfaces.c
===================================================================
RCS file: /cvs/emacs/src/xfaces.c,v
retrieving revision 1.262
retrieving revision 1.263
diff -u -c -r1.262 -r1.263
cvs server: conflicting specifications of output style
*** xfaces.c	13 Sep 2002 19:35:08 -0000	1.262
--- xfaces.c	26 Sep 2002 08:03:52 -0000	1.263
***************
*** 510,516 ****
  static int font_list P_ ((struct frame *, Lisp_Object, Lisp_Object,
  			  Lisp_Object, struct font_name **));
  static int try_font_list P_ ((struct frame *, Lisp_Object *, 
! 			      Lisp_Object, Lisp_Object, struct font_name **));
  static int try_alternative_families P_ ((struct frame *f, Lisp_Object,
  					 Lisp_Object, struct font_name **));
  static int cmp_font_names P_ ((const void *, const void *));
--- 510,517 ----
  static int font_list P_ ((struct frame *, Lisp_Object, Lisp_Object,
  			  Lisp_Object, struct font_name **));
  static int try_font_list P_ ((struct frame *, Lisp_Object *, 
! 			      Lisp_Object, Lisp_Object, struct font_name **,
! 			      int));
  static int try_alternative_families P_ ((struct frame *f, Lisp_Object,
  					 Lisp_Object, struct font_name **));
  static int cmp_font_names P_ ((const void *, const void *));
***************
*** 6324,6344 ****
     match.  A value of nil means include fonts of any registry and
     encoding.
     
     Return in *FONTS a pointer to a vector of font_name structures for
     the fonts matched.  Value is the number of fonts found.  */
  
  static int
! try_font_list (f, attrs, family, registry, fonts)
       struct frame *f;
       Lisp_Object *attrs;
       Lisp_Object family, registry;
       struct font_name **fonts;
  {
    int nfonts = 0;
    Lisp_Object face_family = attrs[LFACE_FAMILY_INDEX];
  
!   if (STRINGP (face_family))
!     nfonts = try_alternative_families (f, face_family, registry, fonts);
  
  #ifdef MAC_OS
    /* When realizing the default face and a font spec does not matched
--- 6325,6352 ----
     match.  A value of nil means include fonts of any registry and
     encoding.
     
+    If PREFER_FACE_FAMILY is nonzero, perfer face's family to FAMILY.
+    Otherwise, prefer FAMILY.
+ 
     Return in *FONTS a pointer to a vector of font_name structures for
     the fonts matched.  Value is the number of fonts found.  */
  
  static int
! try_font_list (f, attrs, family, registry, fonts, prefer_face_family)
       struct frame *f;
       Lisp_Object *attrs;
       Lisp_Object family, registry;
       struct font_name **fonts;
+      int prefer_face_family;
  {
    int nfonts = 0;
    Lisp_Object face_family = attrs[LFACE_FAMILY_INDEX];
+   Lisp_Object try_family;
+ 
+   try_family = prefer_face_family ? face_family : family;
  
!   if (STRINGP (try_family))
!     nfonts = try_alternative_families (f, try_family, registry, fonts);
  
  #ifdef MAC_OS
    /* When realizing the default face and a font spec does not matched
***************
*** 6346,6357 ****
       default font.  On the Mac, this is mac-roman, which does not work
       if the family is -etl-fixed, e.g.  The following widens the
       choices and fixes that problem.  */
!   if (nfonts == 0 && STRINGP (face_family) && STRINGP (registry)
        && xstricmp (SDATA (registry), "mac-roman") == 0)
!     nfonts = try_alternative_families (f, face_family, Qnil, fonts);
  #endif
  
!   if (nfonts == 0 && !NILP (family))
      nfonts = try_alternative_families (f, family, registry, fonts);
  
    /* Try font family of the default face or "fixed".  */
--- 6354,6368 ----
       default font.  On the Mac, this is mac-roman, which does not work
       if the family is -etl-fixed, e.g.  The following widens the
       choices and fixes that problem.  */
!   if (nfonts == 0 && STRINGP (try_family) && STRINGP (registry)
        && xstricmp (SDATA (registry), "mac-roman") == 0)
!     nfonts = try_alternative_families (f, try_family, Qnil, fonts);
  #endif
  
!   if (! prefer_face_family)
!     family = face_family;
! 
!   if (nfonts == 0 && STRINGP (family))
      nfonts = try_alternative_families (f, family, registry, fonts);
  
    /* Try font family of the default face or "fixed".  */
***************
*** 6425,6431 ****
  
    /* Get a list of fonts matching that pattern and choose the
       best match for the specified face attributes from it.  */
!   nfonts = try_font_list (f, attrs, XCAR (pattern), XCDR (pattern), &fonts);
    width_ratio = (SINGLE_BYTE_CHAR_P (c)
  		 ? 1
  		 : CHARSET_WIDTH (CHAR_CHARSET (c)));
--- 6436,6444 ----
  
    /* Get a list of fonts matching that pattern and choose the
       best match for the specified face attributes from it.  */
!   nfonts = try_font_list (f, attrs, XCAR (pattern), XCDR (pattern), &fonts,
! 			  (SINGLE_BYTE_CHAR_P (c)
! 			   || CHAR_CHARSET (c) == charset_latin_iso8859_1));
    width_ratio = (SINGLE_BYTE_CHAR_P (c)
  		 ? 1
  		 : CHARSET_WIDTH (CHAR_CHARSET (c)));

Re: Fontset problem

[Andreas Schwab]

Kenichi Handa <handa@m17n.org> writes:

|> Sorry for the late response.
|> 
|> In article <je4rd3qllv.fsf@sykes.suse.de>, Andreas Schwab <schwab@suse.de> writes:
|> 
|> > Kenichi Handa <handa@etl.go.jp> writes:
|> > |> So, please try this:
|> > |> 
|> > |> Emacs.Fontset-0: -*-courier-medium-r-normal--*-140-*-*-*-*-fontset-startup,\
|> > |> 	mule-unicode-2500-33ff:-*-fixed-*-*-*-*-*-*-*-*-*-*-iso10646-1,\
|> > |> 	mule-unicode-e000-ffff:-*-fixed-*-*-*-*-*-*-*-*-*-*-iso10646-1,\
|> > |> 	mule-unicode-0100-24ff:-*-fixed-*-*-*-*-*-*-*-*-*-*-iso10646-1
|> 
|> > That does not work either, now Emacs is using courier instead of fixed for
|> > the mule-unicode charsets.
|> 
|> Ah, I see.  You actually have courier iso10646-1 fonts.
|> 
|> Hmmm, currently, Emacs tries at first the family specified
|> by face, and if you have that font, Emacs uses it.
|> 
|> Perhaps, for finding a font or non-ASCII, non-Latin-1 chars,
|> Emacs should prefer the family specified in the fontset to
|> the family specified in the face.
|> 
|> I've just installed this change.  Please try it.

Thanks, it fixed the problem for me.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux AG, Deutschherrnstr. 15-19, D-90429 Nürnberg
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: Fontset problem

[Stefan Monnier]

> Sorry for the late response.
> 
> In article <je4rd3qllv.fsf@sykes.suse.de>, Andreas Schwab <schwab@suse.de> writes:
> 
> > Kenichi Handa <handa@etl.go.jp> writes:
> > |> So, please try this:
> > |> 
> > |> Emacs.Fontset-0: -*-courier-medium-r-normal--*-140-*-*-*-*-fontset-startup,\
> > |> 	mule-unicode-2500-33ff:-*-fixed-*-*-*-*-*-*-*-*-*-*-iso10646-1,\
> > |> 	mule-unicode-e000-ffff:-*-fixed-*-*-*-*-*-*-*-*-*-*-iso10646-1,\
> > |> 	mule-unicode-0100-24ff:-*-fixed-*-*-*-*-*-*-*-*-*-*-iso10646-1
> 
> > That does not work either, now Emacs is using courier instead of fixed for
> > the mule-unicode charsets.
> 
> Ah, I see.  You actually have courier iso10646-1 fonts.
> 
> Hmmm, currently, Emacs tries at first the family specified
> by face, and if you have that font, Emacs uses it.
> 
> Perhaps, for finding a font or non-ASCII, non-Latin-1 chars,
> Emacs should prefer the family specified in the fontset to
> the family specified in the face.

I haven't seen the rest of the thread, so I don't know what problem
you're trying to solve, but this seems wrong to me.
If the face says "courier" then Emacs should use courier if it's available.
The fontset's family should only be obeyed for faces that do not set the
`family' property or when the face family cannot be found with the
desired encoding.


	Stefan

Re: Fontset problem

[Andreas Schwab]

"Stefan Monnier" <monnier+gnu/emacs@rum.cs.yale.edu> writes:

|> > Sorry for the late response.
|> > 
|> > In article <je4rd3qllv.fsf@sykes.suse.de>, Andreas Schwab <schwab@suse.de> writes:
|> > 
|> > > Kenichi Handa <handa@etl.go.jp> writes:
|> > > |> So, please try this:
|> > > |> 
|> > > |> Emacs.Fontset-0: -*-courier-medium-r-normal--*-140-*-*-*-*-fontset-startup,\
|> > > |> 	mule-unicode-2500-33ff:-*-fixed-*-*-*-*-*-*-*-*-*-*-iso10646-1,\
|> > > |> 	mule-unicode-e000-ffff:-*-fixed-*-*-*-*-*-*-*-*-*-*-iso10646-1,\
|> > > |> 	mule-unicode-0100-24ff:-*-fixed-*-*-*-*-*-*-*-*-*-*-iso10646-1
|> > 
|> > > That does not work either, now Emacs is using courier instead of fixed for
|> > > the mule-unicode charsets.
|> > 
|> > Ah, I see.  You actually have courier iso10646-1 fonts.
|> > 
|> > Hmmm, currently, Emacs tries at first the family specified
|> > by face, and if you have that font, Emacs uses it.
|> > 
|> > Perhaps, for finding a font or non-ASCII, non-Latin-1 chars,
|> > Emacs should prefer the family specified in the fontset to
|> > the family specified in the face.
|> 
|> I haven't seen the rest of the thread,

See http://mail.gnu.org/pipermail/emacs-devel/2002-September/012110.html
for the start of the thread.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux AG, Deutschherrnstr. 15-19, D-90429 Nürnberg
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: Fontset problem

[Stefan Monnier]

> |> > > |> So, please try this:
> |> > > |> 
> |> > > |> Emacs.Fontset-0: -*-courier-medium-r-normal--*-140-*-*-*-*-fontset-startup,\
> |> > > |> 	mule-unicode-2500-33ff:-*-fixed-*-*-*-*-*-*-*-*-*-*-iso10646-1,\
> |> > > |> 	mule-unicode-e000-ffff:-*-fixed-*-*-*-*-*-*-*-*-*-*-iso10646-1,\
> |> > > |> 	mule-unicode-0100-24ff:-*-fixed-*-*-*-*-*-*-*-*-*-*-iso10646-1
> |> > 
> |> > > That does not work either, now Emacs is using courier instead of fixed for
> |> > > the mule-unicode charsets.

Oh, so could it be that the `family' property of the `default' face
should be `unspecified' rather than `courier' ?


	Stefan

Re: Fontset problem

[Andreas Schwab]

"Stefan Monnier" <monnier+gnu/emacs@rum.cs.yale.edu> writes:

|> > |> > > |> So, please try this:
|> > |> > > |> 
|> > |> > > |> Emacs.Fontset-0: -*-courier-medium-r-normal--*-140-*-*-*-*-fontset-startup,\
|> > |> > > |> 	mule-unicode-2500-33ff:-*-fixed-*-*-*-*-*-*-*-*-*-*-iso10646-1,\
|> > |> > > |> 	mule-unicode-e000-ffff:-*-fixed-*-*-*-*-*-*-*-*-*-*-iso10646-1,\
|> > |> > > |> 	mule-unicode-0100-24ff:-*-fixed-*-*-*-*-*-*-*-*-*-*-iso10646-1
|> > |> > 
|> > |> > > That does not work either, now Emacs is using courier instead of fixed for
|> > |> > > the mule-unicode charsets.
|> 
|> Oh, so could it be that the `family' property of the `default' face
|> should be `unspecified' rather than `courier' ?

??? But I want courier, not something unspecified.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux AG, Deutschherrnstr. 15-19, D-90429 Nürnberg
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: Fontset problem

[Stefan Monnier]

> "Stefan Monnier" <monnier+gnu/emacs@rum.cs.yale.edu> writes:
> 
> |> > |> > > |> So, please try this:
> |> > |> > > |> 
> |> > |> > > |> Emacs.Fontset-0: -*-courier-medium-r-normal--*-140-*-*-*-*-fontset-startup,\
> |> > |> > > |> 	mule-unicode-2500-33ff:-*-fixed-*-*-*-*-*-*-*-*-*-*-iso10646-1,\
> |> > |> > > |> 	mule-unicode-e000-ffff:-*-fixed-*-*-*-*-*-*-*-*-*-*-iso10646-1,\
> |> > |> > > |> 	mule-unicode-0100-24ff:-*-fixed-*-*-*-*-*-*-*-*-*-*-iso10646-1
> |> > |> > 
> |> > |> > > That does not work either, now Emacs is using courier instead of fixed for
> |> > |> > > the mule-unicode charsets.
> |> 
> |> Oh, so could it be that the `family' property of the `default' face
> |> should be `unspecified' rather than `courier' ?
> 
> ??? But I want courier, not something unspecified.

But that's specified in the `font' property already.


	Stefan

Re: Fontset problem

[Andreas Schwab]

"Stefan Monnier" <monnier+gnu/emacs@rum.cs.yale.edu> writes:

|> > "Stefan Monnier" <monnier+gnu/emacs@rum.cs.yale.edu> writes:
|> > 
|> > |> > |> > > |> So, please try this:
|> > |> > |> > > |> 
|> > |> > |> > > |> Emacs.Fontset-0: -*-courier-medium-r-normal--*-140-*-*-*-*-fontset-startup,\
|> > |> > |> > > |> 	mule-unicode-2500-33ff:-*-fixed-*-*-*-*-*-*-*-*-*-*-iso10646-1,\
|> > |> > |> > > |> 	mule-unicode-e000-ffff:-*-fixed-*-*-*-*-*-*-*-*-*-*-iso10646-1,\
|> > |> > |> > > |> 	mule-unicode-0100-24ff:-*-fixed-*-*-*-*-*-*-*-*-*-*-iso10646-1
|> > |> > |> > 
|> > |> > |> > > That does not work either, now Emacs is using courier instead of fixed for
|> > |> > |> > > the mule-unicode charsets.
|> > |> 
|> > |> Oh, so could it be that the `family' property of the `default' face
|> > |> should be `unspecified' rather than `courier' ?
|> > 
|> > ??? But I want courier, not something unspecified.
|> 
|> But that's specified in the `font' property already.

Sorry, I don't understand what you mean.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux AG, Deutschherrnstr. 15-19, D-90429 Nürnberg
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: Fontset problem

[Stefan Monnier]

> "Stefan Monnier" <monnier+gnu/emacs@rum.cs.yale.edu> writes:
> 
> |> > "Stefan Monnier" <monnier+gnu/emacs@rum.cs.yale.edu> writes:
> |> > 
> |> > |> > |> > > |> So, please try this:
> |> > |> > |> > > |> 
> |> > |> > |> > > |> Emacs.Fontset-0: -*-courier-medium-r-normal--*-140-*-*-*-*-fontset-startup,\
> |> > |> > |> > > |> 	mule-unicode-2500-33ff:-*-fixed-*-*-*-*-*-*-*-*-*-*-iso10646-1,\
> |> > |> > |> > > |> 	mule-unicode-e000-ffff:-*-fixed-*-*-*-*-*-*-*-*-*-*-iso10646-1,\
> |> > |> > |> > > |> 	mule-unicode-0100-24ff:-*-fixed-*-*-*-*-*-*-*-*-*-*-iso10646-1
> |> > |> > |> > 
> |> > |> > |> > > That does not work either, now Emacs is using courier instead of fixed for
> |> > |> > |> > > the mule-unicode charsets.
> |> > |> 
> |> > |> Oh, so could it be that the `family' property of the `default' face
> |> > |> should be `unspecified' rather than `courier' ?
> |> > 
> |> > ??? But I want courier, not something unspecified.
> |> 
> |> But that's specified in the `font' property already.
> 
> Sorry, I don't understand what you mean.

Faces have both a `font' property (which specifies a fontset or
a specific font) and a `family' property.
Typically the `font' is set once and for all in the `default' face
and is not touched by any other face: they set their `family'
property instead.

For some reason, the `default' face initializes its `family' property
not to `unspecified' but to the name of the `font', which will thus
stupidly override any other family setting in the fontset.  I think
that's the bug.


	Stefan

Re: Fontset problem

[Kenichi Handa]

In article <200209261448.g8QEmU915967@rum.cs.yale.edu>, "Stefan Monnier" <monnier+gnu/emacs@rum.cs.yale.edu> writes:
> Faces have both a `font' property (which specifies a fontset or
> a specific font) and a `family' property.
> Typically the `font' is set once and for all in the `default' face
> and is not touched by any other face: they set their `family'
> property instead.

> For some reason, the `default' face initializes its `family' property
> not to `unspecified' but to the name of the `font', which will thus
> stupidly override any other family setting in the fontset.  I think
> that's the bug.

I don't think so.  The font related properties of the
`default' face is decided by a font Emacs uses for a frame,
and that font specifies all font related properties.  That
way, it is assured that the other faces inheriting font
realted properties from the `default' face use the same
font.   I think that is the design policy of xfaces.c.

The current problem is that the role of family part
sepecified in a fontset is not clear; is it just giving a
fallback default, or is it to override the normal font
selection mechanism?

Previously, as people didn't have that many non-Latin-1
fonts, the above was not that big problem because the face
family almost always failed (especially for CJK fonts), thus
the fontset family was used if specified.

But, now, there are many variations of non-Latin-1 fonts,
and incomplete iso10646 fonts make the situation worse.  So,
until we implement a better font selection mechanism for
multibyte characters, people need a way to override the
current font selection mechanism.  First of all, such family
names as courier, times, etc are applicable only to Latin
fonts.

I agree that my latest change is not a complete solution,
but it's a quick workaround to solve the current problem.
People who are satisfied with the current font selector
don't have to specify a family part in a fontset.  So, if it
is specified, I think it is better to regard it a sign of a
request to override the current font selector.

---
Ken'ichi HANDA
handa@m17n.org