[d.love@dl.ac.uk: x-font-family-list gives incomplete results]

[d.love@dl.ac.uk: x-font-family-list gives incomplete results]

[Richard Stallman]

This seems like an important issue, so I am forwarding the message here.

------- Start of forwarded message -------
To: bug-gnu-emacs@gnu.org
From: Dave Love <d.love@dl.ac.uk>
Date: 20 Jan 2003 18:32:46 +0000
Subject: x-font-family-list gives incomplete results
Sender: bug-gnu-emacs-bounces+rms=gnu.org@gnu.org

`x-font-family-list' misses a lot, specifically fonts which are
suitable for offering the user as the default.  E.g. on a Debian
stable system I have this:

(x-font-family-list)
  => (("clean" . t) ("clearlyu") ("clearlyu alternate glyphs") ("clearlyu arabic extra") ("clearlyu ligature") ("clearlyu pua") ("fangsong ti" . t) ("fixed" . t) ("gothic" . t) ("mincho" . t) ("newspaper") ("nil" . t) ("open look cursor") ("open look glyph") ("song ti" . t) ("unifont" . t))

(face-font 'default)
  => "-B&H-LucidaTypewriter-Medium-R-Normal-Sans-12-120-75-75-M-70-ISO8859-1"

Note that the lucidatypewriter I'm actually using isn't listed and I
have 64 families according to xfontsel.

The code implementing this looks superficially correct to me.  Am I
missing something?


_______________________________________________
Bug-gnu-emacs mailing list
Bug-gnu-emacs@gnu.org
http://mail.gnu.org/mailman/listinfo/bug-gnu-emacs
------- End of forwarded message -------

Re: [d.love@dl.ac.uk: x-font-family-list gives incomplete results]

[Jan D.]

Richard Stallman wrote:
> This seems like an important issue, so I am forwarding the message here.
> 
> ------- Start of forwarded message -------
> To: bug-gnu-emacs@gnu.org
> From: Dave Love <d.love@dl.ac.uk>
> Date: 20 Jan 2003 18:32:46 +0000
> Subject: x-font-family-list gives incomplete results
> Sender: bug-gnu-emacs-bounces+rms=gnu.org@gnu.org
> 
> `x-font-family-list' misses a lot, specifically fonts which are
> suitable for offering the user as the default.  E.g. on a Debian
> stable system I have this:
> 
> (x-font-family-list)
>   => (("clean" . t) ("clearlyu") ("clearlyu alternate glyphs") ("clearlyu arabic extra") ("clearlyu ligature") ("clearlyu pua") ("fangsong ti" . t) ("fixed" . t) ("gothic" . t) ("mincho" . t) ("newspaper") ("nil" . t) ("open look cursor") ("open look glyph") ("song ti" . t) ("unifont" . t))
> 
> (face-font 'default)
>   => "-B&H-LucidaTypewriter-Medium-R-Normal-Sans-12-120-75-75-M-70-ISO8859-1"
> 
> Note that the lucidatypewriter I'm actually using isn't listed and I
> have 64 families according to xfontsel.
> 
> The code implementing this looks superficially correct to me.  Am I
> missing something?

There is a bug here.  x-font-family-list does this:

   for (limit = 500;;)
     {
       specbind (intern ("font-list-limit"), make_number (limit));
       nfonts = font_list (f, Qnil, Qnil, Qnil, &fonts);

       if (nfonts == limit)
	{
	  free_font_names (fonts, nfonts);
	  limit *= 2;
	}
       else
	break;
     }

The problem is that font_list return a filtered list.  For instance, duplicates 
are removed, and scalable fonts may also be removed (see variable 
scalable-fonts-allowed).  If I do

% xlsfonts | wc

I get 9568, but if I do

% xlsfonts | sort -u | wc

I get 5690.

So if there is just one duplicate or if scalable fonts are filtered, 
x-font-family-list stops increasing limit and stops reading font names.

One alternative solution would be to make maxnames passed to x_list_fonts 
(xterm.c) mean "get all font names" if it is less than zero.  This also affects 
w32 and mac, since those ports also have a function with this name.

Making maxnames < 0 mean "all font names" should perhaps be propagated into 
Fx_list_fonts?  Now it uses an arbitrary limit of 2000 if maximum is not given, 
which is too few on many machines.


	Jan D.

Re: [d.love@dl.ac.uk: x-font-family-list gives incomplete results]

[Jan D.]

[-- Attachment #1: Type: text/plain, Size: 606 bytes --]

Jan D. wrote:

> One alternative solution would be to make maxnames passed to 
> x_list_fonts (xterm.c) mean "get all font names" if it is less than 
> zero.  This also affects w32 and mac, since those ports also have a 
> function with this name.
> 
> Making maxnames < 0 mean "all font names" should perhaps be propagated 
> into Fx_list_fonts?  Now it uses an arbitrary limit of 2000 if maximum 
> is not given, which is too few on many machines.
> 

Here is a proposed patch.  Comments welcome.  I can not do the mac and w32 
parts, because I have no development environment on such systems.

	Jan D.


[-- Attachment #2: listfont.diff --]
[-- Type: text/plain, Size: 7626 bytes --]

Index: src/xfaces.c
===================================================================
RCS file: /cvsroot/emacs/emacs/src/xfaces.c,v
retrieving revision 1.267
diff -c -r1.267 xfaces.c
*** src/xfaces.c	4 Feb 2003 14:03:17 -0000	1.267
--- src/xfaces.c	16 Feb 2003 19:17:34 -0000
***************
*** 474,480 ****
  static int better_font_p P_ ((int *, struct font_name *, struct font_name *,
  			      int, int));
  static int x_face_list_fonts P_ ((struct frame *, char *,
! 				  struct font_name *, int, int));
  static int font_scalable_p P_ ((struct font_name *));
  static int get_lface_attributes P_ ((struct frame *, Lisp_Object, Lisp_Object *, int));
  static int load_pixmap P_ ((struct frame *, Lisp_Object, unsigned *, unsigned *));
--- 474,480 ----
  static int better_font_p P_ ((int *, struct font_name *, struct font_name *,
  			      int, int));
  static int x_face_list_fonts P_ ((struct frame *, char *,
! 				  struct font_name **, int, int));
  static int font_scalable_p P_ ((struct font_name *));
  static int get_lface_attributes P_ ((struct frame *, Lisp_Object, Lisp_Object *, int));
  static int load_pixmap P_ ((struct frame *, Lisp_Object, unsigned *, unsigned *));
***************
*** 2435,2444 ****
     fonts that we can't parse.  Value is the number of fonts found.  */
  
  static int
! x_face_list_fonts (f, pattern, fonts, nfonts, try_alternatives_p)
       struct frame *f;
       char *pattern;
!      struct font_name *fonts;
       int nfonts, try_alternatives_p;
  {
    int n, nignored;
--- 2435,2444 ----
     fonts that we can't parse.  Value is the number of fonts found.  */
  
  static int
! x_face_list_fonts (f, pattern, pfonts, nfonts, try_alternatives_p)
       struct frame *f;
       char *pattern;
!      struct font_name **pfonts;
       int nfonts, try_alternatives_p;
  {
    int n, nignored;
***************
*** 2447,2452 ****
--- 2447,2454 ----
       better to do it the other way around. */
    Lisp_Object lfonts;
    Lisp_Object lpattern, tem;
+   struct font_name *fonts = 0;
+   int num_fonts = nfonts;
  
    lpattern = build_string (pattern);
  
***************
*** 2459,2468 ****
    lfonts = x_list_fonts (f, lpattern, -1, nfonts);
  #endif
  
    /* Make a copy of the font names we got from X, and
       split them into fields.  */
    n = nignored = 0;
!   for (tem = lfonts; CONSP (tem) && n < nfonts; tem = XCDR (tem))
      {
        Lisp_Object elt, tail;
        const char *name = SDATA (XCAR (tem));
--- 2461,2473 ----
    lfonts = x_list_fonts (f, lpattern, -1, nfonts);
  #endif
  
+   if (nfonts < 0 && CONSP (lfonts))
+     num_fonts = Flength (lfonts);
+   
    /* Make a copy of the font names we got from X, and
       split them into fields.  */
    n = nignored = 0;
!   for (tem = lfonts; CONSP (tem) && n < num_fonts; tem = XCDR (tem))
      {
        Lisp_Object elt, tail;
        const char *name = SDATA (XCAR (tem));
***************
*** 2481,2486 ****
--- 2486,2497 ----
  	  continue;
  	}
  
+       if (! fonts)
+         {
+           *pfonts = (struct font_name *) xmalloc (num_fonts * sizeof **pfonts);
+           fonts = *pfonts;
+         }
+ 
        /* Make a copy of the font name.  */
        fonts[n].name = xstrdup (name);
  
***************
*** 2504,2509 ****
--- 2515,2522 ----
      {
        Lisp_Object list = Valternate_fontname_alist;
  
+       if (fonts) xfree (fonts);
+       
        while (CONSP (list))
  	{
  	  Lisp_Object entry = XCAR (list);
***************
*** 2527,2533 ****
  		    already with no success.  */
  		 && (strcmp (SDATA (name), pattern) == 0
  		     || (n = x_face_list_fonts (f, SDATA (name),
! 						fonts, nfonts, 0),
  			 n == 0)))
  	    patterns = XCDR (patterns);
  	}
--- 2540,2546 ----
  		    already with no success.  */
  		 && (strcmp (SDATA (name), pattern) == 0
  		     || (n = x_face_list_fonts (f, SDATA (name),
! 						pfonts, nfonts, 0),
  			 n == 0)))
  	    patterns = XCDR (patterns);
  	}
***************
*** 2556,2566 ****
  
    /* Get the list of fonts matching pattern.  100 should suffice.  */
    nfonts = DEFAULT_FONT_LIST_LIMIT;
!   if (INTEGERP (Vfont_list_limit) && XINT (Vfont_list_limit) > 0)
!     nfonts = XFASTINT (Vfont_list_limit);
  
!   *fonts = (struct font_name *) xmalloc (nfonts * sizeof **fonts);
!   nfonts = x_face_list_fonts (f, pattern, *fonts, nfonts, 1);
  
    /* Sort the resulting array and return it in *FONTS.  If no
       fonts were found, make sure to set *FONTS to null.  */
--- 2569,2578 ----
  
    /* Get the list of fonts matching pattern.  100 should suffice.  */
    nfonts = DEFAULT_FONT_LIST_LIMIT;
!   if (INTEGERP (Vfont_list_limit))
!     nfonts = XINT (Vfont_list_limit);
  
!   nfonts = x_face_list_fonts (f, pattern, fonts, nfonts, 1);
  
    /* Sort the resulting array and return it in *FONTS.  If no
       fonts were found, make sure to set *FONTS to null.  */
***************
*** 2834,2856 ****
    Lisp_Object result;
    struct gcpro gcpro1;
    int count = SPECPDL_INDEX ();
-   int limit;
- 
-   /* Let's consider all fonts.  Increase the limit for matching
-      fonts until we have them all.  */
-   for (limit = 500;;)
-     {
-       specbind (intern ("font-list-limit"), make_number (limit));
-       nfonts = font_list (f, Qnil, Qnil, Qnil, &fonts);
  
!       if (nfonts == limit)
! 	{
! 	  free_font_names (fonts, nfonts);
! 	  limit *= 2;
! 	}
!       else
! 	break;
!     }
  
    result = Qnil;
    GCPRO1 (result);
--- 2846,2855 ----
    Lisp_Object result;
    struct gcpro gcpro1;
    int count = SPECPDL_INDEX ();
  
!   /* Let's consider all fonts.  */
!   specbind (intern ("font-list-limit"), make_number (-1));
!   nfonts = font_list (f, Qnil, Qnil, Qnil, &fonts);
  
    result = Qnil;
    GCPRO1 (result);
***************
*** 2897,2903 ****
    CHECK_STRING (pattern);
  
    if (NILP (maximum))
!     maxnames = 2000;
    else
      {
        CHECK_NATNUM (maximum);
--- 2896,2902 ----
    CHECK_STRING (pattern);
  
    if (NILP (maximum))
!     maxnames = -1;
    else
      {
        CHECK_NATNUM (maximum);
Index: src/xterm.c
===================================================================
RCS file: /cvsroot/emacs/emacs/src/xterm.c,v
retrieving revision 1.772
diff -c -r1.772 xterm.c
*** src/xterm.c	8 Feb 2003 11:18:32 -0000	1.772
--- src/xterm.c	16 Feb 2003 19:17:45 -0000
***************
*** 14536,14543 ****
  	{
  	  /* We try at least 10 fonts because XListFonts will return
  	     auto-scaled fonts at the head.  */
! 	  names = XListFonts (dpy, SDATA (pattern), max (maxnames, 10),
! 			      &num_fonts);
  	  if (x_had_errors_p (dpy))
  	    {
  	      /* This error is perhaps due to insufficient memory on X
--- 14536,14563 ----
  	{
  	  /* We try at least 10 fonts because XListFonts will return
  	     auto-scaled fonts at the head.  */
!           if (maxnames < 0)
!             {
!               int limit;
! 
!               for (limit = 500;;)
!                 {
!                   names = XListFonts (dpy, SDATA (pattern), limit, &num_fonts);
!                   if (num_fonts == limit)
!                     {
!                       BLOCK_INPUT;
!                       XFreeFontNames (names);
!                       UNBLOCK_INPUT;
!                       limit *= 2;
!                     }
!                   else
!                     break;
!                 }
!             }
!           else
!             names = XListFonts (dpy, SDATA (pattern), max (maxnames, 10),
!                                 &num_fonts);
! 
  	  if (x_had_errors_p (dpy))
  	    {
  	      /* This error is perhaps due to insufficient memory on X

[-- Attachment #3: Type: text/plain, Size: 142 bytes --]

_______________________________________________
Emacs-devel mailing list
Emacs-devel@gnu.org
http://mail.gnu.org/mailman/listinfo/emacs-devel

Re: [d.love@dl.ac.uk: x-font-family-list gives incomplete results]

[Jason Rumney]

"Jan D." <jan.h.d@swipnet.se> writes:

> Jan D. wrote:
> 
> > One alternative solution would be to make maxnames passed to
> > x_list_fonts (xterm.c) mean "get all font names" if it is less than
> > zero.
> 
> Here is a proposed patch.  Comments welcome.  I can not do the mac and
> w32 parts, because I have no development environment on such systems.

I have made the necessary change to w32_list_fonts to support this.

Re: [d.love@dl.ac.uk: x-font-family-list gives incomplete results]

[Jan D.]

> "Jan D." <jan.h.d@swipnet.se> writes:
> 
> > Jan D. wrote:
> > 
> > > One alternative solution would be to make maxnames passed to
> > > x_list_fonts (xterm.c) mean "get all font names" if it is less than
> > > zero.
> > 
> > Here is a proposed patch.  Comments welcome.  I can not do the mac and
> > w32 parts, because I have no development environment on such systems.
> 
> I have made the necessary change to w32_list_fonts to support this.

Thanks, I have now checked in xterm.c/xfaces.c with these changes.

	Jan D.

Re: [d.love@dl.ac.uk: x-font-family-list gives incomplete results]

[Dave Love]

"Jan D." <jan.h.d@swipnet.se> writes:

> Here is a proposed patch.  Comments welcome.  I can not do the mac and
> w32 parts, because I have no development environment on such systems.

Thanks a lot for looking at it.

Unfortunately with it I see the following at startup, which I don't
have time to debug now:

Program received signal SIGSEGV, Segmentation fault.
free (mem=0x102bf6fc) at /net/pxfs5/home/fx/esrc/src/../malloc/malloc.c:3369
3369      if (chunk_is_mmapped(p))                       /* release mmapped memory. */
(gdb) bt
#0  free (mem=0x102bf6fc)
    at /net/pxfs5/home/fx/esrc/src/../malloc/malloc.c:3369
#1  0x00101f64 in emacs_blocked_free (ptr=0x102bf6fc)
    at /net/pxfs5/home/fx/esrc/src/alloc.c:694
#2  0x0016a234 in free (mem=0x0)
    at /net/pxfs5/home/fx/esrc/src/../malloc/malloc.c:3359
#3  0x00101cf0 in xfree (block=0x102bf6fc)
    at /net/pxfs5/home/fx/esrc/src/alloc.c:565
#4  0x00070ba0 in sorted_font_list (f=0x526580, 
    pattern=0xefffc958 "-*-helv-*-ISO8859-1", cmpfn=0x70bac <cmp_font_names>, 
    fonts=0xefffcb8c) at /net/pxfs5/home/fx/esrc/src/xfaces.c:2583
#5  0x00070dc0 in font_list_1 (f=0x526580, pattern=-268449448, family=2965540, 
    registry=5939040, fonts=0xefffcb8c)
    at /net/pxfs5/home/fx/esrc/src/xfaces.c:2672
#6  0x00070e50 in font_list (f=0x526580, pattern=271316684, family=808725812, 
    registry=811251612, fonts=0xefffcb8c)
    at /net/pxfs5/home/fx/esrc/src/xfaces.c:2717
#7  0x00078298 in try_alternative_families (f=0x526580, family=808725812, 
    registry=811251612, fonts=0xefffcb8c)
    at /net/pxfs5/home/fx/esrc/src/xfaces.c:6301
#8  0x000783fc in try_font_list (f=0x526580, attrs=0x30342d34, 
    family=811251596, registry=811251612, fonts=0xefffcb8c, 
    prefer_face_family=1) at /net/pxfs5/home/fx/esrc/src/xfaces.c:6361
#9  0x00078604 in choose_face_font (f=0x526580, attrs=0x5a6ed8, fontset=14, 
    c=0, needs_overstrike=0xefffcbfc)
    at /net/pxfs5/home/fx/esrc/src/xfaces.c:6456
#10 0x0006f68c in load_face_font (f=0x526580, face=0x5a6e90, c=0)
    at /net/pxfs5/home/fx/esrc/src/xfaces.c:1265
#11 0x0007916c in realize_face (cache=0x526580, attrs=0x102cc104, c=0, 
    base_face=0x102cc104, former_face_id=271368452)
    at /net/pxfs5/home/fx/esrc/src/xfaces.c:6704
...

Re: [d.love@dl.ac.uk: x-font-family-list gives incomplete results]

[Jan D.]

> "Jan D." <jan.h.d@swipnet.se> writes:
> 
> > Here is a proposed patch.  Comments welcome.  I can not do the mac and
> > w32 parts, because I have no development environment on such systems.
> 
> Thanks a lot for looking at it.
> 
> Unfortunately with it I see the following at startup, which I don't
> have time to debug now:

The patch contained a bug.  I have fixed that in the check in I just did.

	Jan D.

Re: [d.love@dl.ac.uk: x-font-family-list gives incomplete results]

[Kenichi Handa]

"Jan D." <jan.h.d@bogus.example.com> writes:
> The patch contained a bug.  I have fixed that in the check in I just did.

With the latest CVS code, Emacs still crashes as below when
I type C-h h.

Program received signal SIGSEGV, Segmentation fault.
0x40354a90 in free () from /lib/libc.so.6
(gdb) bt
#0  0x40354a90 in free () from /lib/libc.so.6
#1  0x08117c9a in emacs_blocked_free (ptr=0x8750400) at alloc.c:694
#2  0x40354a29 in free () from /lib/libc.so.6
#3  0x08117ab0 in xfree (block=0x8750400) at alloc.c:565
#4  0x0809c927 in sorted_font_list (f=0x851cd18, 
    pattern=0xbfffe2e4 "-*-*-*-devanagari-cdac", 
    cmpfn=0x809c938 <cmp_font_names>, fonts=0xbfffe418) at xfaces.c:2585
#5  0x0809cb1d in font_list_1 (f=0x851cd18, pattern=405469284, 
    family=944773028, registry=944773044, fonts=0xbfffe418) at xfaces.c:2674
#6  0x0809cbd1 in font_list (f=0x851cd18, pattern=405469284, family=944773028, 
    registry=944773044, fonts=0xbfffe418) at xfaces.c:2719
#7  0x080a2d57 in try_alternative_families (f=0x851cd18, family=944773028, 
    registry=944773044, fonts=0xbfffe418) at xfaces.c:6303
#8  0x080a2e84 in try_font_list (f=0x851cd18, attrs=0x874adb8, 
    family=944773028, registry=944773044, fonts=0xbfffe418, 
    prefer_face_family=0) at xfaces.c:6363
#9  0x080a3039 in choose_face_font (f=0x851cd18, attrs=0x874adb8, fontset=11, 
    c=266272, needs_overstrike=0xbfffe458) at xfaces.c:6458
#10 0x0809b70d in load_face_font (f=0x851cd18, face=0x874ad70, c=266272)
    at xfaces.c:1265
#11 0x080a3962 in realize_face (cache=0x851f9f0, attrs=0x8573528, c=266272, 
    base_face=0x85734e0, former_face_id=-1) at xfaces.c:6706
#12 0x080a5bca in lookup_face (f=0x851cd18, attr=0x8573528, c=266272, 
    base_face=0x85734e0) at xfaces.c:5641
#13 0x080d1ec7 in face_for_char (f=0x851cd18, face=0x85734e0, c=266272)
    at fontset.c:483
#14 0x080a82d7 in x_produce_glyphs (it=0xbfffe6b4) at xterm.c:2108
#15 0x0806f07d in display_line (it=0xbfffe6b4) at xdisp.c:13307
...

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

Re: [d.love@dl.ac.uk: x-font-family-list gives incomplete results]

[Jan D.]

> "Jan D." <jan.h.d@bogus.example.com> writes:
> > The patch contained a bug.  I have fixed that in the check in I just did.
> 
> With the latest CVS code, Emacs still crashes as below when
> I type C-h h.

Sorry about that, checked in wrong version.  I fixed that now.

	Jan D.

Re: [d.love@dl.ac.uk: x-font-family-list gives incomplete results]

[Kenichi Handa]

In article <200302190652.h1J6qoFM020719@stubby.bodenonline.com>, "Jan D." <jan.h.d@swipnet.se> writes:
>>  "Jan D." <jan.h.d@bogus.example.com> writes:
>>  > The patch contained a bug.  I have fixed that in the check in I just did.
>>  
>>  With the latest CVS code, Emacs still crashes as below when
>>  I type C-h h.

> Sorry about that, checked in wrong version.  I fixed that now.

Thank you.  I confirmed that the problem is fixed.

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