[Fontconfig] Kochi Mincho causes segfault (was Re: Multi lingual fontconfig names)

[Daichi Kawahata]

On Fri, 25 Feb 2005 21:03:15 -0800
Keith Packard wrote:

> you can see that the name 'Kochi Mincho' has no specific language 

I remember now... it was happened when I had updated to this version
2.2.99. GNU libiconv 1.9.2 with --enable-extra-encodings, FreeType2
9.8.3 (CVS), here they are:

  # fc-cache -v
  fc-cache: "/usr/local/share/fonts/opentype": skipping, 4 fonts, 0 dirs
  fc-cache: "/usr/local/share/fonts/truetype": skipping, 3 fonts, 12 dirs
  fc-cache: "/usr/local/share/fonts/truetype/thai": skipping, 12 fonts, 0 dirs
  fc-cache: "/usr/local/share/fonts/truetype/japanese": Bus error

  $ gdb fc-cache core
  [...]
  #0  libiconv (icd=0xffffffff, inbuf=0x10018c68, inbytesleft=0x7ffd7a2c, outbuf=0x7ffd7a30, outbytesleft=0x7ffd7a34) at iconv.c:426
  426         return cd->lfuncs.loop_convert(icd,
  (gdb) bt
  #0  libiconv (icd=0xffffffff, inbuf=0x10018c68, inbytesleft=0x7ffd7a2c, outbuf=0x7ffd7a30, outbytesleft=0x7ffd7a34) at iconv.c:426
  #1  0x5ffc82f0 in FcFreeTypeQuery (file=0x10016f30 "/usr/local/share/fonts/truetype/japanese/kochi-mincho.ttf", id=0, blanks=0x1001d400, count=0x10075248)
    at fcfreetype.c:626
  #2  0x5ffc574c in FcFileScanConfig (set=0x1001d418, dirs=0x10018cc0, cache=0x0, blanks=0x1001d400, 
    file=0x10016f30 "/usr/local/share/fonts/truetype/japanese/kochi-mincho.ttf", force=0, config=0x0) at fcdir.c:117
  #3  0x5ffc5ca0 in FcDirScanConfig (set=0x1001d418, dirs=0x10018cc0, cache=0x0, blanks=0x1001d400, dir=0x10019788 "/usr/local/share/fonts/truetype/japanese"
, 
    force=0, config=0x0) at fcdir.c:240
  #4  0x10001d88 in scanDirs (list=0x10020dd8, config=0x10015008, program=0x7ffd8000 "fc-cache", force=0, verbose=1) at fc-cache.c:179
  #5  0x10001df8 in scanDirs (list=0x10015278, config=0x10015008, program=0x7ffd8000 "fc-cache", force=0, verbose=1) at fc-cache.c:210
  #6  0x10002254 in main (argc=268522104, argv=0x7ffd7e84) at fc-cache.c:291
  (gdb) l
  291         ret = scanDirs (list, config, argv[0], force, verbose);
  292         /* 
  293          * Now we need to sleep a second  (or two, to be extra sure), to make
  294          * sure that timestamps for changes after this run of fc-cache are later
  295          * then any timestamps we wrote.  We don't use gettimeofday() because
  296          * sleep(3) can't be interrupted by a signal here -- this isn't in the
  297          * library, and there aren't any signals flying around here.
  298          */
  299         sleep (2);
  300         if (verbose)

Regards,
-- 
Daichi