Resolve the unregistered-CCL-programs-get-garbage-collected problem correctly src/ChangeLog addition: 2009-11-15 Aidan Kehoe <kehoea@parhasard.net> * mule-ccl.c (CCL_CALL_FOR_MAP_INSTRUCTION): Assert that we always have a symbol in this macro. (setup_ccl_program): Ensure we're not allocating unreachable memory in this function; all symbols must have been resolved in a given CCL program before this function is called. (find_ccl_program): New function, return a CCL program with all its symbols resolved if it is valid (possibly allocating memory), Qnil otherwise. (get_ccl_program): New function, exported to other files; call find_ccl_program, and error if it gives nil. (Fccl_program_p): Call find_ccl_program from this function instead of implementing the bulk of it here. (Fccl_execute): Call get_ccl_program instead of implementing the bulk of it here. (Fccl_execute_on_string): Ditto. * mule-ccl.h (Vfont_ccl_encoder_alist): Remove this declaration, it hasn't been used in years. (get_ccl_program): Declare this function. * mule-coding.c (ccl_putprop): Use get_ccl_program on any specified encode or decode CCL program property. (fixed_width_putprop): Ditto. * mule-charset.c (Fmake_charset): Use get_ccl_program on any specified ccl-program. (Fset_charset_ccl_program): Ditto.

The files in this directory were converted from data in the mule-ucs
package (mule-ucs/lisp/reldata/*), using this code:

(defun convert-mule-ucs-file (start end)
  (interactive "r")
  (with-output-to-temp-buffer "*mule-ucs-convert*"
    (save-excursion
      (goto-char start)
      (while (re-search-forward "(\\?\\(.\\) \\. \"\\(.*\\)\") ;+ \\(.*\\)$"
				end t)
	(let ((ch (string-to-char (match-string 1)))
	      (codepoint (match-string 2))
	      (name (match-string 3)))
	  (if (= 1 (charset-dimension (char-charset ch)))
	      (princ (format "0x%x   %s  #   %s\n" (char-octet ch) codepoint name))
	    (princ (format "0x%x%x   %s  #   %s\n" (char-octet ch 0) (char-octet ch 1) codepoint name))))))))

Each file is named after the XEmacs charset it represents.  The CNS files
contain more codepoints than those in unicode-consortium/ because they list
codepoints above 0xFFFF, those handled by surrogates (supported starting in
Windows 2000, I think, but not yet by XEmacs).