Move the #'query-coding-region implementation to C. This is necessary because there is no reasonable way to access the corresponding mswindows-multibyte functionality from Lisp, and we need such functionality if we're going to have a reliable and portable #'query-coding-region implementation. However, this change doesn't yet provide #'query-coding-region for the mswindow-multibyte coding systems, there should be no functional differences between an XEmacs with this change and one without it. src/ChangeLog addition: 2009-09-19 Aidan Kehoe <kehoea@parhasard.net> Move the #'query-coding-region implementation to C. This is necessary because there is no reasonable way to access the corresponding mswindows-multibyte functionality from Lisp, and we need such functionality if we're going to have a reliable and portable #'query-coding-region implementation. However, this change doesn't yet provide #'query-coding-region for the mswindow-multibyte coding systems, there should be no functional differences between an XEmacs with this change and one without it. * mule-coding.c (struct fixed_width_coding_system): Add a new coding system type, fixed_width, and implement it. It uses the CCL infrastructure but has a much simpler creation API, and its own query_method, formerly in lisp/mule/mule-coding.el. * unicode.c: Move the Unicode query method implementation here from unicode.el. * lisp.h: Declare Fmake_coding_system_internal, Fcopy_range_table here. * intl-win32.c (complex_vars_of_intl_win32): Use Fmake_coding_system_internal, not Fmake_coding_system. * general-slots.h: Add Qsucceeded, Qunencodable, Qinvalid_sequence here. * file-coding.h (enum coding_system_variant): Add fixed_width_coding_system here. (struct coding_system_methods): Add query_method and query_lstream_method to the coding system methods. Provide flags for the query methods. Declare the default query method; initialise it correctly in INITIALIZE_CODING_SYSTEM_TYPE. * file-coding.c (default_query_method): New function, the default query method for coding systems that do not set it. Moved from coding.el. (make_coding_system_1): Accept new elements in PROPS in #'make-coding-system; aliases, a list of aliases; safe-chars and safe-charsets (these were previously accepted but not saved); and category. (Fmake_coding_system_internal): New function, what used to be #'make-coding-system--on Mule builds, we've now moved some of the functionality of this to Lisp. (Fcoding_system_canonical_name_p): Move this earlier in the file, since it's now called from within make_coding_system_1. (Fquery_coding_region): Move the implementation of this here, from coding.el. (complex_vars_of_file_coding): Call Fmake_coding_system_internal, not Fmake_coding_system; specify safe-charsets properties when we're a mule build. * extents.h (mouse_highlight_priority, Fset_extent_priority, Fset_extent_face, Fmap_extents): Make these available to other C files. lisp/ChangeLog addition: 2009-09-19 Aidan Kehoe <kehoea@parhasard.net> Move the #'query-coding-region implementation to C. * coding.el: Consolidate code that depends on the presence or absence of Mule at the end of this file. (default-query-coding-region, query-coding-region): Move these functions to C. (default-query-coding-region-safe-charset-skip-chars-map): Remove this variable, the corresponding C variable is Vdefault_query_coding_region_chartab_cache in file-coding.c. (query-coding-string): Update docstring to reflect actual multiple values, be more careful about not modifying a range table that we're currently mapping over. (encode-coding-char): Make the implementation of this simpler. (featurep 'mule): Autoload #'make-coding-system from mule/make-coding-system.el if we're a mule build; provide an appropriate compiler macro. Do various non-mule compatibility things if we're not a mule build. * update-elc.el (additional-dump-dependencies): Add mule/make-coding-system as a dump time dependency if we're a mule build. * unicode.el (ccl-encode-to-ucs-2): (decode-char): (encode-char): Move these earlier in the file, for the sake of some byte compile warnings. (unicode-query-coding-region): Move this to unicode.c * mule/make-coding-system.el: New file, not dumped. Contains the functionality to rework the arguments necessary for fixed-width coding systems, and contains the implementation of #'make-coding-system, which now calls #'make-coding-system-internal. * mule/vietnamese.el (viscii): * mule/latin.el (iso-8859-2): (windows-1250): (iso-8859-3): (iso-8859-4): (iso-8859-14): (iso-8859-15): (iso-8859-16): (iso-8859-9): (macintosh): (windows-1252): * mule/hebrew.el (iso-8859-8): * mule/greek.el (iso-8859-7): (windows-1253): * mule/cyrillic.el (iso-8859-5): (koi8-r): (koi8-u): (windows-1251): (alternativnyj): (koi8-ru): (koi8-t): (koi8-c): (koi8-o): * mule/arabic.el (iso-8859-6): (windows-1256): Move all these coding systems to being of type fixed-width, not of type CCL. This allows the distinct query-coding-region for them to be in C, something which will eventually allow us to implement query-coding-region for the mswindows-multibyte coding systems. * mule/general-late.el (posix-charset-to-coding-system-hash): Document why we're pre-emptively persuading the byte compiler that the ELC for this file needs to be written using escape-quoted. Call #'set-unicode-query-skip-chars-args, now the Unicode query-coding-region implementation is in C. * mule/thai-xtis.el (tis-620): Don't bother checking whether we're XEmacs or not here. * mule/mule-coding.el: Move the eight bit fixed-width functionality from this file to make-coding-system.el. tests/ChangeLog addition: 2009-09-19 Aidan Kehoe <kehoea@parhasard.net> * automated/mule-tests.el: Check a coding system's type, not an 8-bit-fixed property, for whether that coding system should be treated as a fixed-width coding system. * automated/query-coding-tests.el: Don't test the query coding functionality for mswindows-multibyte coding systems, it's not yet implemented.

--- a/lisp/mule/general-late.el	Sat Sep 19 17:56:23 2009 +0200
+++ b/lisp/mule/general-late.el	Sat Sep 19 22:53:13 2009 +0100
@@ -65,31 +65,41 @@
 			       Installation-file-coding-system)
 			    Installation-string)
 
-      ;; Convince the byte compiler that, really, this file can't be encoded
-      ;; as binary. Ugh.
-      system-type (symbol-value (intern "\u0073ystem-type"))
+      ;; This used to be here to convince the byte-compiler to encode the
+      ;; output file using escape-quoted. This is no longer necessary, but
+      ;; keeping it here avoids doing the eval-when-compile clause below
+      ;; twice, which is a significant improvement.
+      system-type (symbol-value (intern "\u0073ystem-type")))
 
-      unicode-query-coding-skip-chars-arg
-      (eval-when-compile 
-        (when-fboundp 'map-charset-chars 
-          (loop
-            for charset in (charset-list)
-            with skip-chars-string = ""
-            do
-            (block no-ucs-mapping
-              (map-charset-chars
-               #'(lambda (begin end)
-                   (loop
-                     while (and begin (>= end begin))
-                     do
-                     (when (= -1 (char-to-unicode begin))
-                       (return-from no-ucs-mapping))
-                     (setq begin (int-to-char (1+ begin)))))
-               charset)
-              (setq skip-chars-string
-                    (concat skip-chars-string
-                            (charset-skip-chars-string charset))))
-            finally return skip-chars-string))))
+;; When this file is being compiled, all the charsets have been loaded, so
+;; we can construct the query-skip-chars-arg string correctly. 
+(set-unicode-query-skip-chars-args
+ (eval-when-compile 
+   (when-fboundp 'map-charset-chars 
+     (loop
+       for charset in (charset-list)
+       with skip-chars-string = ""
+       do
+       (block no-ucs-mapping
+         (map-charset-chars
+          #'(lambda (begin end)
+              (loop
+                while (and begin (>= end begin))
+                do
+                (when (= -1 (char-to-unicode begin))
+                  (return-from no-ucs-mapping))
+                (setq begin (int-to-char (1+ begin)))))
+          charset)
+         (setq skip-chars-string
+               (concat skip-chars-string
+                       (charset-skip-chars-string charset))))
+       finally return skip-chars-string)))
+ unicode-invalid-sequence-regexp-range
+ (eval-when-compile
+   (concat (loop
+             for i from #x80 to #xFF
+             collect (aref (decode-coding-string (int-char i)
+                                                 'utf-8) 0)))))
 
 ;; At this point in the dump, all the charsets have been loaded. Now, load
 ;; their Unicode mappings.