Mercurial > hg > xemacs-beta
view src/mule-charset.c @ 853:2b6fa2618f76
[xemacs-hg @ 2002-05-28 08:44:22 by ben]
merge my stderr-proc ws
make-docfile.c: Fix places where we forget to check for EOF.
code-init.el: Don't use CRLF conversion by default on process output. CMD.EXE and
friends work both ways but Cygwin programs don't like the CRs.
code-process.el, multicast.el, process.el: Removed.
Improvements to call-process-internal:
-- allows a buffer to be specified for input and stderr output
-- use it on all systems
-- implement C-g as documented
-- clean up and comment
call-process-region uses new call-process facilities; no temp file.
remove duplicate funs in process.el.
comment exactly how coding systems work and fix various problems.
open-multicast-group now does similar coding-system frobbing to
open-network-stream.
dumped-lisp.el, faces.el, msw-faces.el: Fix some hidden errors due to code not being defined at the right time.
xemacs.mak: Add -DSTRICT.
================================================================
ALLOW SEPARATION OF STDOUT AND STDERR IN PROCESSES
================================================================
Standard output and standard error can be processed separately in
a process. Each can have its own buffer, its own mark in that buffer,
and its filter function. You can specify a separate buffer for stderr
in `start-process' to get things started, or use the new primitives:
set-process-stderr-buffer
process-stderr-buffer
process-stderr-mark
set-process-stderr-filter
process-stderr-filter
Also, process-send-region takes a 4th optional arg, a buffer.
Currently always uses a pipe() under Unix to read the error output.
(#### Would a PTY be better?)
sysdep.h, sysproc.h, unexfreebsd.c, unexsunos4.c, nt.c, emacs.c, callproc.c, symsinit.h, sysdep.c, Makefile.in.in, process-unix.c: Delete callproc.c. Move child_setup() to process-unix.c.
wait_for_termination() now only needed on a few really old systems.
console-msw.h, event-Xt.c, event-msw.c, event-stream.c, event-tty.c, event-unixoid.c, events.h, process-nt.c, process-unix.c, process.c, process.h, procimpl.h: Rewrite the process methods to handle a separate channel for
error input. Create Lstreams for reading in the error channel.
Many process methods need change. In general the changes are
fairly clear as they involve duplicating what's used for reading
the normal stdout and changing for stderr -- although tedious,
as such changes are required throughout the entire process code.
Rewrote the code that reads process output to do two loops, one
for stdout and one for stderr.
gpmevent.c, tooltalk.c: set_process_filter takes an argument for stderr.
================================================================
NEW ERROR-TRAPPING MECHANISM
================================================================
Totally rewrite error trapping code to be unified and support more
features. Basic function is call_trapping_problems(), which lets
you specify, by means of flags, what sorts of problems you want
trapped. these can include
-- quit
-- errors
-- throws past the function
-- creation of "display objects" (e.g. buffers)
-- deletion of already-existing "display objects" (e.g. buffers)
-- modification of already-existing buffers
-- entering the debugger
-- gc
-- errors->warnings (ala suspended errors)
etc. All other error funs rewritten in terms of this one.
Various older mechanisms removed or rewritten.
window.c, insdel.c, console.c, buffer.c, device.c, frame.c: When creating a display object, added call to
note_object_created(), for use with trapping_problems mechanism.
When deleting, call check_allowed_operation() and note_object
deleted().
The trapping-problems code records the objects created since the
call-trapping-problems began. Those objects can be deleted, but
none others (i.e. previously existing ones).
bytecode.c, cmdloop.c: internal_catch takes another arg.
eval.c: Add long comments describing the "five lists" used to maintain
state (backtrace, gcpro, specbind, etc.) in the Lisp engine.
backtrace.h, eval.c: Implement trapping-problems mechanism, eliminate old mechanisms or
redo in terms of new one.
frame.c, gutter.c: Flush out the concept of "critical display section", defined by
the in_display() var. Use an internal_bind() to get it reset,
rather than just doing it at end, because there may be a non-local
exit.
event-msw.c, event-stream.c, console-msw.h, device.c, dialog-msw.c, frame.c, frame.h, intl.c, toolbar.c, menubar-msw.c, redisplay.c, alloc.c, menubar-x.c: Make use of new trapping-errors stuff and rewrite code based on
old mechanisms.
glyphs-widget.c, redisplay.h: Protect calling Lisp in redisplay.
insdel.c: Protect hooks against deleting existing buffers.
frame-msw.c: Use EQ, not EQUAL in hash tables whose keys are just numbers.
Otherwise we run into stickiness in redisplay because
internal_equal() can QUIT.
================================================================
SIGNAL, C-G CHANGES
================================================================
Here we change the way that C-g interacts with event reading. The
idea is that a C-g occurring while we're reading a user event
should be read as C-g, but elsewhere should be a QUIT. The former
code did all sorts of bizarreness -- requiring that no QUIT occurs
anywhere in event-reading code (impossible to enforce given the
stuff called or Lisp code invoked), and having some weird system
involving enqueue/dequeue of a C-g and interaction with Vquit_flag
-- and it didn't work.
Now, we simply enclose all code where we want C-g read as an event
with {begin/end}_dont_check_for_quit(). This completely turns off
the mechanism that checks (and may remove or alter) C-g in the
read-ahead queues, so we just get the C-g normal.
Signal.c documents this very carefully.
cmdloop.c: Correct use of dont_check_for_quit to new scheme, remove old
out-of-date comments.
event-stream.c: Fix C-g handling to actually work.
device-x.c: Disable quit checking when err out.
signal.c: Cleanup. Add large descriptive comment.
process-unix.c, process-nt.c, sysdep.c: Use QUIT instead of REALLY_QUIT.
It's not necessary to use REALLY_QUIT and just confuses the issue.
lisp.h: Comment quit handlers.
================================================================
CONS CHANGES
================================================================
free_cons() now takes a Lisp_Object not the result of XCONS().
car and cdr have been renamed so that they don't get used directly;
go through XCAR(), XCDR() instead.
alloc.c, dired.c, editfns.c, emodules.c, fns.c, glyphs-msw.c, glyphs-x.c, glyphs.c, keymap.c, minibuf.c, search.c, eval.c, lread.c, lisp.h: Correct free_cons calling convention: now takes Lisp_Object,
not Lisp_Cons
chartab.c: Eliminate direct use of ->car, ->cdr, should be black box.
callint.c: Rewrote using EXTERNAL_LIST_LOOP to avoid use of Lisp_Cons.
================================================================
USE INTERNAL-BIND-*
================================================================
eval.c: Cleanups of these funs.
alloc.c, fileio.c, undo.c, specifier.c, text.c, profile.c, lread.c, redisplay.c, menubar-x.c, macros.c: Rewrote to use internal_bind_int() and internal_bind_lisp_object()
in place of whatever varied and cumbersome mechanisms were
formerly there.
================================================================
SPECBIND SANITY
================================================================
backtrace.h: - Improved comments
backtrace.h, bytecode.c, eval.c: Add new mechanism check_specbind_stack_sanity() for sanity
checking code each time the catchlist or specbind stack change.
Removed older prototype of same mechanism.
================================================================
MISC
================================================================
lisp.h, insdel.c, window.c, device.c, console.c, buffer.c: Fleshed out authorship.
device-msw.c: Correct bad Unicode-ization.
print.c: Be more careful when not initialized or in fatal error handling.
search.c: Eliminate running_asynch_code, an FSF holdover.
alloc.c: Added comments about gc-cons-threshold.
dialog-x.c: Use begin_gc_forbidden() around code to build up a widget value
tree, like in menubar-x.c.
gui.c: Use Qunbound not Qnil as the default for
gethash.
lisp-disunion.h, lisp-union.h: Added warnings on use of VOID_TO_LISP().
lisp.h: Use ERROR_CHECK_STRUCTURES to turn on
ERROR_CHECK_TRAPPING_PROBLEMS and ERROR_CHECK_TYPECHECK
lisp.h: Add assert_with_message.
lisp.h: Add macros for gcproing entire arrays. (You could do this before
but it required manual twiddling the gcpro structure.)
lisp.h: Add prototypes for new functions defined elsewhere.
author | ben |
---|---|
date | Tue, 28 May 2002 08:45:36 +0000 |
parents | 6728e641994e |
children | 804517e16990 |
line wrap: on
line source
/* Functions to handle multilingual characters. Copyright (C) 1992, 1995 Free Software Foundation, Inc. Copyright (C) 1995 Sun Microsystems, Inc. Copyright (C) 2001, 2002 Ben Wing. This file is part of XEmacs. XEmacs is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2, or (at your option) any later version. XEmacs is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with XEmacs; see the file COPYING. If not, write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. */ /* Synched up with: FSF 20.3. Not in FSF. */ /* Rewritten by Ben Wing <ben@xemacs.org>. */ #include <config.h> #include "lisp.h" #include "buffer.h" #include "chartab.h" #include "elhash.h" #include "device.h" #include "faces.h" #include "lstream.h" #include "mule-ccl.h" /* The various pre-defined charsets. */ Lisp_Object Vcharset_ascii; Lisp_Object Vcharset_control_1; Lisp_Object Vcharset_latin_iso8859_1; Lisp_Object Vcharset_latin_iso8859_2; Lisp_Object Vcharset_latin_iso8859_3; Lisp_Object Vcharset_latin_iso8859_4; Lisp_Object Vcharset_thai_tis620; Lisp_Object Vcharset_greek_iso8859_7; Lisp_Object Vcharset_arabic_iso8859_6; Lisp_Object Vcharset_hebrew_iso8859_8; Lisp_Object Vcharset_katakana_jisx0201; Lisp_Object Vcharset_latin_jisx0201; Lisp_Object Vcharset_cyrillic_iso8859_5; Lisp_Object Vcharset_latin_iso8859_9; Lisp_Object Vcharset_latin_iso8859_15; Lisp_Object Vcharset_japanese_jisx0208_1978; Lisp_Object Vcharset_chinese_gb2312; Lisp_Object Vcharset_japanese_jisx0208; Lisp_Object Vcharset_korean_ksc5601; Lisp_Object Vcharset_japanese_jisx0212; Lisp_Object Vcharset_chinese_cns11643_1; Lisp_Object Vcharset_chinese_cns11643_2; Lisp_Object Vcharset_chinese_big5_1; Lisp_Object Vcharset_chinese_big5_2; Lisp_Object Vcharset_composite; struct charset_lookup *chlook; static const struct lrecord_description charset_lookup_description_1[] = { { XD_LISP_OBJECT_ARRAY, offsetof (struct charset_lookup, charset_by_leading_byte), NUM_LEADING_BYTES+4*128*2 }, { XD_END } }; static const struct struct_description charset_lookup_description = { sizeof (struct charset_lookup), charset_lookup_description_1 }; Lisp_Object Qcharsetp; /* Qdoc_string, Qdimension, Qchars defined in general.c */ Lisp_Object Qregistry, Qfinal, Qgraphic; Lisp_Object Qdirection; Lisp_Object Qreverse_direction_charset; Lisp_Object Qshort_name, Qlong_name; Lisp_Object Qfrom_unicode, Qto_unicode; Lisp_Object Qlatin_iso8859_1, Qlatin_iso8859_2, Qlatin_iso8859_3, Qlatin_iso8859_4, Qthai_tis620, Qgreek_iso8859_7, Qarabic_iso8859_6, Qhebrew_iso8859_8, Qkatakana_jisx0201, Qlatin_jisx0201, Qcyrillic_iso8859_5, Qlatin_iso8859_9, Qlatin_iso8859_15, Qjapanese_jisx0208_1978, Qchinese_gb2312, Qjapanese_jisx0208, Qkorean_ksc5601, Qjapanese_jisx0212, Qchinese_cns11643_1, Qchinese_cns11643_2, Qchinese_big5_1, Qchinese_big5_2, Qcomposite; Lisp_Object Ql2r, Qr2l; Lisp_Object Vcharset_hash_table; /************************************************************************/ /* charset object */ /************************************************************************/ static Lisp_Object mark_charset (Lisp_Object obj) { Lisp_Charset *cs = XCHARSET (obj); mark_object (cs->short_name); mark_object (cs->long_name); mark_object (cs->doc_string); mark_object (cs->registry); mark_object (cs->ccl_program); return cs->name; } static void print_charset (Lisp_Object obj, Lisp_Object printcharfun, int escapeflag) { Lisp_Charset *cs = XCHARSET (obj); if (print_readably) printing_unreadable_object ("#<charset %s 0x%x>", XSTRING_DATA (XSYMBOL (CHARSET_NAME (cs))-> name), cs->header.uid); write_fmt_string_lisp (printcharfun, "#<charset %s %S %S %S", 4, CHARSET_NAME (cs), CHARSET_SHORT_NAME (cs), CHARSET_LONG_NAME (cs), CHARSET_DOC_STRING (cs)); write_fmt_string (printcharfun, " %s %s cols=%d g%d final='%c' reg=", CHARSET_TYPE (cs) == CHARSET_TYPE_94 ? "94" : CHARSET_TYPE (cs) == CHARSET_TYPE_96 ? "96" : CHARSET_TYPE (cs) == CHARSET_TYPE_94X94 ? "94x94" : "96x96", CHARSET_DIRECTION (cs) == CHARSET_LEFT_TO_RIGHT ? "l2r" : "r2l", CHARSET_COLUMNS (cs), CHARSET_GRAPHIC (cs), CHARSET_FINAL (cs)); print_internal (CHARSET_REGISTRY (cs), printcharfun, 0); write_fmt_string (printcharfun, " 0x%x>", cs->header.uid); } static void finalize_charset (void *header, int for_disksave) { /* See mule-charset.h, definition of Lisp_Charset. */ Lisp_Object charset = wrap_charset ((Lisp_Charset *) header); if (for_disksave && XCHARSET_TO_UNICODE_TABLE (charset)) { /* Control-1, ASCII, Composite don't have tables */ free_charset_unicode_tables (charset); XCHARSET_TO_UNICODE_TABLE (charset) = 0; XCHARSET_FROM_UNICODE_TABLE (charset) = 0; } } static const struct lrecord_description charset_description[] = { { XD_LISP_OBJECT, offsetof (Lisp_Charset, name) }, { XD_LISP_OBJECT, offsetof (Lisp_Charset, doc_string) }, { XD_LISP_OBJECT, offsetof (Lisp_Charset, registry) }, { XD_LISP_OBJECT, offsetof (Lisp_Charset, short_name) }, { XD_LISP_OBJECT, offsetof (Lisp_Charset, long_name) }, { XD_LISP_OBJECT, offsetof (Lisp_Charset, reverse_direction_charset) }, { XD_LISP_OBJECT, offsetof (Lisp_Charset, ccl_program) }, #if 0 /* #### XD_UNION not yet implemented! pdump version of XEmacs will not work! */ { XD_UNION, offsetof (Lisp_Charset, to_unicode_table), XD_INDIRECT (offsetof (Lisp_Charset, dimension), 0), to_unicode_description }, { XD_UNION, offsetof (Lisp_Charset, from_unicode_table), XD_INDIRECT (offsetof (Lisp_Charset, from_unicode_levels), 0), from_unicode_description }, #endif { XD_END } }; DEFINE_LRECORD_IMPLEMENTATION ("charset", charset, mark_charset, print_charset, finalize_charset, 0, 0, charset_description, Lisp_Charset); /* Make a new charset. */ /* #### SJT Should generic properties be allowed? */ static Lisp_Object make_charset (int id, Lisp_Object name, int rep_bytes, int type, int columns, int graphic, Intbyte final, int direction, Lisp_Object short_name, Lisp_Object long_name, Lisp_Object doc, Lisp_Object reg, int overwrite) { Lisp_Object obj; Lisp_Charset *cs; if (!overwrite) { cs = alloc_lcrecord_type (Lisp_Charset, &lrecord_charset); zero_lcrecord (cs); obj = wrap_charset (cs); if (final) { /* some charsets do not have final characters. This includes ASCII, Control-1, Composite, and the two faux private charsets. */ assert (NILP (chlook-> charset_by_attributes[type][final][direction])); chlook->charset_by_attributes[type][final][direction] = obj; } assert (NILP (chlook->charset_by_leading_byte[id - MIN_LEADING_BYTE])); chlook->charset_by_leading_byte[id - MIN_LEADING_BYTE] = obj; } else { Lisp_Object ret; /* Actually overwrite the properties of the existing charset. We do this because until now charsets could never be "deleted", so parts of the code don't bother to GC charsets. */ obj = chlook->charset_by_leading_byte[id - MIN_LEADING_BYTE]; cs = XCHARSET (obj); assert (EQ (chlook->charset_by_attributes[type][final][direction], obj)); ret = Fremhash (XCHARSET_NAME (obj), Vcharset_hash_table); assert (!NILP (ret)); } CHARSET_ID (cs) = id; CHARSET_NAME (cs) = name; CHARSET_SHORT_NAME (cs) = short_name; CHARSET_LONG_NAME (cs) = long_name; CHARSET_REP_BYTES (cs) = rep_bytes; CHARSET_DIRECTION (cs) = direction; CHARSET_TYPE (cs) = type; CHARSET_COLUMNS (cs) = columns; CHARSET_GRAPHIC (cs) = graphic; CHARSET_FINAL (cs) = final; CHARSET_DOC_STRING (cs) = doc; CHARSET_REGISTRY (cs) = reg; CHARSET_CCL_PROGRAM (cs) = Qnil; CHARSET_REVERSE_DIRECTION_CHARSET (cs) = Qnil; CHARSET_DIMENSION (cs) = (CHARSET_TYPE (cs) == CHARSET_TYPE_94 || CHARSET_TYPE (cs) == CHARSET_TYPE_96) ? 1 : 2; CHARSET_CHARS (cs) = (CHARSET_TYPE (cs) == CHARSET_TYPE_94 || CHARSET_TYPE (cs) == CHARSET_TYPE_94X94) ? 94 : 96; if (id == LEADING_BYTE_ASCII || id == LEADING_BYTE_CONTROL_1 #ifdef ENABLE_COMPOSITE_CHARS || id == LEADING_BYTE_COMPOSITE #endif ) assert (!overwrite); else { if (overwrite) free_charset_unicode_tables (obj); init_charset_unicode_tables (obj); } /* Some charsets are "faux" and don't have names or really exist at all except in the leading-byte table. */ if (!NILP (name)) { assert (NILP (Fgethash (name, Vcharset_hash_table, Qnil))); Fputhash (name, obj, Vcharset_hash_table); } recalculate_unicode_precedence (); return obj; } static int get_unallocated_leading_byte (int dimension) { int lb; if (dimension == 1) { if (chlook->next_allocated_1_byte_leading_byte > MAX_LEADING_BYTE_PRIVATE_1) lb = 0; else lb = chlook->next_allocated_1_byte_leading_byte++; } else { if (chlook->next_allocated_2_byte_leading_byte > MAX_LEADING_BYTE_PRIVATE_2) lb = 0; else lb = chlook->next_allocated_2_byte_leading_byte++; } if (!lb) invalid_operation ("No more character sets free for this dimension", make_int (dimension)); return lb; } /************************************************************************/ /* Basic charset Lisp functions */ /************************************************************************/ void get_charset_limits (Lisp_Object charset, int *low, int *high) { Lisp_Charset *cs = XCHARSET (charset); if (EQ (charset, Vcharset_ascii)) *low = 0, *high = 127; else if (EQ (charset, Vcharset_control_1)) *low = 0, *high = 31; else if (CHARSET_CHARS (cs) == 94) *low = 33, *high = 126; else /* CHARSET_CHARS (cs) == 96) */ *low = 32, *high = 127; } DEFUN ("charsetp", Fcharsetp, 1, 1, 0, /* Return non-nil if OBJECT is a charset. */ (object)) { return CHARSETP (object) ? Qt : Qnil; } DEFUN ("find-charset", Ffind_charset, 1, 1, 0, /* Retrieve the charset of the given name. If CHARSET-OR-NAME is a charset object, it is simply returned. Otherwise, CHARSET-OR-NAME should be a symbol. If there is no such charset, nil is returned. Otherwise the associated charset object is returned. */ (charset_or_name)) { if (CHARSETP (charset_or_name)) return charset_or_name; CHECK_SYMBOL (charset_or_name); return Fgethash (charset_or_name, Vcharset_hash_table, Qnil); } DEFUN ("get-charset", Fget_charset, 1, 1, 0, /* Retrieve the charset of the given name. Same as `find-charset' except an error is signalled if there is no such charset instead of returning nil. */ (name)) { Lisp_Object charset = Ffind_charset (name); if (NILP (charset)) invalid_argument ("No such charset", name); return charset; } /* We store the charsets in hash tables with the names as the key and the actual charset object as the value. Occasionally we need to use them in a list format. These routines provide us with that. */ struct charset_list_closure { Lisp_Object *charset_list; }; static int add_charset_to_list_mapper (Lisp_Object key, Lisp_Object value, void *charset_list_closure) { /* This function can GC */ struct charset_list_closure *chcl = (struct charset_list_closure*) charset_list_closure; Lisp_Object *charset_list = chcl->charset_list; *charset_list = Fcons (XCHARSET_NAME (value), *charset_list); return 0; } DEFUN ("charset-list", Fcharset_list, 0, 0, 0, /* Return a list of the names of all defined charsets. */ ()) { Lisp_Object charset_list = Qnil; struct gcpro gcpro1; struct charset_list_closure charset_list_closure; GCPRO1 (charset_list); charset_list_closure.charset_list = &charset_list; elisp_maphash (add_charset_to_list_mapper, Vcharset_hash_table, &charset_list_closure); UNGCPRO; return charset_list; } DEFUN ("charset-name", Fcharset_name, 1, 1, 0, /* Return the name of charset CHARSET. */ (charset)) { return XCHARSET_NAME (Fget_charset (charset)); } /* #### SJT Should generic properties be allowed? */ DEFUN ("make-charset", Fmake_charset, 3, 3, 0, /* Define a new character set. This function is for use with Mule support. NAME is a symbol, the name by which the character set is normally referred. DOC-STRING is a string describing the character set. PROPS is a property list, describing the specific nature of the character set. Recognized properties are: 'short-name Short version of the charset name (ex: Latin-1) 'long-name Long version of the charset name (ex: ISO8859-1 (Latin-1)) 'registry A regular expression matching the font registry field for this character set. 'dimension Number of octets used to index a character in this charset. Either 1 or 2. Defaults to 1. 'columns Number of columns used to display a character in this charset. Only used in TTY mode. (Under X, the actual width of a character can be derived from the font used to display the characters.) If unspecified, defaults to the dimension (this is almost always the correct value). 'chars Number of characters in each dimension (94 or 96). Defaults to 94. Note that if the dimension is 2, the character set thus described is 94x94 or 96x96. 'final Final byte of ISO 2022 escape sequence. Must be supplied. Each combination of (DIMENSION, CHARS) defines a separate namespace for final bytes. Note that ISO 2022 restricts the final byte to the range 0x30 - 0x7E if dimension == 1, and 0x30 - 0x5F if dimension == 2. Note also that final bytes in the range 0x30 - 0x3F are reserved for user-defined (not official) character sets. 'graphic 0 (use left half of font on output) or 1 (use right half of font on output). Defaults to 0. For example, for a font whose registry is ISO8859-1, the left half (octets 0x20 - 0x7F) is the `ascii' character set, while the right half (octets 0xA0 - 0xFF) is the `latin-1' character set. With 'graphic set to 0, the octets will have their high bit cleared; with it set to 1, the octets will have their high bit set. 'direction 'l2r (left-to-right) or 'r2l (right-to-left). Defaults to 'l2r. 'ccl-program A compiled CCL program used to convert a character in this charset into an index into the font. This is in addition to the 'graphic property. The CCL program is passed the octets of the character, with the high bit cleared and set depending upon whether the value of the 'graphic property is 0 or 1. */ (name, doc_string, props)) { int id, dimension = 1, chars = 94, graphic = 0, columns = -1; Intbyte final = 0; int direction = CHARSET_LEFT_TO_RIGHT; int type; Lisp_Object registry = Qnil; Lisp_Object charset = Qnil; Lisp_Object ccl_program = Qnil; Lisp_Object short_name = Qnil, long_name = Qnil; Lisp_Object existing_charset; int temporary = UNBOUNDP (name); /* NOTE: name == Qunbound is a directive from the iso2022 code to create a temporary charset for an unknown final. We allow the final to be overwritten with a real charset later on. */ if (!NILP (doc_string)) CHECK_STRING (doc_string); if (!UNBOUNDP (name)) { CHECK_SYMBOL (name); charset = Ffind_charset (name); if (!NILP (charset)) invalid_operation ("Cannot redefine existing charset", name); } { EXTERNAL_PROPERTY_LIST_LOOP_3 (keyword, value, props) { if (EQ (keyword, Qshort_name)) { CHECK_STRING (value); short_name = value; } else if (EQ (keyword, Qlong_name)) { CHECK_STRING (value); long_name = value; } else if (EQ (keyword, Qdimension)) { CHECK_INT (value); dimension = XINT (value); if (dimension < 1 || dimension > 2) invalid_constant ("Invalid value for 'dimension", value); } else if (EQ (keyword, Qchars)) { CHECK_INT (value); chars = XINT (value); if (chars != 94 && chars != 96) invalid_constant ("Invalid value for 'chars", value); } else if (EQ (keyword, Qcolumns)) { CHECK_INT (value); columns = XINT (value); if (columns != 1 && columns != 2) invalid_constant ("Invalid value for 'columns", value); } else if (EQ (keyword, Qgraphic)) { CHECK_INT (value); graphic = XINT (value); if (graphic < 0 || graphic > 1) invalid_constant ("Invalid value for 'graphic", value); } else if (EQ (keyword, Qregistry)) { CHECK_STRING (value); registry = value; } else if (EQ (keyword, Qdirection)) { if (EQ (value, Ql2r)) direction = CHARSET_LEFT_TO_RIGHT; else if (EQ (value, Qr2l)) direction = CHARSET_RIGHT_TO_LEFT; else invalid_constant ("Invalid value for 'direction", value); } else if (EQ (keyword, Qfinal)) { CHECK_CHAR_COERCE_INT (value); final = XCHAR (value); if (final < '0' || final > '~') invalid_constant ("Invalid value for 'final", value); } else if (EQ (keyword, Qccl_program)) { struct ccl_program test_ccl; if (setup_ccl_program (&test_ccl, value) < 0) invalid_argument ("Invalid value for 'ccl-program", value); ccl_program = value; } else invalid_constant ("Unrecognized property", keyword); } } if (!final) invalid_argument ("'final must be specified", Qunbound); if (dimension == 2 && final > 0x5F) invalid_constant ("Final must be in the range 0x30 - 0x5F for dimension == 2", make_char (final)); if (dimension == 1) type = (chars == 94) ? CHARSET_TYPE_94 : CHARSET_TYPE_96; else type = (chars == 94) ? CHARSET_TYPE_94X94 : CHARSET_TYPE_96X96; existing_charset = charset_by_attributes (type, final, direction); if (!NILP (existing_charset) && !XCHARSET (existing_charset)->temporary) invalid_argument ("Character set already defined for this DIMENSION/CHARS/FINAL/DIRECTION combo", existing_charset); if (!NILP (existing_charset)) /* Reuse same leading byte */ id = XCHARSET_ID (existing_charset); else id = get_unallocated_leading_byte (dimension); if (temporary) { Intbyte tempname[80]; qxesprintf (tempname, "___temporary___%d__", id); name = intern_int (tempname); } if (NILP (doc_string)) doc_string = build_string (""); if (NILP (registry)) registry = build_string (""); if (NILP (short_name)) short_name = XSYMBOL (name)->name; if (NILP (long_name)) long_name = doc_string; if (columns == -1) columns = dimension; charset = make_charset (id, name, dimension + 2, type, columns, graphic, final, direction, short_name, long_name, doc_string, registry, !NILP (existing_charset)); XCHARSET (charset)->temporary = temporary; if (!NILP (ccl_program)) XCHARSET_CCL_PROGRAM (charset) = ccl_program; { Lisp_Object revdircs = charset_by_attributes (type, final, direction == CHARSET_LEFT_TO_RIGHT ? CHARSET_RIGHT_TO_LEFT : CHARSET_LEFT_TO_RIGHT); if (!NILP (revdircs)) { XCHARSET_REVERSE_DIRECTION_CHARSET (revdircs) = charset; XCHARSET_REVERSE_DIRECTION_CHARSET (charset) = revdircs; } } return charset; } DEFUN ("make-reverse-direction-charset", Fmake_reverse_direction_charset, 2, 2, 0, /* Make a charset equivalent to CHARSET but which goes in the opposite direction. NEW-NAME is the name of the new charset. Return the new charset. */ (charset, new_name)) { Lisp_Object new_charset = Qnil; int id, dimension, columns, graphic; Intbyte final; int direction, type; Lisp_Object registry, doc_string, short_name, long_name; Lisp_Charset *cs; charset = Fget_charset (charset); if (!NILP (XCHARSET_REVERSE_DIRECTION_CHARSET (charset))) invalid_operation ("Charset already has reverse-direction charset", charset); CHECK_SYMBOL (new_name); if (!NILP (Ffind_charset (new_name))) invalid_operation ("Cannot redefine existing charset", new_name); cs = XCHARSET (charset); type = CHARSET_TYPE (cs); columns = CHARSET_COLUMNS (cs); dimension = CHARSET_DIMENSION (cs); id = get_unallocated_leading_byte (dimension); graphic = CHARSET_GRAPHIC (cs); final = CHARSET_FINAL (cs); direction = CHARSET_RIGHT_TO_LEFT; if (CHARSET_DIRECTION (cs) == CHARSET_RIGHT_TO_LEFT) direction = CHARSET_LEFT_TO_RIGHT; doc_string = CHARSET_DOC_STRING (cs); short_name = CHARSET_SHORT_NAME (cs); long_name = CHARSET_LONG_NAME (cs); registry = CHARSET_REGISTRY (cs); new_charset = make_charset (id, new_name, dimension + 2, type, columns, graphic, final, direction, short_name, long_name, doc_string, registry, 0); CHARSET_REVERSE_DIRECTION_CHARSET (cs) = new_charset; XCHARSET_REVERSE_DIRECTION_CHARSET (new_charset) = charset; return new_charset; } DEFUN ("charset-reverse-direction-charset", Fcharset_reverse_direction_charset, 1, 1, 0, /* Return the reverse-direction charset parallel to CHARSET, if any. This is the charset with the same properties (in particular, the same dimension, number of characters per dimension, and final byte) as CHARSET but whose characters are displayed in the opposite direction. */ (charset)) { charset = Fget_charset (charset); return XCHARSET_REVERSE_DIRECTION_CHARSET (charset); } DEFUN ("charset-from-attributes", Fcharset_from_attributes, 3, 4, 0, /* Return a charset with the given DIMENSION, CHARS, FINAL, and DIRECTION. If DIRECTION is omitted, both directions will be checked (left-to-right will be returned if character sets exist for both directions). */ (dimension, chars, final, direction)) { int dm, ch, fi, di = -1; int type; Lisp_Object obj = Qnil; CHECK_INT (dimension); dm = XINT (dimension); if (dm < 1 || dm > 2) invalid_constant ("Invalid value for DIMENSION", dimension); CHECK_INT (chars); ch = XINT (chars); if (ch != 94 && ch != 96) invalid_constant ("Invalid value for CHARS", chars); CHECK_CHAR_COERCE_INT (final); fi = XCHAR (final); if (fi < '0' || fi > '~') invalid_constant ("Invalid value for FINAL", final); if (EQ (direction, Ql2r)) di = CHARSET_LEFT_TO_RIGHT; else if (EQ (direction, Qr2l)) di = CHARSET_RIGHT_TO_LEFT; else if (!NILP (direction)) invalid_constant ("Invalid value for DIRECTION", direction); if (dm == 2 && fi > 0x5F) invalid_constant ("Final must be in the range 0x30 - 0x5F for dimension == 2", final); if (dm == 1) type = (ch == 94) ? CHARSET_TYPE_94 : CHARSET_TYPE_96; else type = (ch == 94) ? CHARSET_TYPE_94X94 : CHARSET_TYPE_96X96; if (di == -1) { obj = charset_by_attributes (type, fi, CHARSET_LEFT_TO_RIGHT); if (NILP (obj)) obj = charset_by_attributes (type, fi, CHARSET_RIGHT_TO_LEFT); } else obj = charset_by_attributes (type, fi, di); if (CHARSETP (obj)) return XCHARSET_NAME (obj); return obj; } DEFUN ("charset-short-name", Fcharset_short_name, 1, 1, 0, /* Return short name of CHARSET. */ (charset)) { return XCHARSET_SHORT_NAME (Fget_charset (charset)); } DEFUN ("charset-long-name", Fcharset_long_name, 1, 1, 0, /* Return long name of CHARSET. */ (charset)) { return XCHARSET_LONG_NAME (Fget_charset (charset)); } DEFUN ("charset-description", Fcharset_description, 1, 1, 0, /* Return description of CHARSET. */ (charset)) { return XCHARSET_DOC_STRING (Fget_charset (charset)); } DEFUN ("charset-dimension", Fcharset_dimension, 1, 1, 0, /* Return dimension of CHARSET. */ (charset)) { return make_int (XCHARSET_DIMENSION (Fget_charset (charset))); } DEFUN ("charset-property", Fcharset_property, 2, 2, 0, /* Return property PROP of CHARSET, a charset object or symbol naming a charset. Recognized properties are those listed in `make-charset', as well as 'name and 'doc-string. */ (charset, prop)) { Lisp_Charset *cs; charset = Fget_charset (charset); cs = XCHARSET (charset); CHECK_SYMBOL (prop); if (EQ (prop, Qname)) return CHARSET_NAME (cs); if (EQ (prop, Qshort_name)) return CHARSET_SHORT_NAME (cs); if (EQ (prop, Qlong_name)) return CHARSET_LONG_NAME (cs); if (EQ (prop, Qdoc_string)) return CHARSET_DOC_STRING (cs); if (EQ (prop, Qdimension)) return make_int (CHARSET_DIMENSION (cs)); if (EQ (prop, Qcolumns)) return make_int (CHARSET_COLUMNS (cs)); if (EQ (prop, Qgraphic)) return make_int (CHARSET_GRAPHIC (cs)); if (EQ (prop, Qfinal)) return make_char (CHARSET_FINAL (cs)); if (EQ (prop, Qchars)) return make_int (CHARSET_CHARS (cs)); if (EQ (prop, Qregistry)) return CHARSET_REGISTRY (cs); if (EQ (prop, Qccl_program)) return CHARSET_CCL_PROGRAM (cs); if (EQ (prop, Qdirection)) return CHARSET_DIRECTION (cs) == CHARSET_LEFT_TO_RIGHT ? Ql2r : Qr2l; if (EQ (prop, Qreverse_direction_charset)) { Lisp_Object obj = CHARSET_REVERSE_DIRECTION_CHARSET (cs); /* #### Is this translation OK? If so, error checking sufficient? */ return CHARSETP (obj) ? XCHARSET_NAME (obj) : obj; } invalid_constant ("Unrecognized charset property name", prop); RETURN_NOT_REACHED (Qnil) } DEFUN ("charset-id", Fcharset_id, 1, 1, 0, /* Return charset identification number of CHARSET. */ (charset)) { return make_int (XCHARSET_LEADING_BYTE (Fget_charset (charset))); } /* #### We need to figure out which properties we really want to allow to be set. */ DEFUN ("set-charset-ccl-program", Fset_charset_ccl_program, 2, 2, 0, /* Set the 'ccl-program property of CHARSET to CCL-PROGRAM. */ (charset, ccl_program)) { struct ccl_program test_ccl; charset = Fget_charset (charset); if (setup_ccl_program (&test_ccl, ccl_program) < 0) invalid_argument ("Invalid ccl-program", ccl_program); XCHARSET_CCL_PROGRAM (charset) = ccl_program; face_property_was_changed (Vdefault_face, Qfont, Qglobal); return Qnil; } static void invalidate_charset_font_caches (Lisp_Object charset) { /* Invalidate font cache entries for charset on all devices. */ Lisp_Object devcons, concons, hash_table; DEVICE_LOOP_NO_BREAK (devcons, concons) { struct device *d = XDEVICE (XCAR (devcons)); hash_table = Fgethash (charset, d->charset_font_cache, Qunbound); if (!UNBOUNDP (hash_table)) Fclrhash (hash_table); } } /* Japanese folks may want to (set-charset-registry 'ascii "jisx0201") */ DEFUN ("set-charset-registry", Fset_charset_registry, 2, 2, 0, /* Set the 'registry property of CHARSET to REGISTRY. */ (charset, registry)) { charset = Fget_charset (charset); CHECK_STRING (registry); XCHARSET_REGISTRY (charset) = registry; invalidate_charset_font_caches (charset); face_property_was_changed (Vdefault_face, Qfont, Qglobal); return Qnil; } /************************************************************************/ /* memory usage */ /************************************************************************/ #ifdef MEMORY_USAGE_STATS struct charset_stats { int from_unicode; int to_unicode; int other; }; static void compute_charset_usage (Lisp_Object charset, struct charset_stats *stats, struct overhead_stats *ovstats) { struct Lisp_Charset *c = XCHARSET (charset); xzero (*stats); stats->other += malloced_storage_size (c, sizeof (*c), ovstats); stats->from_unicode += compute_from_unicode_table_size (charset, ovstats); stats->to_unicode += compute_to_unicode_table_size (charset, ovstats); } DEFUN ("charset-memory-usage", Fcharset_memory_usage, 1, 1, 0, /* Return stats about the memory usage of charset CHARSET. The values returned are in the form of an alist of usage types and byte counts. The byte counts attempt to encompass all the memory used by the charset (separate from the memory logically associated with a charset or frame), including internal structures and any malloc() overhead associated with them. In practice, the byte counts are underestimated for various reasons, e.g. because certain memory usage is very hard to determine \(e.g. the amount of memory used inside the Xt library or inside the X server). Multiple slices of the total memory usage may be returned, separated by a nil. Each slice represents a particular view of the memory, a particular way of partitioning it into groups. Within a slice, there is no overlap between the groups of memory, and each slice collectively represents all the memory concerned. */ (charset)) { struct charset_stats stats; struct overhead_stats ovstats; Lisp_Object val = Qnil; charset = Fget_charset (charset); xzero (ovstats); compute_charset_usage (charset, &stats, &ovstats); val = acons (Qfrom_unicode, make_int (stats.from_unicode), val); val = acons (Qto_unicode, make_int (stats.to_unicode), val); val = Fcons (Qnil, val); val = acons (Qactually_requested, make_int (ovstats.was_requested), val); val = acons (Qmalloc_overhead, make_int (ovstats.malloc_overhead), val); val = acons (Qgap_overhead, make_int (ovstats.gap_overhead), val); val = acons (Qdynarr_overhead, make_int (ovstats.dynarr_overhead), val); return Fnreverse (val); } #endif /* MEMORY_USAGE_STATS */ /************************************************************************/ /* initialization */ /************************************************************************/ void syms_of_mule_charset (void) { INIT_LRECORD_IMPLEMENTATION (charset); DEFSUBR (Fcharsetp); DEFSUBR (Ffind_charset); DEFSUBR (Fget_charset); DEFSUBR (Fcharset_list); DEFSUBR (Fcharset_name); DEFSUBR (Fmake_charset); DEFSUBR (Fmake_reverse_direction_charset); DEFSUBR (Fcharset_reverse_direction_charset); DEFSUBR (Fcharset_from_attributes); DEFSUBR (Fcharset_short_name); DEFSUBR (Fcharset_long_name); DEFSUBR (Fcharset_description); DEFSUBR (Fcharset_dimension); DEFSUBR (Fcharset_property); DEFSUBR (Fcharset_id); DEFSUBR (Fset_charset_ccl_program); DEFSUBR (Fset_charset_registry); #ifdef MEMORY_USAGE_STATS DEFSUBR (Fcharset_memory_usage); #endif DEFSYMBOL (Qcharsetp); DEFSYMBOL (Qregistry); DEFSYMBOL (Qfinal); DEFSYMBOL (Qgraphic); DEFSYMBOL (Qdirection); DEFSYMBOL (Qreverse_direction_charset); DEFSYMBOL (Qshort_name); DEFSYMBOL (Qlong_name); DEFSYMBOL (Qfrom_unicode); DEFSYMBOL (Qto_unicode); DEFSYMBOL (Ql2r); DEFSYMBOL (Qr2l); /* Charsets, compatible with FSF 20.3 Naming convention is Script-Charset[-Edition] */ DEFSYMBOL (Qlatin_iso8859_1); DEFSYMBOL (Qlatin_iso8859_2); DEFSYMBOL (Qlatin_iso8859_3); DEFSYMBOL (Qlatin_iso8859_4); DEFSYMBOL (Qthai_tis620); DEFSYMBOL (Qgreek_iso8859_7); DEFSYMBOL (Qarabic_iso8859_6); DEFSYMBOL (Qhebrew_iso8859_8); DEFSYMBOL (Qkatakana_jisx0201); DEFSYMBOL (Qlatin_jisx0201); DEFSYMBOL (Qcyrillic_iso8859_5); DEFSYMBOL (Qlatin_iso8859_9); DEFSYMBOL (Qlatin_iso8859_15); DEFSYMBOL (Qjapanese_jisx0208_1978); DEFSYMBOL (Qchinese_gb2312); DEFSYMBOL (Qjapanese_jisx0208); DEFSYMBOL (Qkorean_ksc5601); DEFSYMBOL (Qjapanese_jisx0212); DEFSYMBOL (Qchinese_cns11643_1); DEFSYMBOL (Qchinese_cns11643_2); DEFSYMBOL (Qchinese_big5_1); DEFSYMBOL (Qchinese_big5_2); DEFSYMBOL (Qcomposite); } static int init_charset_unicode_tables_mapper (Lisp_Object key, Lisp_Object value, void *closure) { init_charset_unicode_tables (value); return 0; } void init_mule_charset (void) { /* See mule-charset.h, definition of Lisp_Charset. */ if (initialized) elisp_maphash (init_charset_unicode_tables_mapper, Vcharset_hash_table, 0); } void vars_of_mule_charset (void) { int i, j, k; chlook = xnew_and_zero (struct charset_lookup); /* zero for Purify. */ dump_add_root_struct_ptr (&chlook, &charset_lookup_description); /* Table of charsets indexed by leading byte. */ for (i = 0; i < countof (chlook->charset_by_leading_byte); i++) chlook->charset_by_leading_byte[i] = Qnil; /* Table of charsets indexed by type/final-byte/direction. */ for (i = 0; i < countof (chlook->charset_by_attributes); i++) for (j = 0; j < countof (chlook->charset_by_attributes[0]); j++) for (k = 0; k < countof (chlook->charset_by_attributes[0][0]); k++) chlook->charset_by_attributes[i][j][k] = Qnil; chlook->next_allocated_1_byte_leading_byte = MIN_LEADING_BYTE_PRIVATE_1; chlook->next_allocated_2_byte_leading_byte = MIN_LEADING_BYTE_PRIVATE_2; staticpro (&Vcharset_hash_table); Vcharset_hash_table = make_lisp_hash_table (50, HASH_TABLE_NON_WEAK, HASH_TABLE_EQ); } void complex_vars_of_mule_charset (void) { /* Predefined character sets. We store them into variables for ease of access. */ staticpro (&Vcharset_ascii); Vcharset_ascii = make_charset (LEADING_BYTE_ASCII, Qascii, 1, CHARSET_TYPE_94, 1, 0, 'B', CHARSET_LEFT_TO_RIGHT, build_string ("ASCII"), build_msg_string ("ASCII"), build_msg_string ("ASCII (ISO646 IRV)"), build_string ("\\(iso8859-[0-9]*\\|-ascii\\)"), 0); staticpro (&Vcharset_control_1); Vcharset_control_1 = make_charset (LEADING_BYTE_CONTROL_1, Qcontrol_1, 2, CHARSET_TYPE_94, 1, 1, 0, CHARSET_LEFT_TO_RIGHT, build_string ("C1"), build_msg_string ("Control characters"), build_msg_string ("Control characters 128-191"), build_string (""), 0); staticpro (&Vcharset_latin_iso8859_1); Vcharset_latin_iso8859_1 = make_charset (LEADING_BYTE_LATIN_ISO8859_1, Qlatin_iso8859_1, 2, CHARSET_TYPE_96, 1, 1, 'A', CHARSET_LEFT_TO_RIGHT, build_string ("Latin-1"), build_msg_string ("ISO8859-1 (Latin-1)"), build_msg_string ("ISO8859-1 (Latin-1)"), build_string ("iso8859-1"), 0); staticpro (&Vcharset_latin_iso8859_2); Vcharset_latin_iso8859_2 = make_charset (LEADING_BYTE_LATIN_ISO8859_2, Qlatin_iso8859_2, 2, CHARSET_TYPE_96, 1, 1, 'B', CHARSET_LEFT_TO_RIGHT, build_string ("Latin-2"), build_msg_string ("ISO8859-2 (Latin-2)"), build_msg_string ("ISO8859-2 (Latin-2)"), build_string ("iso8859-2"), 0); staticpro (&Vcharset_latin_iso8859_3); Vcharset_latin_iso8859_3 = make_charset (LEADING_BYTE_LATIN_ISO8859_3, Qlatin_iso8859_3, 2, CHARSET_TYPE_96, 1, 1, 'C', CHARSET_LEFT_TO_RIGHT, build_string ("Latin-3"), build_msg_string ("ISO8859-3 (Latin-3)"), build_msg_string ("ISO8859-3 (Latin-3)"), build_string ("iso8859-3"), 0); staticpro (&Vcharset_latin_iso8859_4); Vcharset_latin_iso8859_4 = make_charset (LEADING_BYTE_LATIN_ISO8859_4, Qlatin_iso8859_4, 2, CHARSET_TYPE_96, 1, 1, 'D', CHARSET_LEFT_TO_RIGHT, build_string ("Latin-4"), build_msg_string ("ISO8859-4 (Latin-4)"), build_msg_string ("ISO8859-4 (Latin-4)"), build_string ("iso8859-4"), 0); staticpro (&Vcharset_thai_tis620); Vcharset_thai_tis620 = make_charset (LEADING_BYTE_THAI_TIS620, Qthai_tis620, 2, CHARSET_TYPE_96, 1, 1, 'T', CHARSET_LEFT_TO_RIGHT, build_string ("TIS620"), build_msg_string ("TIS620 (Thai)"), build_msg_string ("TIS620.2529 (Thai)"), build_string ("tis620"),0); staticpro (&Vcharset_greek_iso8859_7); Vcharset_greek_iso8859_7 = make_charset (LEADING_BYTE_GREEK_ISO8859_7, Qgreek_iso8859_7, 2, CHARSET_TYPE_96, 1, 1, 'F', CHARSET_LEFT_TO_RIGHT, build_string ("ISO8859-7"), build_msg_string ("ISO8859-7 (Greek)"), build_msg_string ("ISO8859-7 (Greek)"), build_string ("iso8859-7"), 0); staticpro (&Vcharset_arabic_iso8859_6); Vcharset_arabic_iso8859_6 = make_charset (LEADING_BYTE_ARABIC_ISO8859_6, Qarabic_iso8859_6, 2, CHARSET_TYPE_96, 1, 1, 'G', CHARSET_RIGHT_TO_LEFT, build_string ("ISO8859-6"), build_msg_string ("ISO8859-6 (Arabic)"), build_msg_string ("ISO8859-6 (Arabic)"), build_string ("iso8859-6"), 0); staticpro (&Vcharset_hebrew_iso8859_8); Vcharset_hebrew_iso8859_8 = make_charset (LEADING_BYTE_HEBREW_ISO8859_8, Qhebrew_iso8859_8, 2, CHARSET_TYPE_96, 1, 1, 'H', CHARSET_RIGHT_TO_LEFT, build_string ("ISO8859-8"), build_msg_string ("ISO8859-8 (Hebrew)"), build_msg_string ("ISO8859-8 (Hebrew)"), build_string ("iso8859-8"), 0); staticpro (&Vcharset_katakana_jisx0201); Vcharset_katakana_jisx0201 = make_charset (LEADING_BYTE_KATAKANA_JISX0201, Qkatakana_jisx0201, 2, CHARSET_TYPE_94, 1, 1, 'I', CHARSET_LEFT_TO_RIGHT, build_string ("JISX0201 Kana"), build_msg_string ("JISX0201.1976 (Japanese Kana)"), build_msg_string ("JISX0201.1976 Japanese Kana"), build_string ("jisx0201.1976"), 0); staticpro (&Vcharset_latin_jisx0201); Vcharset_latin_jisx0201 = make_charset (LEADING_BYTE_LATIN_JISX0201, Qlatin_jisx0201, 2, CHARSET_TYPE_94, 1, 0, 'J', CHARSET_LEFT_TO_RIGHT, build_string ("JISX0201 Roman"), build_msg_string ("JISX0201.1976 (Japanese Roman)"), build_msg_string ("JISX0201.1976 Japanese Roman"), build_string ("jisx0201.1976"), 0); staticpro (&Vcharset_cyrillic_iso8859_5); Vcharset_cyrillic_iso8859_5 = make_charset (LEADING_BYTE_CYRILLIC_ISO8859_5, Qcyrillic_iso8859_5, 2, CHARSET_TYPE_96, 1, 1, 'L', CHARSET_LEFT_TO_RIGHT, build_string ("ISO8859-5"), build_msg_string ("ISO8859-5 (Cyrillic)"), build_msg_string ("ISO8859-5 (Cyrillic)"), build_string ("iso8859-5"), 0); staticpro (&Vcharset_latin_iso8859_9); Vcharset_latin_iso8859_9 = make_charset (LEADING_BYTE_LATIN_ISO8859_9, Qlatin_iso8859_9, 2, CHARSET_TYPE_96, 1, 1, 'M', CHARSET_LEFT_TO_RIGHT, build_string ("Latin-5"), build_msg_string ("ISO8859-9 (Latin-5)"), build_msg_string ("ISO8859-9 (Latin-5)"), build_string ("iso8859-9"), 0); staticpro (&Vcharset_latin_iso8859_15); Vcharset_latin_iso8859_15 = make_charset (LEADING_BYTE_LATIN_ISO8859_15, Qlatin_iso8859_15, 2, CHARSET_TYPE_96, 1, 1, 'b', CHARSET_LEFT_TO_RIGHT, build_string ("Latin-9"), build_msg_string ("ISO8859-15 (Latin-9)"), build_msg_string ("ISO8859-15 (Latin-9)"), build_string ("iso8859-15"), 0); staticpro (&Vcharset_japanese_jisx0208_1978); Vcharset_japanese_jisx0208_1978 = make_charset (LEADING_BYTE_JAPANESE_JISX0208_1978, Qjapanese_jisx0208_1978, 3, CHARSET_TYPE_94X94, 2, 0, '@', CHARSET_LEFT_TO_RIGHT, build_string ("JISX0208.1978"), build_msg_string ("JISX0208.1978 (Japanese)"), build_msg_string ("JISX0208.1978 Japanese Kanji (so called \"old JIS\")"), build_string ("\\(jisx0208\\|jisc6226\\)\\.1978"), 0); staticpro (&Vcharset_chinese_gb2312); Vcharset_chinese_gb2312 = make_charset (LEADING_BYTE_CHINESE_GB2312, Qchinese_gb2312, 3, CHARSET_TYPE_94X94, 2, 0, 'A', CHARSET_LEFT_TO_RIGHT, build_string ("GB2312"), build_msg_string ("GB2312)"), build_msg_string ("GB2312 Chinese simplified"), build_string ("gb2312"), 0); staticpro (&Vcharset_japanese_jisx0208); Vcharset_japanese_jisx0208 = make_charset (LEADING_BYTE_JAPANESE_JISX0208, Qjapanese_jisx0208, 3, CHARSET_TYPE_94X94, 2, 0, 'B', CHARSET_LEFT_TO_RIGHT, build_string ("JISX0208"), build_msg_string ("JISX0208.1983/1990 (Japanese)"), build_msg_string ("JISX0208.1983/1990 Japanese Kanji"), build_string ("jisx0208.19\\(83\\|90\\)"), 0); staticpro (&Vcharset_korean_ksc5601); Vcharset_korean_ksc5601 = make_charset (LEADING_BYTE_KOREAN_KSC5601, Qkorean_ksc5601, 3, CHARSET_TYPE_94X94, 2, 0, 'C', CHARSET_LEFT_TO_RIGHT, build_string ("KSC5601"), build_msg_string ("KSC5601 (Korean"), build_msg_string ("KSC5601 Korean Hangul and Hanja"), build_string ("ksc5601"), 0); staticpro (&Vcharset_japanese_jisx0212); Vcharset_japanese_jisx0212 = make_charset (LEADING_BYTE_JAPANESE_JISX0212, Qjapanese_jisx0212, 3, CHARSET_TYPE_94X94, 2, 0, 'D', CHARSET_LEFT_TO_RIGHT, build_string ("JISX0212"), build_msg_string ("JISX0212 (Japanese)"), build_msg_string ("JISX0212 Japanese Supplement"), build_string ("jisx0212"), 0); #define CHINESE_CNS_PLANE_RE(n) "cns11643[.-]\\(.*[.-]\\)?" n "$" staticpro (&Vcharset_chinese_cns11643_1); Vcharset_chinese_cns11643_1 = make_charset (LEADING_BYTE_CHINESE_CNS11643_1, Qchinese_cns11643_1, 3, CHARSET_TYPE_94X94, 2, 0, 'G', CHARSET_LEFT_TO_RIGHT, build_string ("CNS11643-1"), build_msg_string ("CNS11643-1 (Chinese traditional)"), build_msg_string ("CNS 11643 Plane 1 Chinese traditional"), build_string (CHINESE_CNS_PLANE_RE("1")), 0); staticpro (&Vcharset_chinese_cns11643_2); Vcharset_chinese_cns11643_2 = make_charset (LEADING_BYTE_CHINESE_CNS11643_2, Qchinese_cns11643_2, 3, CHARSET_TYPE_94X94, 2, 0, 'H', CHARSET_LEFT_TO_RIGHT, build_string ("CNS11643-2"), build_msg_string ("CNS11643-2 (Chinese traditional)"), build_msg_string ("CNS 11643 Plane 2 Chinese traditional"), build_string (CHINESE_CNS_PLANE_RE("2")), 0); staticpro (&Vcharset_chinese_big5_1); Vcharset_chinese_big5_1 = make_charset (LEADING_BYTE_CHINESE_BIG5_1, Qchinese_big5_1, 3, CHARSET_TYPE_94X94, 2, 0, '0', CHARSET_LEFT_TO_RIGHT, build_string ("Big5"), build_msg_string ("Big5 (Level-1)"), build_msg_string ("Big5 Level-1 Chinese traditional"), build_string ("big5"), 0); staticpro (&Vcharset_chinese_big5_2); Vcharset_chinese_big5_2 = make_charset (LEADING_BYTE_CHINESE_BIG5_2, Qchinese_big5_2, 3, CHARSET_TYPE_94X94, 2, 0, '1', CHARSET_LEFT_TO_RIGHT, build_string ("Big5"), build_msg_string ("Big5 (Level-2)"), build_msg_string ("Big5 Level-2 Chinese traditional"), build_string ("big5"), 0); #ifdef ENABLE_COMPOSITE_CHARS /* #### For simplicity, we put composite chars into a 96x96 charset. This is going to lead to problems because you can run out of room, esp. as we don't yet recycle numbers. */ staticpro (&Vcharset_composite); Vcharset_composite = make_charset (LEADING_BYTE_COMPOSITE, Qcomposite, 3, CHARSET_TYPE_96X96, 2, 0, 0, CHARSET_LEFT_TO_RIGHT, build_string ("Composite"), build_msg_string ("Composite characters"), build_msg_string ("Composite characters"), build_string (""), 0); #else /* We create a hack so that we have a way of storing ESC 0 and ESC 1 sequences as "characters", so that they will be output correctly. */ staticpro (&Vcharset_composite); Vcharset_composite = make_charset (LEADING_BYTE_COMPOSITE_REPLACEMENT, Qcomposite, 2, CHARSET_TYPE_96, 1, 1, '|', CHARSET_LEFT_TO_RIGHT, build_string ("Composite hack"), build_msg_string ("Composite characters hack"), build_msg_string ("Composite characters hack"), build_string (""), 0); #endif /* ENABLE_COMPOSITE_CHARS */ }