1/* Conversion to and from IBM939.
2 Copyright (C) 2000-2021 Free Software Foundation, Inc.
3 This file is part of the GNU C Library.
4 Contributed by Masahide Washizawa <washi@yamato.ibm.co.jp>, 2000.
5
6 The GNU C Library is free software; you can redistribute it and/or
7 modify it under the terms of the GNU Lesser General Public
8 License as published by the Free Software Foundation; either
9 version 2.1 of the License, or (at your option) any later version.
10
11 The GNU C Library is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 Lesser General Public License for more details.
15
16 You should have received a copy of the GNU Lesser General Public
17 License along with the GNU C Library; if not, see
18 <https://www.gnu.org/licenses/>. */
19
20/* IBM939 is designed for the representation of Japanese Latin/Kanji
21 using a stateful EBCDIC encoding scheme. It is also known as
22 CCSID 939 or CP939. See:
23 https://www-01.ibm.com/software/globalization/ccsid/ccsid939.html */
24
25#include <dlfcn.h>
26#include <stdint.h>
27#include <wchar.h>
28#include <byteswap.h>
29#include "ibm939.h"
30
31/* The shift sequences for this charset (it does not use ESC). */
32#define SI 0x0F /* Shift In, host code to turn DBCS off. */
33#define SO 0x0E /* Shift Out, host code to turn DBCS on. */
34
35/* Definitions used in the body of the `gconv' function. */
36#define CHARSET_NAME "IBM939//"
37#define FROM_LOOP from_ibm939
38#define TO_LOOP to_ibm939
39#define ONE_DIRECTION 0
40#define FROM_LOOP_MIN_NEEDED_FROM 1
41#define FROM_LOOP_MAX_NEEDED_FROM 2
42#define FROM_LOOP_MIN_NEEDED_TO 4
43#define FROM_LOOP_MAX_NEEDED_TO 4
44#define TO_LOOP_MIN_NEEDED_FROM 4
45#define TO_LOOP_MAX_NEEDED_FROM 4
46#define TO_LOOP_MIN_NEEDED_TO 1
47#define TO_LOOP_MAX_NEEDED_TO 3
48#define PREPARE_LOOP \
49 int save_curcs; \
50 int *curcsp = &data->__statep->__count;
51#define EXTRA_LOOP_ARGS , curcsp
52
53/* Definitions of initialization and destructor function. */
54#define DEFINE_INIT 1
55#define DEFINE_FINI 1
56
57
58/* Since this is a stateful encoding we have to provide code which resets
59 the output state to the initial state. This has to be done during the
60 flushing. */
61#define EMIT_SHIFT_TO_INIT \
62 if ((data->__statep->__count & ~7) != sb) \
63 { \
64 if (FROM_DIRECTION) \
65 data->__statep->__count &= 7; \
66 else \
67 { \
68 /* We are not in the initial state. To switch back we have \
69 to emit `SI'. */ \
70 if (__glibc_unlikely (outbuf >= outend)) \
71 /* We don't have enough room in the output buffer. */ \
72 status = __GCONV_FULL_OUTPUT; \
73 else \
74 { \
75 /* Write out the shift sequence. */ \
76 *outbuf++ = SI; \
77 data->__statep->__count &= 7; \
78 } \
79 } \
80 }
81
82
83/* Since we might have to reset input pointer we must be able to save
84 and retore the state. */
85#define SAVE_RESET_STATE(Save) \
86 if (Save) \
87 save_curcs = *curcsp; \
88 else \
89 *curcsp = save_curcs
90
91
92/* Current codeset type. */
93enum
94{
95 sb = 0,
96 db = 64
97};
98
99/* First, define the conversion function from IBM-939 to UCS4. */
100#define MIN_NEEDED_INPUT FROM_LOOP_MIN_NEEDED_FROM
101#define MAX_NEEDED_INPUT FROM_LOOP_MAX_NEEDED_FROM
102#define MIN_NEEDED_OUTPUT FROM_LOOP_MIN_NEEDED_TO
103#define MAX_NEEDED_OUTPUT FROM_LOOP_MAX_NEEDED_TO
104#define LOOPFCT FROM_LOOP
105#define BODY \
106 { \
107 uint32_t ch = *inptr; \
108 uint32_t res; \
109 \
110 if (__builtin_expect (ch, 0) == SO) \
111 { \
112 /* Shift OUT, change to DBCS converter (redundant escape okay). */ \
113 curcs = db; \
114 ++inptr; \
115 continue; \
116 } \
117 else if (__builtin_expect (ch, 0) == SI) \
118 { \
119 /* Shift IN, change to SBCS converter (redundant escape okay). */ \
120 curcs = sb; \
121 ++inptr; \
122 continue; \
123 } \
124 \
125 if (curcs == sb) \
126 { \
127 /* Use the IBM939 table for single byte. */ \
128 res = __ibm939sb_to_ucs4[ch]; \
129 if (__builtin_expect (res == L'\0', 0) && ch != '\0') \
130 { \
131 /* This is an illegal character. */ \
132 STANDARD_FROM_LOOP_ERR_HANDLER (1); \
133 } \
134 else \
135 { \
136 put32 (outptr, res); \
137 outptr += 4; \
138 } \
139 ++inptr; \
140 } \
141 else \
142 { \
143 /* Use the IBM939 table for double byte. */ \
144 const struct gap *rp2 = __ibm939db_to_ucs4_idx; \
145 \
146 assert (curcs == db); \
147 \
148 if (__glibc_unlikely (inptr + 1 >= inend)) \
149 { \
150 /* The second character is not available. Store the \
151 intermediate result. */ \
152 result = __GCONV_INCOMPLETE_INPUT; \
153 break; \
154 } \
155 \
156 ch = (ch * 0x100) + inptr[1]; \
157 while (ch > rp2->end) \
158 ++rp2; \
159 \
160 if (__builtin_expect (rp2->start == 0xffff, 0) \
161 || __builtin_expect (ch < rp2->start, 0) \
162 || (res = __ibm939db_to_ucs4[ch + rp2->idx], \
163 __builtin_expect (res, L'\1') == L'\0' && ch != '\0')) \
164 { \
165 /* This is an illegal character. */ \
166 STANDARD_FROM_LOOP_ERR_HANDLER (2); \
167 } \
168 else \
169 { \
170 put32 (outptr, res); \
171 outptr += 4; \
172 } \
173 inptr += 2; \
174 } \
175 }
176#define LOOP_NEED_FLAGS
177#define EXTRA_LOOP_DECLS , int *curcsp
178#define INIT_PARAMS int curcs = *curcsp & ~7
179#define UPDATE_PARAMS *curcsp = curcs
180#include <iconv/loop.c>
181
182/* Next, define the other direction */
183#define MIN_NEEDED_INPUT TO_LOOP_MIN_NEEDED_FROM
184#define MAX_NEEDED_INPUT TO_LOOP_MAX_NEEDED_FROM
185#define MIN_NEEDED_OUTPUT TO_LOOP_MIN_NEEDED_TO
186#define MAX_NEEDED_OUTPUT TO_LOOP_MAX_NEEDED_TO
187#define LOOPFCT TO_LOOP
188#define BODY \
189 { \
190 uint32_t ch = get32 (inptr); \
191 const struct gap *rp1 = __ucs4_to_ibm939sb_idx; \
192 const struct gap *rp2 = __ucs4_to_ibm939db_idx; \
193 \
194 if (__glibc_unlikely (ch >= 0xffff)) \
195 { \
196 UNICODE_TAG_HANDLER (ch, 4); \
197 goto ibm939_invalid_char; \
198 } \
199 \
200 while (ch > rp1->end) \
201 ++rp1; \
202 \
203 /* Use the UCS4 table for single byte. */ \
204 unsigned char sbconv; \
205 if (__builtin_expect (ch < rp1->start, 0) \
206 || (sbconv = __ucs4_to_ibm939sb[ch + rp1->idx], \
207 __builtin_expect (sbconv, L'\1') == L'\0' && ch != '\0')) \
208 { \
209 /* Use the UCS4 table for double byte. */ \
210 while (ch > rp2->end) \
211 ++rp2; \
212 \
213 const char *cp; \
214 if (__builtin_expect (ch < rp2->start, 0) \
215 || (cp = __ucs4_to_ibm939db[ch + rp2->idx], \
216 __builtin_expect (cp[0], L'\1')==L'\0' && ch != '\0')) \
217 { \
218 /* This is an illegal character. */ \
219 ibm939_invalid_char: \
220 STANDARD_TO_LOOP_ERR_HANDLER (4); \
221 } \
222 else \
223 { \
224 if (curcs == sb) \
225 { \
226 if (__glibc_unlikely (outptr + 1 > outend)) \
227 { \
228 result = __GCONV_FULL_OUTPUT; \
229 break; \
230 } \
231 *outptr++ = SO; \
232 curcs = db; \
233 } \
234 \
235 if (__glibc_unlikely (outptr + 2 > outend)) \
236 { \
237 result = __GCONV_FULL_OUTPUT; \
238 break; \
239 } \
240 *outptr++ = cp[0]; \
241 *outptr++ = cp[1]; \
242 } \
243 } \
244 else \
245 { \
246 if (curcs == db) \
247 { \
248 if (__glibc_unlikely (outptr + 1 > outend)) \
249 { \
250 result = __GCONV_FULL_OUTPUT; \
251 break; \
252 } \
253 *outptr++ = SI; \
254 curcs = sb; \
255 } \
256 \
257 if (__glibc_unlikely (outptr + 1 > outend)) \
258 { \
259 result = __GCONV_FULL_OUTPUT; \
260 break; \
261 } \
262 if (ch == 0x7e) \
263 *outptr++ = 0xa0; \
264 else if (ch == 0x5c) \
265 *outptr++ = 0xb2; \
266 else \
267 *outptr++ = sbconv; \
268 } \
269 \
270 /* Now that we wrote the output increment the input pointer. */ \
271 inptr += 4; \
272 }
273#define LOOP_NEED_FLAGS
274#define EXTRA_LOOP_DECLS , int *curcsp
275#define INIT_PARAMS int curcs = *curcsp & ~7
276#define REINIT_PARAMS curcs = *curcsp & ~7
277#define UPDATE_PARAMS *curcsp = curcs
278#include <iconv/loop.c>
279
280/* Now define the toplevel functions. */
281#include <iconv/skeleton.c>
282