1 /* Conversion to and from IBM939.
2    Copyright (C) 2000-2022 Free Software Foundation, Inc.
3    This file is part of the GNU C Library.
4 
5    The GNU C Library is free software; you can redistribute it and/or
6    modify it under the terms of the GNU Lesser General Public
7    License as published by the Free Software Foundation; either
8    version 2.1 of the License, or (at your option) any later version.
9 
10    The GNU C Library is distributed in the hope that it will be useful,
11    but WITHOUT ANY WARRANTY; without even the implied warranty of
12    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
13    Lesser General Public License for more details.
14 
15    You should have received a copy of the GNU Lesser General Public
16    License along with the GNU C Library; if not, see
17    <https://www.gnu.org/licenses/>.  */
18 
19 /* IBM939 is designed for the representation of Japanese Latin/Kanji
20    using a stateful EBCDIC encoding scheme.  It is also known as
21    CCSID 939 or CP939. See:
22    https://www-01.ibm.com/software/globalization/ccsid/ccsid939.html */
23 
24 #include <dlfcn.h>
25 #include <stdint.h>
26 #include <wchar.h>
27 #include <byteswap.h>
28 #include "ibm939.h"
29 
30 /* The shift sequences for this charset (it does not use ESC).  */
31 #define SI 		0x0F  /* Shift In, host code to turn DBCS off.  */
32 #define SO 		0x0E  /* Shift Out, host code to turn DBCS on.  */
33 
34 /* Definitions used in the body of the `gconv' function.  */
35 #define CHARSET_NAME	"IBM939//"
36 #define FROM_LOOP	from_ibm939
37 #define TO_LOOP		to_ibm939
38 #define ONE_DIRECTION			0
39 #define FROM_LOOP_MIN_NEEDED_FROM	1
40 #define FROM_LOOP_MAX_NEEDED_FROM	2
41 #define FROM_LOOP_MIN_NEEDED_TO		4
42 #define FROM_LOOP_MAX_NEEDED_TO		4
43 #define TO_LOOP_MIN_NEEDED_FROM		4
44 #define TO_LOOP_MAX_NEEDED_FROM		4
45 #define TO_LOOP_MIN_NEEDED_TO		1
46 #define TO_LOOP_MAX_NEEDED_TO		3
47 #define PREPARE_LOOP \
48   int save_curcs;							      \
49   int *curcsp = &data->__statep->__count;
50 #define EXTRA_LOOP_ARGS		, curcsp
51 
52 /* Definitions of initialization and destructor function.  */
53 #define DEFINE_INIT	1
54 #define DEFINE_FINI	1
55 
56 
57 /* Since this is a stateful encoding we have to provide code which resets
58    the output state to the initial state.  This has to be done during the
59    flushing.  */
60 #define EMIT_SHIFT_TO_INIT \
61   if ((data->__statep->__count & ~7) != sb)				      \
62     {									      \
63       if (FROM_DIRECTION)						      \
64 	data->__statep->__count &= 7;					      \
65       else								      \
66 	{								      \
67 	  /* We are not in the initial state.  To switch back we have	      \
68 	     to emit `SI'.  */						      \
69 	  if (__glibc_unlikely (outbuf >= outend))			      \
70 	    /* We don't have enough room in the output buffer.  */	      \
71 	    status = __GCONV_FULL_OUTPUT;				      \
72 	  else								      \
73 	    {								      \
74 	      /* Write out the shift sequence.  */			      \
75 	      *outbuf++ = SI;						      \
76 	      data->__statep->__count &= 7;				      \
77 	    }								      \
78 	}								      \
79     }
80 
81 
82 /* Since we might have to reset input pointer we must be able to save
83    and retore the state.  */
84 #define SAVE_RESET_STATE(Save) \
85   if (Save)								      \
86     save_curcs = *curcsp;						      \
87   else									      \
88     *curcsp = save_curcs
89 
90 
91 /* Current codeset type.  */
92 enum
93 {
94   sb = 0,
95   db = 64
96 };
97 
98 /* First, define the conversion function from IBM-939 to UCS4.  */
99 #define MIN_NEEDED_INPUT	FROM_LOOP_MIN_NEEDED_FROM
100 #define MAX_NEEDED_INPUT	FROM_LOOP_MAX_NEEDED_FROM
101 #define MIN_NEEDED_OUTPUT	FROM_LOOP_MIN_NEEDED_TO
102 #define MAX_NEEDED_OUTPUT	FROM_LOOP_MAX_NEEDED_TO
103 #define LOOPFCT 		FROM_LOOP
104 #define BODY \
105   {									      \
106     uint32_t ch = *inptr;						      \
107     uint32_t res;							      \
108 									      \
109     if (__builtin_expect (ch, 0) == SO)					      \
110       {									      \
111 	/* Shift OUT, change to DBCS converter (redundant escape okay).  */   \
112 	curcs = db;							      \
113 	++inptr;							      \
114 	continue;							      \
115       }									      \
116     else if (__builtin_expect (ch, 0) == SI)				      \
117       {									      \
118 	/* Shift IN, change to SBCS converter (redundant escape okay).  */    \
119 	curcs = sb;							      \
120 	++inptr;							      \
121 	continue;							      \
122       }									      \
123 									      \
124     if (curcs == sb)							      \
125       {									      \
126 	/* Use the IBM939 table for single byte.  */			      \
127 	res = __ibm939sb_to_ucs4[ch];					      \
128 	if (__builtin_expect (res == L'\0', 0) && ch != '\0')		      \
129 	  {								      \
130 	    /* This is an illegal character.  */			      \
131 	    STANDARD_FROM_LOOP_ERR_HANDLER (1);				      \
132 	  }								      \
133 	else								      \
134 	  {								      \
135 	    put32 (outptr, res);					      \
136 	    outptr += 4;						      \
137 	  }								      \
138 	++inptr;							      \
139       }									      \
140     else								      \
141       {									      \
142 	/* Use the IBM939 table for double byte.  */			      \
143 	const struct gap *rp2 = __ibm939db_to_ucs4_idx;			      \
144 									      \
145 	assert (curcs == db);						      \
146 									      \
147 	if (__glibc_unlikely (inptr + 1 >= inend))			      \
148 	  {								      \
149 	    /* The second character is not available.  Store the	      \
150 	       intermediate result. */					      \
151 	    result = __GCONV_INCOMPLETE_INPUT;				      \
152 	    break;							      \
153 	  }								      \
154 									      \
155 	ch = (ch * 0x100) + inptr[1];					      \
156 	while (ch > rp2->end)						      \
157 	  ++rp2;							      \
158 									      \
159 	if (__builtin_expect (rp2->start == 0xffff, 0)			      \
160 	    || __builtin_expect (ch < rp2->start, 0)			      \
161 	    || (res = __ibm939db_to_ucs4[ch + rp2->idx],		      \
162 		__builtin_expect (res, L'\1') == L'\0' && ch != '\0'))	      \
163 	  {								      \
164 	    /* This is an illegal character.  */			      \
165 	    STANDARD_FROM_LOOP_ERR_HANDLER (2);				      \
166 	  }								      \
167 	else								      \
168 	  {								      \
169 	    put32 (outptr, res);					      \
170 	    outptr += 4;						      \
171 	  }								      \
172 	inptr += 2;							      \
173       }									      \
174   }
175 #define LOOP_NEED_FLAGS
176 #define EXTRA_LOOP_DECLS	, int *curcsp
177 #define INIT_PARAMS		int curcs = *curcsp & ~7
178 #define UPDATE_PARAMS		*curcsp = curcs
179 #include <iconv/loop.c>
180 
181 /* Next, define the other direction */
182 #define MIN_NEEDED_INPUT	TO_LOOP_MIN_NEEDED_FROM
183 #define MAX_NEEDED_INPUT	TO_LOOP_MAX_NEEDED_FROM
184 #define MIN_NEEDED_OUTPUT	TO_LOOP_MIN_NEEDED_TO
185 #define MAX_NEEDED_OUTPUT	TO_LOOP_MAX_NEEDED_TO
186 #define LOOPFCT			TO_LOOP
187 #define BODY \
188   {									      \
189     uint32_t ch = get32 (inptr);					      \
190     const struct gap *rp1 = __ucs4_to_ibm939sb_idx;			      \
191     const struct gap *rp2 = __ucs4_to_ibm939db_idx;			      \
192 									      \
193     if (__glibc_unlikely (ch >= 0xffff))				      \
194       {									      \
195 	UNICODE_TAG_HANDLER (ch, 4);					      \
196 	goto ibm939_invalid_char;					      \
197       }									      \
198 									      \
199     while (ch > rp1->end)						      \
200       ++rp1;								      \
201 									      \
202     /* Use the UCS4 table for single byte.  */				      \
203     unsigned char sbconv;						      \
204     if (__builtin_expect (ch < rp1->start, 0)				      \
205 	|| (sbconv = __ucs4_to_ibm939sb[ch + rp1->idx],			      \
206 	    __builtin_expect (sbconv, L'\1') == L'\0' && ch != '\0'))	      \
207       {									      \
208 	/* Use the UCS4 table for double byte.  */			      \
209 	while (ch > rp2->end)						      \
210 	  ++rp2;							      \
211 									      \
212 	const char *cp;							      \
213 	if (__builtin_expect (ch < rp2->start, 0)			      \
214 	    || (cp = __ucs4_to_ibm939db[ch + rp2->idx],			      \
215 		__builtin_expect (cp[0], L'\1')==L'\0' && ch != '\0'))	      \
216 	  {								      \
217 	    /* This is an illegal character.  */			      \
218 	  ibm939_invalid_char:						      \
219 	    STANDARD_TO_LOOP_ERR_HANDLER (4);				      \
220 	  }								      \
221 	else								      \
222 	  {								      \
223 	    if (curcs == sb)						      \
224 	      {								      \
225 		if (__glibc_unlikely (outptr + 1 > outend))		      \
226 		  {							      \
227 		    result = __GCONV_FULL_OUTPUT;			      \
228 		    break;						      \
229 		  }							      \
230 		*outptr++ = SO;						      \
231 		curcs = db;						      \
232 	      }								      \
233 									      \
234 	    if (__glibc_unlikely (outptr + 2 > outend))			      \
235 	      {								      \
236 		result = __GCONV_FULL_OUTPUT;				      \
237 		break;							      \
238 	      }								      \
239 	    *outptr++ = cp[0];						      \
240 	    *outptr++ = cp[1];						      \
241 	  }								      \
242       }									      \
243     else								      \
244       {									      \
245 	if (curcs == db)						      \
246 	  {								      \
247 	    if (__glibc_unlikely (outptr + 1 > outend))			      \
248 	      {								      \
249 		result = __GCONV_FULL_OUTPUT;				      \
250 		break;							      \
251 	      }								      \
252 	    *outptr++ = SI;						      \
253 	    curcs = sb;							      \
254 	  }								      \
255 									      \
256 	if (__glibc_unlikely (outptr + 1 > outend))			      \
257 	  {								      \
258 	    result = __GCONV_FULL_OUTPUT;				      \
259 	    break;							      \
260 	  }								      \
261 	if (ch == 0x7e)							      \
262 	  *outptr++ = 0xa0;						      \
263 	else if (ch == 0x5c)						      \
264 	  *outptr++ = 0xb2;						      \
265 	else								      \
266 	  *outptr++ = sbconv;						      \
267       }									      \
268 									      \
269     /* Now that we wrote the output increment the input pointer.  */	      \
270     inptr += 4;								      \
271   }
272 #define LOOP_NEED_FLAGS
273 #define EXTRA_LOOP_DECLS	, int *curcsp
274 #define INIT_PARAMS		int curcs = *curcsp & ~7
275 #define REINIT_PARAMS		curcs = *curcsp & ~7
276 #define UPDATE_PARAMS		*curcsp = curcs
277 #include <iconv/loop.c>
278 
279 /* Now define the toplevel functions. */
280 #include <iconv/skeleton.c>
281