Update.
[glibc.git] / iconv / loop.c
blobae83894dc9dea3a90220e16b1d12a79e281cc72b
1 /* Conversion loop frame work.
2 Copyright (C) 1998, 1999, 2000, 2001 Free Software Foundation, Inc.
3 This file is part of the GNU C Library.
4 Contributed by Ulrich Drepper <drepper@cygnus.com>, 1998.
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.
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.
16 You should have received a copy of the GNU Lesser General Public
17 License along with the GNU C Library; if not, write to the Free
18 Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
19 02111-1307 USA. */
21 /* This file provides a frame for the reader loop in all conversion modules.
22 The actual code must (of course) be provided in the actual module source
23 code but certain actions can be written down generically, with some
24 customization options which are these:
26 MIN_NEEDED_INPUT minimal number of input bytes needed for the next
27 conversion.
28 MIN_NEEDED_OUTPUT minimal number of bytes produced by the next round
29 of conversion.
31 MAX_NEEDED_INPUT you guess it, this is the maximal number of input
32 bytes needed. It defaults to MIN_NEEDED_INPUT
33 MAX_NEEDED_OUTPUT likewise for output bytes.
35 LOOPFCT name of the function created. If not specified
36 the name is `loop' but this prevents the use
37 of multiple functions in the same file.
39 BODY this is supposed to expand to the body of the loop.
40 The user must provide this.
42 EXTRA_LOOP_DECLS extra arguments passed from converion loop call.
44 INIT_PARAMS code to define and initialize variables from params.
45 UPDATE_PARAMS code to store result in params.
48 #include <assert.h>
49 #include <endian.h>
50 #include <gconv.h>
51 #include <stdint.h>
52 #include <string.h>
53 #include <wchar.h>
54 #include <sys/param.h> /* For MIN. */
55 #define __need_size_t
56 #include <stddef.h>
59 /* We have to provide support for machines which are not able to handled
60 unaligned memory accesses. Some of the character encodings have
61 representations with a fixed width of 2 or 4 bytes. But if we cannot
62 access unaligned memory we still have to read byte-wise. */
63 #undef FCTNAME2
64 #if defined _STRING_ARCH_unaligned || !defined DEFINE_UNALIGNED
65 /* We can handle unaligned memory access. */
66 # define get16(addr) *((__const uint16_t *) (addr))
67 # define get32(addr) *((__const uint32_t *) (addr))
69 /* We need no special support for writing values either. */
70 # define put16(addr, val) *((uint16_t *) (addr)) = (val)
71 # define put32(addr, val) *((uint32_t *) (addr)) = (val)
73 # define FCTNAME2(name) name
74 #else
75 /* Distinguish between big endian and little endian. */
76 # if __BYTE_ORDER == __LITTLE_ENDIAN
77 # define get16(addr) \
78 (((__const unsigned char *) (addr))[1] << 8 \
79 | ((__const unsigned char *) (addr))[0])
80 # define get32(addr) \
81 (((((__const unsigned char *) (addr))[3] << 8 \
82 | ((__const unsigned char *) (addr))[2]) << 8 \
83 | ((__const unsigned char *) (addr))[1]) << 8 \
84 | ((__const unsigned char *) (addr))[0])
86 # define put16(addr, val) \
87 ({ uint16_t __val = (val); \
88 ((unsigned char *) (addr))[0] = __val; \
89 ((unsigned char *) (addr))[1] = __val >> 8; \
90 (void) 0; })
91 # define put32(addr, val) \
92 ({ uint32_t __val = (val); \
93 ((unsigned char *) (addr))[0] = __val; \
94 __val >>= 8; \
95 ((unsigned char *) (addr))[1] = __val; \
96 __val >>= 8; \
97 ((unsigned char *) (addr))[2] = __val; \
98 __val >>= 8; \
99 ((unsigned char *) (addr))[3] = __val; \
100 (void) 0; })
101 # else
102 # define get16(addr) \
103 (((__const unsigned char *) (addr))[0] << 8 \
104 | ((__const unsigned char *) (addr))[1])
105 # define get32(addr) \
106 (((((__const unsigned char *) (addr))[0] << 8 \
107 | ((__const unsigned char *) (addr))[1]) << 8 \
108 | ((__const unsigned char *) (addr))[2]) << 8 \
109 | ((__const unsigned char *) (addr))[3])
111 # define put16(addr, val) \
112 ({ uint16_t __val = (val); \
113 ((unsigned char *) (addr))[1] = __val; \
114 ((unsigned char *) (addr))[0] = __val >> 8; \
115 (void) 0; })
116 # define put32(addr, val) \
117 ({ uint32_t __val = (val); \
118 ((unsigned char *) (addr))[3] = __val; \
119 __val >>= 8; \
120 ((unsigned char *) (addr))[2] = __val; \
121 __val >>= 8; \
122 ((unsigned char *) (addr))[1] = __val; \
123 __val >>= 8; \
124 ((unsigned char *) (addr))[0] = __val; \
125 (void) 0; })
126 # endif
128 # define FCTNAME2(name) name##_unaligned
129 #endif
130 #define FCTNAME(name) FCTNAME2(name)
133 /* We need at least one byte for the next round. */
134 #ifndef MIN_NEEDED_INPUT
135 # error "MIN_NEEDED_INPUT definition missing"
136 #endif
138 /* Let's see how many bytes we produce. */
139 #ifndef MAX_NEEDED_INPUT
140 # define MAX_NEEDED_INPUT MIN_NEEDED_INPUT
141 #endif
143 /* We produce at least one byte in the next round. */
144 #ifndef MIN_NEEDED_OUTPUT
145 # error "MIN_NEEDED_OUTPUT definition missing"
146 #endif
148 /* Let's see how many bytes we produce. */
149 #ifndef MAX_NEEDED_OUTPUT
150 # define MAX_NEEDED_OUTPUT MIN_NEEDED_OUTPUT
151 #endif
153 /* Default name for the function. */
154 #ifndef LOOPFCT
155 # define LOOPFCT loop
156 #endif
158 /* Make sure we have a loop body. */
159 #ifndef BODY
160 # error "Definition of BODY missing for function" LOOPFCT
161 #endif
164 /* If no arguments have to passed to the loop function define the macro
165 as empty. */
166 #ifndef EXTRA_LOOP_DECLS
167 # define EXTRA_LOOP_DECLS
168 #endif
171 /* To make it easier for the writers of the modules, we define a macro
172 to test whether we have to ignore errors. */
173 #define ignore_errors_p() \
174 (irreversible != NULL && (flags & __GCONV_IGNORE_ERRORS))
177 /* Error handling with transliteration/transcription function use and
178 ignoring of errors. Note that we cannot use the do while (0) trick
179 since `break' and `continue' must reach certain points. */
180 #define STANDARD_ERR_HANDLER(Incr) \
182 struct __gconv_trans_data *trans; \
184 result = __GCONV_ILLEGAL_INPUT; \
186 if (irreversible == NULL) \
187 /* This means we are in call from __gconv_transliterate. In this \
188 case we are not doing any error recovery outself. */ \
189 break; \
191 /* First try the transliteration methods. */ \
192 for (trans = step_data->__trans; trans != NULL; trans = trans->__next) \
194 result = DL_CALL_FCT (trans->__trans_fct, \
195 (step, step_data, trans->__data, *inptrp, \
196 &inptr, inend, &outptr, irreversible)); \
197 if (result != __GCONV_ILLEGAL_INPUT) \
198 break; \
200 /* If any of them recognized the input continue with the loop. */ \
201 if (result != __GCONV_ILLEGAL_INPUT) \
202 continue; \
204 /* Next see whether we have to ignore the error. If not, stop. */ \
205 if (! ignore_errors_p ()) \
206 break; \
208 /* When we come here it means we ignore the character. */ \
209 ++*irreversible; \
210 inptr += Incr; \
211 continue; \
215 /* Handling of Unicode 3.1 TAG characters. Unicode recommends
216 "If language codes are not relevant to the particular processing
217 operation, then they should be ignored."
218 This macro is usually called right before STANDARD_ERR_HANDLER (Incr). */
219 #define UNICODE_TAG_HANDLER(Character, Incr) \
221 /* TAG characters are those in the range U+E0000..U+E007F. */ \
222 if (((Character) >> 7) == (0xe0000 >> 7)) \
224 inptr += Incr; \
225 continue; \
230 /* The function returns the status, as defined in gconv.h. */
231 static inline int
232 FCTNAME (LOOPFCT) (struct __gconv_step *step,
233 struct __gconv_step_data *step_data,
234 const unsigned char **inptrp, const unsigned char *inend,
235 unsigned char **outptrp, const unsigned char *outend,
236 size_t *irreversible EXTRA_LOOP_DECLS)
238 #ifdef LOOP_NEED_STATE
239 mbstate_t *state = step_data->__statep;
240 #endif
241 #ifdef LOOP_NEED_FLAGS
242 int flags = step_data->__flags;
243 #endif
244 #ifdef LOOP_NEED_DATA
245 void *data = step->__data;
246 #endif
247 int result = __GCONV_EMPTY_INPUT;
248 const unsigned char *inptr = *inptrp;
249 unsigned char *outptr = *outptrp;
251 #ifdef INIT_PARAMS
252 INIT_PARAMS;
253 #endif
255 while (inptr != inend)
257 /* `if' cases for MIN_NEEDED_OUTPUT ==/!= 1 is made to help the
258 compiler generating better code. They will be optimized away
259 since MIN_NEEDED_OUTPUT is always a constant. */
260 if ((MIN_NEEDED_OUTPUT != 1
261 && __builtin_expect (outptr + MIN_NEEDED_OUTPUT > outend, 0))
262 || (MIN_NEEDED_OUTPUT == 1
263 && __builtin_expect (outptr >= outend, 0)))
265 /* Overflow in the output buffer. */
266 result = __GCONV_FULL_OUTPUT;
267 break;
269 if (MIN_NEEDED_INPUT > 1
270 && __builtin_expect (inptr + MIN_NEEDED_INPUT > inend, 0))
272 /* We don't have enough input for another complete input
273 character. */
274 result = __GCONV_INCOMPLETE_INPUT;
275 break;
278 /* Here comes the body the user provides. It can stop with
279 RESULT set to GCONV_INCOMPLETE_INPUT (if the size of the
280 input characters vary in size), GCONV_ILLEGAL_INPUT, or
281 GCONV_FULL_OUTPUT (if the output characters vary in size). */
282 BODY
285 /* Update the pointers pointed to by the parameters. */
286 *inptrp = inptr;
287 *outptrp = outptr;
288 #ifdef UPDATE_PARAMS
289 UPDATE_PARAMS;
290 #endif
292 return result;
296 /* Include the file a second time to define the function to handle
297 unaligned access. */
298 #if !defined DEFINE_UNALIGNED && !defined _STRING_ARCH_unaligned \
299 && MIN_NEEDED_FROM != 1 && MAX_NEEDED_FROM % MIN_NEEDED_FROM == 0 \
300 && MIN_NEEDED_TO != 1 && MAX_NEEDED_TO % MIN_NEEDED_TO == 0
301 # undef get16
302 # undef get32
303 # undef put16
304 # undef put32
305 # undef unaligned
307 # define DEFINE_UNALIGNED
308 # include "loop.c"
309 # undef DEFINE_UNALIGNED
310 #endif
313 #if MAX_NEEDED_INPUT > 1
314 # define SINGLE(fct) SINGLE2 (fct)
315 # define SINGLE2(fct) fct##_single
316 static inline int
317 SINGLE(LOOPFCT) (struct __gconv_step *step,
318 struct __gconv_step_data *step_data,
319 const unsigned char **inptrp, const unsigned char *inend,
320 unsigned char **outptrp, unsigned char *outend,
321 size_t *irreversible EXTRA_LOOP_DECLS)
323 mbstate_t *state = step_data->__statep;
324 #ifdef LOOP_NEED_FLAGS
325 int flags = step_data->__flags;
326 #endif
327 #ifdef LOOP_NEED_DATA
328 void *data = step->__data;
329 #endif
330 int result = __GCONV_OK;
331 unsigned char bytebuf[MAX_NEEDED_INPUT];
332 const unsigned char *inptr = *inptrp;
333 unsigned char *outptr = *outptrp;
334 size_t inlen;
336 #ifdef INIT_PARAMS
337 INIT_PARAMS;
338 #endif
340 #ifdef UNPACK_BYTES
341 UNPACK_BYTES
342 #else
343 /* Add the bytes from the state to the input buffer. */
344 for (inlen = 0; inlen < (size_t) (state->__count & 7); ++inlen)
345 bytebuf[inlen] = state->__value.__wchb[inlen];
346 #endif
348 /* Are there enough bytes in the input buffer? */
349 if (__builtin_expect (inptr + (MIN_NEEDED_INPUT - inlen) > inend, 0))
351 *inptrp = inend;
352 #ifdef STORE_REST
353 inptr = bytebuf;
354 inptrp = &inptr;
355 inend = &bytebuf[inlen];
357 STORE_REST
358 #else
359 /* We don't have enough input for another complete input
360 character. */
361 while (inptr < inend)
362 state->__value.__wchb[inlen++] = *inptr++;
363 #endif
365 return __GCONV_INCOMPLETE_INPUT;
368 /* Enough space in output buffer. */
369 if ((MIN_NEEDED_OUTPUT != 1 && outptr + MIN_NEEDED_OUTPUT > outend)
370 || (MIN_NEEDED_OUTPUT == 1 && outptr >= outend))
371 /* Overflow in the output buffer. */
372 return __GCONV_FULL_OUTPUT;
374 /* Now add characters from the normal input buffer. */
376 bytebuf[inlen++] = *inptr++;
377 while (inlen < MAX_NEEDED_INPUT && inptr < inend);
379 inptr = bytebuf;
380 inend = &bytebuf[inlen];
384 BODY
386 while (0);
388 /* Now we either have produced an output character and consumed all the
389 bytes from the state and at least one more, or the character is still
390 incomplete, or we have some other error (like illegal input character,
391 no space in output buffer). */
392 if (__builtin_expect (inptr != bytebuf, 1))
394 /* We found a new character. */
395 assert (inptr - bytebuf > (state->__count & 7));
397 *inptrp += inptr - bytebuf - (state->__count & 7);
398 *outptrp = outptr;
400 result = __GCONV_OK;
402 /* Clear the state buffer. */
403 state->__count &= ~7;
405 else if (result == __GCONV_INCOMPLETE_INPUT)
407 /* This can only happen if we have less than MAX_NEEDED_INPUT bytes
408 available. */
409 assert (inend != &bytebuf[MAX_NEEDED_INPUT]);
411 *inptrp += inend - bytebuf - (state->__count & 7);
412 #ifdef STORE_REST
413 inptrp = &inptr;
415 STORE_REST
416 #else
417 /* We don't have enough input for another complete input
418 character. */
419 while (inptr < inend)
420 state->__value.__wchb[inlen++] = *inptr++;
421 #endif
424 return result;
426 # undef SINGLE
427 # undef SINGLE2
428 #endif
431 /* We remove the macro definitions so that we can include this file again
432 for the definition of another function. */
433 #undef MIN_NEEDED_INPUT
434 #undef MAX_NEEDED_INPUT
435 #undef MIN_NEEDED_OUTPUT
436 #undef MAX_NEEDED_OUTPUT
437 #undef LOOPFCT
438 #undef BODY
439 #undef LOOPFCT
440 #undef EXTRA_LOOP_DECLS
441 #undef INIT_PARAMS
442 #undef UPDATE_PARAMS
443 #undef UNPACK_BYTES
444 #undef LOOP_NEED_STATE
445 #undef LOOP_NEED_FLAGS
446 #undef LOOP_NEED_DATA
447 #undef get16
448 #undef get32
449 #undef put16
450 #undef put32
451 #undef unaligned