1 /* Conversion loop frame work. 2 Copyright (C) 1998-2002, 2003 Free Software Foundation, Inc. 3 This file is part of the GNU C Library. 4 Contributed by Ulrich Drepper <drepper@cygnus.com>, 1998. 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, write to the Free 18 Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 19 02111-1307 USA. */ 20 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: 25 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. 30 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. 34 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. 38 39 BODY this is supposed to expand to the body of the loop. 40 The user must provide this. 41 42 EXTRA_LOOP_DECLS extra arguments passed from converion loop call. 43 44 INIT_PARAMS code to define and initialize variables from params. 45 UPDATE_PARAMS code to store result in params. 46 47 ONEBYTE_BODY body of the specialized conversion function for a 48 single byte from the current character set to INTERNAL. 49 */ 50 51 #include <assert.h> 52 #include <endian.h> 53 #include <gconv.h> 54 #include <stdint.h> 55 #include <string.h> 56 #include <wchar.h> 57 #include <sys/param.h> /* For MIN. */ 58 #define __need_size_t 59 #include <stddef.h> 60 61 62 /* We have to provide support for machines which are not able to handled 63 unaligned memory accesses. Some of the character encodings have 64 representations with a fixed width of 2 or 4 bytes. But if we cannot 65 access unaligned memory we still have to read byte-wise. */ 66 #undef FCTNAME2 67 #if defined _STRING_ARCH_unaligned || !defined DEFINE_UNALIGNED 68 /* We can handle unaligned memory access. */ 69 # define get16(addr) *((__const uint16_t *) (addr)) 70 # define get32(addr) *((__const uint32_t *) (addr)) 71 72 /* We need no special support for writing values either. */ 73 # define put16(addr, val) *((uint16_t *) (addr)) = (val) 74 # define put32(addr, val) *((uint32_t *) (addr)) = (val) 75 76 # define FCTNAME2(name) name 77 #else 78 /* Distinguish between big endian and little endian. */ 79 # if __BYTE_ORDER == __LITTLE_ENDIAN 80 # define get16(addr) \ 81 (((__const unsigned char *) (addr))[1] << 8 \ 82 | ((__const unsigned char *) (addr))[0]) 83 # define get32(addr) \ 84 (((((__const unsigned char *) (addr))[3] << 8 \ 85 | ((__const unsigned char *) (addr))[2]) << 8 \ 86 | ((__const unsigned char *) (addr))[1]) << 8 \ 87 | ((__const unsigned char *) (addr))[0]) 88 89 # define put16(addr, val) \ 90 ({ uint16_t __val = (val); \ 91 ((unsigned char *) (addr))[0] = __val; \ 92 ((unsigned char *) (addr))[1] = __val >> 8; \ 93 (void) 0; }) 94 # define put32(addr, val) \ 95 ({ uint32_t __val = (val); \ 96 ((unsigned char *) (addr))[0] = __val; \ 97 __val >>= 8; \ 98 ((unsigned char *) (addr))[1] = __val; \ 99 __val >>= 8; \ 100 ((unsigned char *) (addr))[2] = __val; \ 101 __val >>= 8; \ 102 ((unsigned char *) (addr))[3] = __val; \ 103 (void) 0; }) 104 # else 105 # define get16(addr) \ 106 (((__const unsigned char *) (addr))[0] << 8 \ 107 | ((__const unsigned char *) (addr))[1]) 108 # define get32(addr) \ 109 (((((__const unsigned char *) (addr))[0] << 8 \ 110 | ((__const unsigned char *) (addr))[1]) << 8 \ 111 | ((__const unsigned char *) (addr))[2]) << 8 \ 112 | ((__const unsigned char *) (addr))[3]) 113 114 # define put16(addr, val) \ 115 ({ uint16_t __val = (val); \ 116 ((unsigned char *) (addr))[1] = __val; \ 117 ((unsigned char *) (addr))[0] = __val >> 8; \ 118 (void) 0; }) 119 # define put32(addr, val) \ 120 ({ uint32_t __val = (val); \ 121 ((unsigned char *) (addr))[3] = __val; \ 122 __val >>= 8; \ 123 ((unsigned char *) (addr))[2] = __val; \ 124 __val >>= 8; \ 125 ((unsigned char *) (addr))[1] = __val; \ 126 __val >>= 8; \ 127 ((unsigned char *) (addr))[0] = __val; \ 128 (void) 0; }) 129 # endif 130 131 # define FCTNAME2(name) name##_unaligned 132 #endif 133 #define FCTNAME(name) FCTNAME2(name) 134 135 136 /* We need at least one byte for the next round. */ 137 #ifndef MIN_NEEDED_INPUT 138 # error "MIN_NEEDED_INPUT definition missing" 139 #elif MIN_NEEDED_INPUT < 1 140 # error "MIN_NEEDED_INPUT must be >= 1" 141 #endif 142 143 /* Let's see how many bytes we produce. */ 144 #ifndef MAX_NEEDED_INPUT 145 # define MAX_NEEDED_INPUT MIN_NEEDED_INPUT 146 #endif 147 148 /* We produce at least one byte in the next round. */ 149 #ifndef MIN_NEEDED_OUTPUT 150 # error "MIN_NEEDED_OUTPUT definition missing" 151 #elif MIN_NEEDED_OUTPUT < 1 152 # error "MIN_NEEDED_OUTPUT must be >= 1" 153 #endif 154 155 /* Let's see how many bytes we produce. */ 156 #ifndef MAX_NEEDED_OUTPUT 157 # define MAX_NEEDED_OUTPUT MIN_NEEDED_OUTPUT 158 #endif 159 160 /* Default name for the function. */ 161 #ifndef LOOPFCT 162 # define LOOPFCT loop 163 #endif 164 165 /* Make sure we have a loop body. */ 166 #ifndef BODY 167 # error "Definition of BODY missing for function" LOOPFCT 168 #endif 169 170 171 /* If no arguments have to passed to the loop function define the macro 172 as empty. */ 173 #ifndef EXTRA_LOOP_DECLS 174 # define EXTRA_LOOP_DECLS 175 #endif 176 177 178 /* To make it easier for the writers of the modules, we define a macro 179 to test whether we have to ignore errors. */ 180 #define ignore_errors_p() \ 181 (irreversible != NULL && (flags & __GCONV_IGNORE_ERRORS)) 182 183 184 /* Error handling for the FROM_LOOP direction, with ignoring of errors. 185 Note that we cannot use the do while (0) trick since `break' and 186 `continue' must reach certain points. */ 187 #define STANDARD_FROM_LOOP_ERR_HANDLER(Incr) \ 188 { \ 189 result = __GCONV_ILLEGAL_INPUT; \ 190 \ 191 if (! ignore_errors_p ()) \ 192 break; \ 193 \ 194 /* We ignore the invalid input byte sequence. */ \ 195 inptr += (Incr); \ 196 ++*irreversible; \ 197 /* But we keep result == __GCONV_ILLEGAL_INPUT, because of the constraint \ 198 that "iconv -c" must give the same exitcode as "iconv". */ \ 199 continue; \ 200 } 201 202 /* Error handling for the TO_LOOP direction, with use of transliteration/ 203 transcription functions and ignoring of errors. Note that we cannot use 204 the do while (0) trick since `break' and `continue' must reach certain 205 points. */ 206 #define STANDARD_TO_LOOP_ERR_HANDLER(Incr) \ 207 { \ 208 result = __GCONV_ILLEGAL_INPUT; \ 209 \ 210 if (irreversible == NULL) \ 211 /* This means we are in call from __gconv_transliterate. In this \ 212 case we are not doing any error recovery outself. */ \ 213 break; \ 214 \ 215 /* See whether we have to ignore the error. If not, stop. */ \ 216 if (! ignore_errors_p ()) \ 217 break; \ 218 \ 219 /* When we come here it means we ignore the character. */ \ 220 ++*irreversible; \ 221 inptr += Incr; \ 222 /* But we keep result == __GCONV_ILLEGAL_INPUT, because of the constraint \ 223 that "iconv -c" must give the same exitcode as "iconv". */ \ 224 continue; \ 225 } 226 227 228 /* Handling of Unicode 3.1 TAG characters. Unicode recommends 229 "If language codes are not relevant to the particular processing 230 operation, then they should be ignored." This macro is usually 231 called right before STANDARD_TO_LOOP_ERR_HANDLER (Incr). */ 232 #define UNICODE_TAG_HANDLER(Character, Incr) \ 233 { \ 234 /* TAG characters are those in the range U+E0000..U+E007F. */ \ 235 if (((Character) >> 7) == (0xe0000 >> 7)) \ 236 { \ 237 inptr += Incr; \ 238 continue; \ 239 } \ 240 } 241 242 243 /* The function returns the status, as defined in gconv.h. */ 244 static inline int 245 FCTNAME (LOOPFCT) (struct __gconv_step *step, 246 struct __gconv_step_data *step_data, 247 const unsigned char **inptrp, const unsigned char *inend, 248 unsigned char **outptrp, const unsigned char *outend, 249 size_t *irreversible EXTRA_LOOP_DECLS) 250 { 251 #ifdef LOOP_NEED_STATE 252 mbstate_t *state = step_data->__statep; 253 #endif 254 #ifdef LOOP_NEED_FLAGS 255 int flags = step_data->__flags; 256 #endif 257 #ifdef LOOP_NEED_DATA 258 void *data = step->__data; 259 #endif 260 int result = __GCONV_EMPTY_INPUT; 261 const unsigned char *inptr = *inptrp; 262 unsigned char *outptr = *outptrp; 263 264 #ifdef INIT_PARAMS 265 INIT_PARAMS; 266 #endif 267 268 while (inptr != inend) 269 { 270 /* `if' cases for MIN_NEEDED_OUTPUT ==/!= 1 is made to help the 271 compiler generating better code. They will be optimized away 272 since MIN_NEEDED_OUTPUT is always a constant. */ 273 if (MIN_NEEDED_INPUT > 1 274 && __builtin_expect (inptr + MIN_NEEDED_INPUT > inend, 0)) 275 { 276 /* We don't have enough input for another complete input 277 character. */ 278 result = __GCONV_INCOMPLETE_INPUT; 279 break; 280 } 281 if ((MIN_NEEDED_OUTPUT != 1 282 && __builtin_expect (outptr + MIN_NEEDED_OUTPUT > outend, 0)) 283 || (MIN_NEEDED_OUTPUT == 1 284 && __builtin_expect (outptr >= outend, 0))) 285 { 286 /* Overflow in the output buffer. */ 287 result = __GCONV_FULL_OUTPUT; 288 break; 289 } 290 291 /* Here comes the body the user provides. It can stop with 292 RESULT set to GCONV_INCOMPLETE_INPUT (if the size of the 293 input characters vary in size), GCONV_ILLEGAL_INPUT, or 294 GCONV_FULL_OUTPUT (if the output characters vary in size). */ 295 BODY 296 } 297 298 /* Update the pointers pointed to by the parameters. */ 299 *inptrp = inptr; 300 *outptrp = outptr; 301 #ifdef UPDATE_PARAMS 302 UPDATE_PARAMS; 303 #endif 304 305 return result; 306 } 307 308 309 /* Include the file a second time to define the function to handle 310 unaligned access. */ 311 #if !defined DEFINE_UNALIGNED && !defined _STRING_ARCH_unaligned \ 312 && MIN_NEEDED_INPUT != 1 && MAX_NEEDED_INPUT % MIN_NEEDED_INPUT == 0 \ 313 && MIN_NEEDED_OUTPUT != 1 && MAX_NEEDED_OUTPUT % MIN_NEEDED_OUTPUT == 0 314 # undef get16 315 # undef get32 316 # undef put16 317 # undef put32 318 # undef unaligned 319 320 # define DEFINE_UNALIGNED 321 # include "loop.c" 322 # undef DEFINE_UNALIGNED 323 #endif 324 325 326 #if MAX_NEEDED_INPUT > 1 327 # define SINGLE(fct) SINGLE2 (fct) 328 # define SINGLE2(fct) fct##_single 329 static inline int 330 SINGLE(LOOPFCT) (struct __gconv_step *step, 331 struct __gconv_step_data *step_data, 332 const unsigned char **inptrp, const unsigned char *inend, 333 unsigned char **outptrp, unsigned char *outend, 334 size_t *irreversible EXTRA_LOOP_DECLS) 335 { 336 mbstate_t *state = step_data->__statep; 337 #ifdef LOOP_NEED_FLAGS 338 int flags = step_data->__flags; 339 #endif 340 #ifdef LOOP_NEED_DATA 341 void *data = step->__data; 342 #endif 343 int result = __GCONV_OK; 344 unsigned char bytebuf[MAX_NEEDED_INPUT]; 345 const unsigned char *inptr = *inptrp; 346 unsigned char *outptr = *outptrp; 347 size_t inlen; 348 349 #ifdef INIT_PARAMS 350 INIT_PARAMS; 351 #endif 352 353 #ifdef UNPACK_BYTES 354 UNPACK_BYTES 355 #else 356 /* Add the bytes from the state to the input buffer. */ 357 for (inlen = 0; inlen < (size_t) (state->__count & 7); ++inlen) 358 bytebuf[inlen] = state->__value.__wchb[inlen]; 359 #endif 360 361 /* Are there enough bytes in the input buffer? */ 362 if (__builtin_expect (inptr + (MIN_NEEDED_INPUT - inlen) > inend, 0)) 363 { 364 *inptrp = inend; 365 #ifdef STORE_REST 366 inptr = bytebuf; 367 inptrp = &inptr; 368 inend = &bytebuf[inlen]; 369 370 STORE_REST 371 #else 372 /* We don't have enough input for another complete input 373 character. */ 374 while (inptr < inend) 375 state->__value.__wchb[inlen++] = *inptr++; 376 #endif 377 378 return __GCONV_INCOMPLETE_INPUT; 379 } 380 381 /* Enough space in output buffer. */ 382 if ((MIN_NEEDED_OUTPUT != 1 && outptr + MIN_NEEDED_OUTPUT > outend) 383 || (MIN_NEEDED_OUTPUT == 1 && outptr >= outend)) 384 /* Overflow in the output buffer. */ 385 return __GCONV_FULL_OUTPUT; 386 387 /* Now add characters from the normal input buffer. */ 388 do 389 bytebuf[inlen++] = *inptr++; 390 while (inlen < MAX_NEEDED_INPUT && inptr < inend); 391 392 inptr = bytebuf; 393 inend = &bytebuf[inlen]; 394 395 do 396 { 397 BODY 398 } 399 while (0); 400 401 /* Now we either have produced an output character and consumed all the 402 bytes from the state and at least one more, or the character is still 403 incomplete, or we have some other error (like illegal input character, 404 no space in output buffer). */ 405 if (__builtin_expect (inptr != bytebuf, 1)) 406 { 407 /* We found a new character. */ 408 assert (inptr - bytebuf > (state->__count & 7)); 409 410 *inptrp += inptr - bytebuf - (state->__count & 7); 411 *outptrp = outptr; 412 413 result = __GCONV_OK; 414 415 /* Clear the state buffer. */ 416 #ifdef CLEAR_STATE 417 CLEAR_STATE; 418 #else 419 state->__count &= ~7; 420 #endif 421 } 422 else if (result == __GCONV_INCOMPLETE_INPUT) 423 { 424 /* This can only happen if we have less than MAX_NEEDED_INPUT bytes 425 available. */ 426 assert (inend != &bytebuf[MAX_NEEDED_INPUT]); 427 428 *inptrp += inend - bytebuf - (state->__count & 7); 429 #ifdef STORE_REST 430 inptrp = &inptr; 431 432 STORE_REST 433 #else 434 /* We don't have enough input for another complete input 435 character. */ 436 while (inptr < inend) 437 state->__value.__wchb[inlen++] = *inptr++; 438 #endif 439 } 440 441 return result; 442 } 443 # undef SINGLE 444 # undef SINGLE2 445 #endif 446 447 448 #ifdef ONEBYTE_BODY 449 /* Define the shortcut function for btowc. */ 450 static wint_t 451 gconv_btowc (struct __gconv_step *step, unsigned char c) 452 ONEBYTE_BODY 453 # define FROM_ONEBYTE gconv_btowc 454 #endif 455 456 457 /* We remove the macro definitions so that we can include this file again 458 for the definition of another function. */ 459 #undef MIN_NEEDED_INPUT 460 #undef MAX_NEEDED_INPUT 461 #undef MIN_NEEDED_OUTPUT 462 #undef MAX_NEEDED_OUTPUT 463 #undef LOOPFCT 464 #undef BODY 465 #undef LOOPFCT 466 #undef EXTRA_LOOP_DECLS 467 #undef INIT_PARAMS 468 #undef UPDATE_PARAMS 469 #undef ONEBYTE_BODY 470 #undef UNPACK_BYTES 471 #undef CLEAR_STATE 472 #undef LOOP_NEED_STATE 473 #undef LOOP_NEED_FLAGS 474 #undef LOOP_NEED_DATA 475 #undef get16 476 #undef get32 477 #undef put16 478 #undef put32 479 #undef unaligned 480