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 struct __gconv_trans_data *trans; \ 209 \ 210 result = __GCONV_ILLEGAL_INPUT; \ 211 \ 212 if (irreversible == NULL) \ 213 /* This means we are in call from __gconv_transliterate. In this \ 214 case we are not doing any error recovery outself. */ \ 215 break; \ 216 \ 217 /* First try the transliteration methods. */ \ 218 for (trans = step_data->__trans; trans != NULL; trans = trans->__next) \ 219 { \ 220 result = DL_CALL_FCT (trans->__trans_fct, \ 221 (step, step_data, trans->__data, *inptrp, \ 222 &inptr, inend, &outptr, irreversible)); \ 223 if (result != __GCONV_ILLEGAL_INPUT) \ 224 break; \ 225 } \ 226 /* If any of them recognized the input continue with the loop. */ \ 227 if (result != __GCONV_ILLEGAL_INPUT) \ 228 continue; \ 229 \ 230 /* Next see whether we have to ignore the error. If not, stop. */ \ 231 if (! ignore_errors_p ()) \ 232 break; \ 233 \ 234 /* When we come here it means we ignore the character. */ \ 235 ++*irreversible; \ 236 inptr += Incr; \ 237 /* But we keep result == __GCONV_ILLEGAL_INPUT, because of the constraint \ 238 that "iconv -c" must give the same exitcode as "iconv". */ \ 239 continue; \ 240 } 241 242 243 /* Handling of Unicode 3.1 TAG characters. Unicode recommends 244 "If language codes are not relevant to the particular processing 245 operation, then they should be ignored." This macro is usually 246 called right before STANDARD_TO_LOOP_ERR_HANDLER (Incr). */ 247 #define UNICODE_TAG_HANDLER(Character, Incr) \ 248 { \ 249 /* TAG characters are those in the range U+E0000..U+E007F. */ \ 250 if (((Character) >> 7) == (0xe0000 >> 7)) \ 251 { \ 252 inptr += Incr; \ 253 continue; \ 254 } \ 255 } 256 257 258 /* The function returns the status, as defined in gconv.h. */ 259 static inline int 260 FCTNAME (LOOPFCT) (struct __gconv_step *step, 261 struct __gconv_step_data *step_data, 262 const unsigned char **inptrp, const unsigned char *inend, 263 unsigned char **outptrp, const unsigned char *outend, 264 size_t *irreversible EXTRA_LOOP_DECLS) 265 { 266 #ifdef LOOP_NEED_STATE 267 mbstate_t *state = step_data->__statep; 268 #endif 269 #ifdef LOOP_NEED_FLAGS 270 int flags = step_data->__flags; 271 #endif 272 #ifdef LOOP_NEED_DATA 273 void *data = step->__data; 274 #endif 275 int result = __GCONV_EMPTY_INPUT; 276 const unsigned char *inptr = *inptrp; 277 unsigned char *outptr = *outptrp; 278 279 #ifdef INIT_PARAMS 280 INIT_PARAMS; 281 #endif 282 283 while (inptr != inend) 284 { 285 /* `if' cases for MIN_NEEDED_OUTPUT ==/!= 1 is made to help the 286 compiler generating better code. They will be optimized away 287 since MIN_NEEDED_OUTPUT is always a constant. */ 288 if (MIN_NEEDED_INPUT > 1 289 && __builtin_expect (inptr + MIN_NEEDED_INPUT > inend, 0)) 290 { 291 /* We don't have enough input for another complete input 292 character. */ 293 result = __GCONV_INCOMPLETE_INPUT; 294 break; 295 } 296 if ((MIN_NEEDED_OUTPUT != 1 297 && __builtin_expect (outptr + MIN_NEEDED_OUTPUT > outend, 0)) 298 || (MIN_NEEDED_OUTPUT == 1 299 && __builtin_expect (outptr >= outend, 0))) 300 { 301 /* Overflow in the output buffer. */ 302 result = __GCONV_FULL_OUTPUT; 303 break; 304 } 305 306 /* Here comes the body the user provides. It can stop with 307 RESULT set to GCONV_INCOMPLETE_INPUT (if the size of the 308 input characters vary in size), GCONV_ILLEGAL_INPUT, or 309 GCONV_FULL_OUTPUT (if the output characters vary in size). */ 310 BODY 311 } 312 313 /* Update the pointers pointed to by the parameters. */ 314 *inptrp = inptr; 315 *outptrp = outptr; 316 #ifdef UPDATE_PARAMS 317 UPDATE_PARAMS; 318 #endif 319 320 return result; 321 } 322 323 324 /* Include the file a second time to define the function to handle 325 unaligned access. */ 326 #if !defined DEFINE_UNALIGNED && !defined _STRING_ARCH_unaligned \ 327 && MIN_NEEDED_INPUT != 1 && MAX_NEEDED_INPUT % MIN_NEEDED_INPUT == 0 \ 328 && MIN_NEEDED_OUTPUT != 1 && MAX_NEEDED_OUTPUT % MIN_NEEDED_OUTPUT == 0 329 # undef get16 330 # undef get32 331 # undef put16 332 # undef put32 333 # undef unaligned 334 335 # define DEFINE_UNALIGNED 336 # include "loop.c" 337 # undef DEFINE_UNALIGNED 338 #endif 339 340 341 #if MAX_NEEDED_INPUT > 1 342 # define SINGLE(fct) SINGLE2 (fct) 343 # define SINGLE2(fct) fct##_single 344 static inline int 345 SINGLE(LOOPFCT) (struct __gconv_step *step, 346 struct __gconv_step_data *step_data, 347 const unsigned char **inptrp, const unsigned char *inend, 348 unsigned char **outptrp, unsigned char *outend, 349 size_t *irreversible EXTRA_LOOP_DECLS) 350 { 351 mbstate_t *state = step_data->__statep; 352 #ifdef LOOP_NEED_FLAGS 353 int flags = step_data->__flags; 354 #endif 355 #ifdef LOOP_NEED_DATA 356 void *data = step->__data; 357 #endif 358 int result = __GCONV_OK; 359 unsigned char bytebuf[MAX_NEEDED_INPUT]; 360 const unsigned char *inptr = *inptrp; 361 unsigned char *outptr = *outptrp; 362 size_t inlen; 363 364 #ifdef INIT_PARAMS 365 INIT_PARAMS; 366 #endif 367 368 #ifdef UNPACK_BYTES 369 UNPACK_BYTES 370 #else 371 /* Add the bytes from the state to the input buffer. */ 372 for (inlen = 0; inlen < (size_t) (state->__count & 7); ++inlen) 373 bytebuf[inlen] = state->__value.__wchb[inlen]; 374 #endif 375 376 /* Are there enough bytes in the input buffer? */ 377 if (__builtin_expect (inptr + (MIN_NEEDED_INPUT - inlen) > inend, 0)) 378 { 379 *inptrp = inend; 380 #ifdef STORE_REST 381 inptr = bytebuf; 382 inptrp = &inptr; 383 inend = &bytebuf[inlen]; 384 385 STORE_REST 386 #else 387 /* We don't have enough input for another complete input 388 character. */ 389 while (inptr < inend) 390 state->__value.__wchb[inlen++] = *inptr++; 391 #endif 392 393 return __GCONV_INCOMPLETE_INPUT; 394 } 395 396 /* Enough space in output buffer. */ 397 if ((MIN_NEEDED_OUTPUT != 1 && outptr + MIN_NEEDED_OUTPUT > outend) 398 || (MIN_NEEDED_OUTPUT == 1 && outptr >= outend)) 399 /* Overflow in the output buffer. */ 400 return __GCONV_FULL_OUTPUT; 401 402 /* Now add characters from the normal input buffer. */ 403 do 404 bytebuf[inlen++] = *inptr++; 405 while (inlen < MAX_NEEDED_INPUT && inptr < inend); 406 407 inptr = bytebuf; 408 inend = &bytebuf[inlen]; 409 410 do 411 { 412 BODY 413 } 414 while (0); 415 416 /* Now we either have produced an output character and consumed all the 417 bytes from the state and at least one more, or the character is still 418 incomplete, or we have some other error (like illegal input character, 419 no space in output buffer). */ 420 if (__builtin_expect (inptr != bytebuf, 1)) 421 { 422 /* We found a new character. */ 423 assert (inptr - bytebuf > (state->__count & 7)); 424 425 *inptrp += inptr - bytebuf - (state->__count & 7); 426 *outptrp = outptr; 427 428 result = __GCONV_OK; 429 430 /* Clear the state buffer. */ 431 #ifdef CLEAR_STATE 432 CLEAR_STATE; 433 #else 434 state->__count &= ~7; 435 #endif 436 } 437 else if (result == __GCONV_INCOMPLETE_INPUT) 438 { 439 /* This can only happen if we have less than MAX_NEEDED_INPUT bytes 440 available. */ 441 assert (inend != &bytebuf[MAX_NEEDED_INPUT]); 442 443 *inptrp += inend - bytebuf - (state->__count & 7); 444 #ifdef STORE_REST 445 inptrp = &inptr; 446 447 STORE_REST 448 #else 449 /* We don't have enough input for another complete input 450 character. */ 451 while (inptr < inend) 452 state->__value.__wchb[inlen++] = *inptr++; 453 #endif 454 } 455 456 return result; 457 } 458 # undef SINGLE 459 # undef SINGLE2 460 #endif 461 462 463 #ifdef ONEBYTE_BODY 464 /* Define the shortcut function for btowc. */ 465 static wint_t 466 gconv_btowc (struct __gconv_step *step, unsigned char c) 467 ONEBYTE_BODY 468 # define FROM_ONEBYTE gconv_btowc 469 #endif 470 471 472 /* We remove the macro definitions so that we can include this file again 473 for the definition of another function. */ 474 #undef MIN_NEEDED_INPUT 475 #undef MAX_NEEDED_INPUT 476 #undef MIN_NEEDED_OUTPUT 477 #undef MAX_NEEDED_OUTPUT 478 #undef LOOPFCT 479 #undef BODY 480 #undef LOOPFCT 481 #undef EXTRA_LOOP_DECLS 482 #undef INIT_PARAMS 483 #undef UPDATE_PARAMS 484 #undef ONEBYTE_BODY 485 #undef UNPACK_BYTES 486 #undef CLEAR_STATE 487 #undef LOOP_NEED_STATE 488 #undef LOOP_NEED_FLAGS 489 #undef LOOP_NEED_DATA 490 #undef get16 491 #undef get32 492 #undef put16 493 #undef put32 494 #undef unaligned 495