]>
Commit | Line | Data |
---|---|---|
09376451 | 1 | /* Skeleton for a conversion module. |
45541bdf | 2 | Copyright (C) 1998-2002, 2005 Free Software Foundation, Inc. |
8619129f UD |
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 | |
41bdb6e2 AJ |
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. | |
8619129f UD |
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 | |
41bdb6e2 | 14 | Lesser General Public License for more details. |
8619129f | 15 | |
41bdb6e2 AJ |
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. */ | |
8619129f UD |
20 | |
21 | /* This file can be included to provide definitions of several things | |
22 | many modules have in common. It can be customized using the following | |
23 | macros: | |
24 | ||
25 | DEFINE_INIT define the default initializer. This requires the | |
26 | following symbol to be defined. | |
27 | ||
28 | CHARSET_NAME string with official name of the coded character | |
29 | set (in all-caps) | |
30 | ||
31 | DEFINE_FINI define the default destructor function. | |
32 | ||
33 | MIN_NEEDED_FROM minimal number of bytes needed for the from-charset. | |
34 | MIN_NEEDED_TO likewise for the to-charset. | |
35 | ||
36 | MAX_NEEDED_FROM maximal number of bytes needed for the from-charset. | |
37 | This macro is optional, it defaults to MIN_NEEDED_FROM. | |
38 | MAX_NEEDED_TO likewise for the to-charset. | |
39 | ||
faaa6f62 UD |
40 | FROM_LOOP_MIN_NEEDED_FROM |
41 | FROM_LOOP_MAX_NEEDED_FROM | |
42 | minimal/maximal number of bytes needed on input | |
43 | of one round through the FROM_LOOP. Defaults | |
44 | to MIN_NEEDED_FROM and MAX_NEEDED_FROM, respectively. | |
45 | FROM_LOOP_MIN_NEEDED_TO | |
46 | FROM_LOOP_MAX_NEEDED_TO | |
47 | minimal/maximal number of bytes needed on output | |
48 | of one round through the FROM_LOOP. Defaults | |
49 | to MIN_NEEDED_TO and MAX_NEEDED_TO, respectively. | |
50 | TO_LOOP_MIN_NEEDED_FROM | |
51 | TO_LOOP_MAX_NEEDED_FROM | |
52 | minimal/maximal number of bytes needed on input | |
53 | of one round through the TO_LOOP. Defaults | |
54 | to MIN_NEEDED_TO and MAX_NEEDED_TO, respectively. | |
55 | TO_LOOP_MIN_NEEDED_TO | |
56 | TO_LOOP_MAX_NEEDED_TO | |
57 | minimal/maximal number of bytes needed on output | |
58 | of one round through the TO_LOOP. Defaults | |
59 | to MIN_NEEDED_FROM and MAX_NEEDED_FROM, respectively. | |
60 | ||
8619129f UD |
61 | FROM_DIRECTION this macro is supposed to return a value != 0 |
62 | if we convert from the current character set, | |
63 | otherwise it return 0. | |
64 | ||
65 | EMIT_SHIFT_TO_INIT this symbol is optional. If it is defined it | |
66 | defines some code which writes out a sequence | |
faaa6f62 | 67 | of bytes which bring the current state into |
8619129f UD |
68 | the initial state. |
69 | ||
70 | FROM_LOOP name of the function implementing the conversion | |
faaa6f62 | 71 | from the current character set. |
8619129f UD |
72 | TO_LOOP likewise for the other direction |
73 | ||
f6ad4726 UD |
74 | ONE_DIRECTION optional. If defined to 1, only one conversion |
75 | direction is defined instead of two. In this | |
76 | case, FROM_DIRECTION should be defined to 1, and | |
77 | FROM_LOOP and TO_LOOP should have the same value. | |
78 | ||
79 | SAVE_RESET_STATE in case of an error we must reset the state for | |
8619129f UD |
80 | the rerun so this macro must be defined for |
81 | stateful encodings. It takes an argument which | |
82 | is nonzero when saving. | |
83 | ||
84 | RESET_INPUT_BUFFER If the input character sets allow this the macro | |
85 | can be defined to reset the input buffer pointers | |
86 | to cover only those characters up to the error. | |
87 | ||
88 | FUNCTION_NAME if not set the conversion function is named `gconv'. | |
28f1c862 UD |
89 | |
90 | PREPARE_LOOP optional code preparing the conversion loop. Can | |
91 | contain variable definitions. | |
918b9d72 | 92 | END_LOOP also optional, may be used to store information |
28f1c862 UD |
93 | |
94 | EXTRA_LOOP_ARGS optional macro specifying extra arguments passed | |
95 | to loop function. | |
faaa6f62 | 96 | |
f9ad060c UD |
97 | STORE_REST optional, needed only when MAX_NEEDED_FROM > 4. |
98 | This macro stores the seen but unconverted input bytes | |
99 | in the state. | |
100 | ||
101 | FROM_ONEBYTE optional. If defined, should be the name of a | |
102 | specialized conversion function for a single byte | |
103 | from the current character set to INTERNAL. This | |
104 | function has prototype | |
105 | wint_t | |
106 | FROM_ONEBYTE (struct __gconv_step *, unsigned char); | |
107 | and does a special conversion: | |
108 | - The input is a single byte. | |
109 | - The output is a single uint32_t. | |
110 | - The state before the conversion is the initial state; | |
111 | the state after the conversion is irrelevant. | |
112 | - No transliteration. | |
113 | - __invocation_counter = 0. | |
114 | - __internal_use = 1. | |
115 | - do_flush = 0. | |
116 | ||
faaa6f62 UD |
117 | Modules can use mbstate_t to store conversion state as follows: |
118 | ||
119 | * Bits 2..0 of '__count' contain the number of lookahead input bytes | |
120 | stored in __value.__wchb. Always zero if the converter never | |
121 | returns __GCONV_INCOMPLETE_INPUT. | |
122 | ||
123 | * Bits 31..3 of '__count' are module dependent shift state. | |
124 | ||
125 | * __value: When STORE_REST/UNPACK_BYTES aren't defined and when the | |
126 | converter has returned __GCONV_INCOMPLETE_INPUT, this contains | |
127 | at most 4 lookahead bytes. Converters with an mb_cur_max > 4 | |
128 | (currently only UTF-8) must find a way to store their state | |
129 | in __value.__wch and define STORE_REST/UNPACK_BYTES appropriately. | |
130 | ||
131 | When __value contains lookahead, __count must not be zero, because | |
132 | the converter is not in the initial state then, and mbsinit() -- | |
133 | defined as a (__count == 0) test -- must reflect this. | |
8619129f UD |
134 | */ |
135 | ||
136 | #include <assert.h> | |
137 | #include <gconv.h> | |
138 | #include <string.h> | |
139 | #define __need_size_t | |
140 | #define __need_NULL | |
141 | #include <stddef.h> | |
c66dbe00 UD |
142 | |
143 | #ifndef STATIC_GCONV | |
144 | # include <dlfcn.h> | |
c66dbe00 | 145 | #endif |
8619129f | 146 | |
a808d541 UD |
147 | #ifndef DL_CALL_FCT |
148 | # define DL_CALL_FCT(fct, args) fct args | |
149 | #endif | |
150 | ||
8619129f | 151 | /* The direction objects. */ |
45541bdf | 152 | #if DEFINE_INIT |
8619129f | 153 | # ifndef FROM_DIRECTION |
45541bdf UD |
154 | # define FROM_DIRECTION_VAL NULL |
155 | # define TO_DIRECTION_VAL ((void *) ~((uintptr_t) 0)) | |
156 | # define FROM_DIRECTION (step->__data == FROM_DIRECTION_VAL) | |
8619129f UD |
157 | # endif |
158 | #else | |
159 | # ifndef FROM_DIRECTION | |
dbc92dc6 | 160 | # error "FROM_DIRECTION must be provided if non-default init is used" |
8619129f UD |
161 | # endif |
162 | #endif | |
163 | ||
164 | ||
165 | /* How many bytes are needed at most for the from-charset. */ | |
166 | #ifndef MAX_NEEDED_FROM | |
167 | # define MAX_NEEDED_FROM MIN_NEEDED_FROM | |
168 | #endif | |
169 | ||
170 | /* Same for the to-charset. */ | |
171 | #ifndef MAX_NEEDED_TO | |
172 | # define MAX_NEEDED_TO MIN_NEEDED_TO | |
173 | #endif | |
174 | ||
faaa6f62 UD |
175 | /* Defaults for the per-direction min/max constants. */ |
176 | #ifndef FROM_LOOP_MIN_NEEDED_FROM | |
177 | # define FROM_LOOP_MIN_NEEDED_FROM MIN_NEEDED_FROM | |
178 | #endif | |
179 | #ifndef FROM_LOOP_MAX_NEEDED_FROM | |
180 | # define FROM_LOOP_MAX_NEEDED_FROM MAX_NEEDED_FROM | |
181 | #endif | |
182 | #ifndef FROM_LOOP_MIN_NEEDED_TO | |
183 | # define FROM_LOOP_MIN_NEEDED_TO MIN_NEEDED_TO | |
184 | #endif | |
185 | #ifndef FROM_LOOP_MAX_NEEDED_TO | |
186 | # define FROM_LOOP_MAX_NEEDED_TO MAX_NEEDED_TO | |
187 | #endif | |
188 | #ifndef TO_LOOP_MIN_NEEDED_FROM | |
189 | # define TO_LOOP_MIN_NEEDED_FROM MIN_NEEDED_TO | |
190 | #endif | |
191 | #ifndef TO_LOOP_MAX_NEEDED_FROM | |
192 | # define TO_LOOP_MAX_NEEDED_FROM MAX_NEEDED_TO | |
193 | #endif | |
194 | #ifndef TO_LOOP_MIN_NEEDED_TO | |
195 | # define TO_LOOP_MIN_NEEDED_TO MIN_NEEDED_FROM | |
196 | #endif | |
197 | #ifndef TO_LOOP_MAX_NEEDED_TO | |
198 | # define TO_LOOP_MAX_NEEDED_TO MAX_NEEDED_FROM | |
199 | #endif | |
200 | ||
8619129f | 201 | |
77e1d15a UD |
202 | /* Define macros which can access unaligned buffers. These macros are |
203 | supposed to be used only in code outside the inner loops. For the inner | |
204 | loops we have other definitions which allow optimized access. */ | |
205 | #ifdef _STRING_ARCH_unaligned | |
206 | /* We can handle unaligned memory access. */ | |
8c0b7170 UD |
207 | # define get16u(addr) *((__const uint16_t *) (addr)) |
208 | # define get32u(addr) *((__const uint32_t *) (addr)) | |
77e1d15a UD |
209 | |
210 | /* We need no special support for writing values either. */ | |
c1db8b0d UD |
211 | # define put16u(addr, val) *((uint16_t *) (addr)) = (val) |
212 | # define put32u(addr, val) *((uint32_t *) (addr)) = (val) | |
77e1d15a UD |
213 | #else |
214 | /* Distinguish between big endian and little endian. */ | |
215 | # if __BYTE_ORDER == __LITTLE_ENDIAN | |
c1db8b0d | 216 | # define get16u(addr) \ |
77e1d15a UD |
217 | (((__const unsigned char *) (addr))[1] << 8 \ |
218 | | ((__const unsigned char *) (addr))[0]) | |
c1db8b0d | 219 | # define get32u(addr) \ |
77e1d15a UD |
220 | (((((__const unsigned char *) (addr))[3] << 8 \ |
221 | | ((__const unsigned char *) (addr))[2]) << 8 \ | |
222 | | ((__const unsigned char *) (addr))[1]) << 8 \ | |
223 | | ((__const unsigned char *) (addr))[0]) | |
224 | ||
c1db8b0d | 225 | # define put16u(addr, val) \ |
77e1d15a | 226 | ({ uint16_t __val = (val); \ |
cb2c5501 UD |
227 | ((unsigned char *) (addr))[0] = __val; \ |
228 | ((unsigned char *) (addr))[1] = __val >> 8; \ | |
77e1d15a | 229 | (void) 0; }) |
c1db8b0d | 230 | # define put32u(addr, val) \ |
cb2c5501 UD |
231 | ({ uint32_t __val = (val); \ |
232 | ((unsigned char *) (addr))[0] = __val; \ | |
77e1d15a | 233 | __val >>= 8; \ |
cb2c5501 | 234 | ((unsigned char *) (addr))[1] = __val; \ |
77e1d15a | 235 | __val >>= 8; \ |
cb2c5501 | 236 | ((unsigned char *) (addr))[2] = __val; \ |
77e1d15a | 237 | __val >>= 8; \ |
cb2c5501 | 238 | ((unsigned char *) (addr))[3] = __val; \ |
77e1d15a UD |
239 | (void) 0; }) |
240 | # else | |
c1db8b0d | 241 | # define get16u(addr) \ |
77e1d15a UD |
242 | (((__const unsigned char *) (addr))[0] << 8 \ |
243 | | ((__const unsigned char *) (addr))[1]) | |
c1db8b0d | 244 | # define get32u(addr) \ |
77e1d15a UD |
245 | (((((__const unsigned char *) (addr))[0] << 8 \ |
246 | | ((__const unsigned char *) (addr))[1]) << 8 \ | |
247 | | ((__const unsigned char *) (addr))[2]) << 8 \ | |
248 | | ((__const unsigned char *) (addr))[3]) | |
249 | ||
c1db8b0d | 250 | # define put16u(addr, val) \ |
77e1d15a | 251 | ({ uint16_t __val = (val); \ |
cb2c5501 | 252 | ((unsigned char *) (addr))[1] = __val; \ |
a1303dc8 | 253 | ((unsigned char *) (addr))[0] = __val >> 8; \ |
77e1d15a | 254 | (void) 0; }) |
c1db8b0d | 255 | # define put32u(addr, val) \ |
cb2c5501 UD |
256 | ({ uint32_t __val = (val); \ |
257 | ((unsigned char *) (addr))[3] = __val; \ | |
77e1d15a | 258 | __val >>= 8; \ |
cb2c5501 | 259 | ((unsigned char *) (addr))[2] = __val; \ |
77e1d15a | 260 | __val >>= 8; \ |
cb2c5501 | 261 | ((unsigned char *) (addr))[1] = __val; \ |
77e1d15a | 262 | __val >>= 8; \ |
cb2c5501 | 263 | ((unsigned char *) (addr))[0] = __val; \ |
77e1d15a UD |
264 | (void) 0; }) |
265 | # endif | |
266 | #endif | |
267 | ||
268 | ||
f6ad4726 UD |
269 | /* For conversions from a fixed width character set to another fixed width |
270 | character set we can define RESET_INPUT_BUFFER in a very fast way. */ | |
8619129f | 271 | #if !defined RESET_INPUT_BUFFER && !defined SAVE_RESET_STATE |
faaa6f62 UD |
272 | # if FROM_LOOP_MIN_NEEDED_FROM == FROM_LOOP_MAX_NEEDED_FROM \ |
273 | && FROM_LOOP_MIN_NEEDED_TO == FROM_LOOP_MAX_NEEDED_TO \ | |
274 | && TO_LOOP_MIN_NEEDED_FROM == TO_LOOP_MAX_NEEDED_FROM \ | |
275 | && TO_LOOP_MIN_NEEDED_TO == TO_LOOP_MAX_NEEDED_TO | |
276 | /* We have to use these `if's here since the compiler cannot know that | |
277 | (outbuf - outerr) is always divisible by FROM/TO_LOOP_MIN_NEEDED_TO. | |
278 | The ?:1 avoids division by zero warnings that gcc 3.2 emits even for | |
279 | obviously unreachable code. */ | |
280 | # define RESET_INPUT_BUFFER \ | |
281 | if (FROM_DIRECTION) \ | |
282 | { \ | |
283 | if (FROM_LOOP_MIN_NEEDED_FROM % FROM_LOOP_MIN_NEEDED_TO == 0) \ | |
284 | *inptrp -= (outbuf - outerr) \ | |
285 | * (FROM_LOOP_MIN_NEEDED_FROM / FROM_LOOP_MIN_NEEDED_TO); \ | |
286 | else if (FROM_LOOP_MIN_NEEDED_TO % FROM_LOOP_MIN_NEEDED_FROM == 0) \ | |
287 | *inptrp -= (outbuf - outerr) \ | |
288 | / (FROM_LOOP_MIN_NEEDED_TO / FROM_LOOP_MIN_NEEDED_FROM \ | |
289 | ? : 1); \ | |
290 | else \ | |
291 | *inptrp -= ((outbuf - outerr) / FROM_LOOP_MIN_NEEDED_TO) \ | |
292 | * FROM_LOOP_MIN_NEEDED_FROM; \ | |
293 | } \ | |
294 | else \ | |
295 | { \ | |
296 | if (TO_LOOP_MIN_NEEDED_FROM % TO_LOOP_MIN_NEEDED_TO == 0) \ | |
297 | *inptrp -= (outbuf - outerr) \ | |
298 | * (TO_LOOP_MIN_NEEDED_FROM / TO_LOOP_MIN_NEEDED_TO); \ | |
299 | else if (TO_LOOP_MIN_NEEDED_TO % TO_LOOP_MIN_NEEDED_FROM == 0) \ | |
300 | *inptrp -= (outbuf - outerr) \ | |
301 | / (TO_LOOP_MIN_NEEDED_TO / TO_LOOP_MIN_NEEDED_FROM ? : 1); \ | |
302 | else \ | |
303 | *inptrp -= ((outbuf - outerr) / TO_LOOP_MIN_NEEDED_TO) \ | |
304 | * TO_LOOP_MIN_NEEDED_FROM; \ | |
305 | } | |
8619129f UD |
306 | # endif |
307 | #endif | |
308 | ||
309 | ||
310 | /* The default init function. It simply matches the name and initializes | |
311 | the step data to point to one of the objects above. */ | |
312 | #if DEFINE_INIT | |
313 | # ifndef CHARSET_NAME | |
314 | # error "CHARSET_NAME not defined" | |
315 | # endif | |
316 | ||
a1620a4c | 317 | extern int gconv_init (struct __gconv_step *step); |
8619129f | 318 | int |
d64b6ad0 | 319 | gconv_init (struct __gconv_step *step) |
8619129f UD |
320 | { |
321 | /* Determine which direction. */ | |
d64b6ad0 | 322 | if (strcmp (step->__from_name, CHARSET_NAME) == 0) |
918b9d72 | 323 | { |
45541bdf | 324 | step->__data = FROM_DIRECTION_VAL; |
390955cb | 325 | |
faaa6f62 UD |
326 | step->__min_needed_from = FROM_LOOP_MIN_NEEDED_FROM; |
327 | step->__max_needed_from = FROM_LOOP_MAX_NEEDED_FROM; | |
328 | step->__min_needed_to = FROM_LOOP_MIN_NEEDED_TO; | |
329 | step->__max_needed_to = FROM_LOOP_MAX_NEEDED_TO; | |
f9ad060c UD |
330 | |
331 | #ifdef FROM_ONEBYTE | |
332 | step->__btowc_fct = FROM_ONEBYTE; | |
333 | #endif | |
918b9d72 | 334 | } |
365afefc | 335 | else if (__builtin_expect (strcmp (step->__to_name, CHARSET_NAME), 0) == 0) |
918b9d72 | 336 | { |
45541bdf | 337 | step->__data = TO_DIRECTION_VAL; |
390955cb | 338 | |
faaa6f62 UD |
339 | step->__min_needed_from = TO_LOOP_MIN_NEEDED_FROM; |
340 | step->__max_needed_from = TO_LOOP_MAX_NEEDED_FROM; | |
341 | step->__min_needed_to = TO_LOOP_MIN_NEEDED_TO; | |
342 | step->__max_needed_to = TO_LOOP_MAX_NEEDED_TO; | |
918b9d72 | 343 | } |
390955cb | 344 | else |
d64b6ad0 | 345 | return __GCONV_NOCONV; |
8619129f | 346 | |
f6ad4726 | 347 | #ifdef SAVE_RESET_STATE |
d64b6ad0 | 348 | step->__stateful = 1; |
9ce5071a | 349 | #else |
d64b6ad0 | 350 | step->__stateful = 0; |
9ce5071a UD |
351 | #endif |
352 | ||
d64b6ad0 | 353 | return __GCONV_OK; |
8619129f UD |
354 | } |
355 | #endif | |
356 | ||
357 | ||
358 | /* The default destructor function does nothing in the moment and so | |
f6ad4726 UD |
359 | we don't define it at all. But we still provide the macro just in |
360 | case we need it some day. */ | |
8619129f UD |
361 | #if DEFINE_FINI |
362 | #endif | |
363 | ||
364 | ||
28f1c862 UD |
365 | /* If no arguments have to passed to the loop function define the macro |
366 | as empty. */ | |
367 | #ifndef EXTRA_LOOP_ARGS | |
368 | # define EXTRA_LOOP_ARGS | |
369 | #endif | |
370 | ||
371 | ||
8619129f UD |
372 | /* This is the actual conversion function. */ |
373 | #ifndef FUNCTION_NAME | |
374 | # define FUNCTION_NAME gconv | |
375 | #endif | |
376 | ||
fd1b5c0f UD |
377 | /* The macros are used to access the function to convert single characters. */ |
378 | #define SINGLE(fct) SINGLE2 (fct) | |
379 | #define SINGLE2(fct) fct##_single | |
380 | ||
381 | ||
a1620a4c UD |
382 | extern int FUNCTION_NAME (struct __gconv_step *step, |
383 | struct __gconv_step_data *data, | |
384 | const unsigned char **inptrp, | |
385 | const unsigned char *inend, | |
386 | unsigned char **outbufstart, size_t *irreversible, | |
387 | int do_flush, int consume_incomplete); | |
8619129f | 388 | int |
d64b6ad0 | 389 | FUNCTION_NAME (struct __gconv_step *step, struct __gconv_step_data *data, |
fd1b5c0f | 390 | const unsigned char **inptrp, const unsigned char *inend, |
f1d5c60d | 391 | unsigned char **outbufstart, size_t *irreversible, int do_flush, |
55985355 | 392 | int consume_incomplete) |
8619129f | 393 | { |
d64b6ad0 UD |
394 | struct __gconv_step *next_step = step + 1; |
395 | struct __gconv_step_data *next_data = data + 1; | |
a334319f | 396 | __gconv_fct fct; |
8619129f UD |
397 | int status; |
398 | ||
a334319f | 399 | fct = (data->__flags & __GCONV_IS_LAST) ? NULL : next_step->__fct; |
85830c4c | 400 | |
8619129f UD |
401 | /* If the function is called with no input this means we have to reset |
402 | to the initial state. The possibly partly converted input is | |
403 | dropped. */ | |
978ce92b | 404 | if (__builtin_expect (do_flush, 0)) |
8619129f | 405 | { |
f1d5c60d UD |
406 | /* This should never happen during error handling. */ |
407 | assert (outbufstart == NULL); | |
408 | ||
2373b30e UD |
409 | status = __GCONV_OK; |
410 | ||
390955cb | 411 | #ifdef EMIT_SHIFT_TO_INIT |
2373b30e UD |
412 | if (do_flush == 1) |
413 | { | |
414 | /* We preserve the initial values of the pointer variables. */ | |
415 | unsigned char *outbuf = data->__outbuf; | |
416 | unsigned char *outstart = outbuf; | |
417 | unsigned char *outend = data->__outbufend; | |
418 | ||
419 | # ifdef PREPARE_LOOP | |
420 | PREPARE_LOOP | |
421 | # endif | |
422 | ||
423 | # ifdef SAVE_RESET_STATE | |
424 | SAVE_RESET_STATE (1); | |
425 | # endif | |
426 | ||
427 | /* Emit the escape sequence to reset the state. */ | |
428 | EMIT_SHIFT_TO_INIT; | |
429 | ||
430 | /* Call the steps down the chain if there are any but only if we | |
431 | successfully emitted the escape sequence. This should only | |
432 | fail if the output buffer is full. If the input is invalid | |
433 | it should be discarded since the user wants to start from a | |
434 | clean state. */ | |
435 | if (status == __GCONV_OK) | |
436 | { | |
437 | if (data->__flags & __GCONV_IS_LAST) | |
438 | /* Store information about how many bytes are available. */ | |
439 | data->__outbuf = outbuf; | |
440 | else | |
441 | { | |
442 | /* Write out all output which was produced. */ | |
443 | if (outbuf > outstart) | |
444 | { | |
445 | const unsigned char *outerr = outstart; | |
446 | int result; | |
447 | ||
448 | result = DL_CALL_FCT (fct, (next_step, next_data, | |
449 | &outerr, outbuf, NULL, | |
450 | irreversible, 0, | |
451 | consume_incomplete)); | |
452 | ||
453 | if (result != __GCONV_EMPTY_INPUT) | |
454 | { | |
455 | if (__builtin_expect (outerr != outbuf, 0)) | |
456 | { | |
457 | /* We have a problem. Undo the conversion. */ | |
458 | outbuf = outstart; | |
459 | ||
460 | /* Restore the state. */ | |
461 | # ifdef SAVE_RESET_STATE | |
462 | SAVE_RESET_STATE (0); | |
463 | # endif | |
464 | } | |
465 | ||
466 | /* Change the status. */ | |
467 | status = result; | |
468 | } | |
469 | } | |
470 | ||
471 | if (status == __GCONV_OK) | |
472 | /* Now flush the remaining steps. */ | |
473 | status = DL_CALL_FCT (fct, (next_step, next_data, NULL, | |
474 | NULL, NULL, irreversible, 1, | |
475 | consume_incomplete)); | |
476 | } | |
477 | } | |
478 | } | |
479 | else | |
8619129f | 480 | #endif |
2373b30e UD |
481 | { |
482 | /* Clear the state object. There might be bytes in there from | |
483 | previous calls with CONSUME_INCOMPLETE == 1. But don't emit | |
484 | escape sequences. */ | |
485 | memset (data->__statep, '\0', sizeof (*data->__statep)); | |
486 | ||
487 | if (! (data->__flags & __GCONV_IS_LAST)) | |
488 | /* Now flush the remaining steps. */ | |
489 | status = DL_CALL_FCT (fct, (next_step, next_data, NULL, NULL, | |
490 | NULL, irreversible, do_flush, | |
491 | consume_incomplete)); | |
492 | } | |
8619129f UD |
493 | } |
494 | else | |
495 | { | |
8619129f | 496 | /* We preserve the initial values of the pointer variables. */ |
fd1b5c0f | 497 | const unsigned char *inptr = *inptrp; |
f1d5c60d UD |
498 | unsigned char *outbuf = (__builtin_expect (outbufstart == NULL, 1) |
499 | ? data->__outbuf : *outbufstart); | |
d64b6ad0 | 500 | unsigned char *outend = data->__outbufend; |
b117f744 | 501 | unsigned char *outstart; |
3aaad0b6 UD |
502 | /* This variable is used to count the number of characters we |
503 | actually converted. */ | |
38677ace | 504 | size_t lirreversible = 0; |
b572c2da | 505 | size_t *lirreversiblep = irreversible ? &lirreversible : NULL; |
faaa6f62 UD |
506 | |
507 | /* The following assumes that encodings, which have a variable length | |
508 | what might unalign a buffer even though it is a aligned in the | |
509 | beginning, either don't have the minimal number of bytes as a divisor | |
510 | of the maximum length or have a minimum length of 1. This is true | |
511 | for all known and supported encodings. | |
512 | We use && instead of || to combine the subexpression for the FROM | |
513 | encoding and for the TO encoding, because usually one of them is | |
514 | INTERNAL, for which the subexpression evaluates to 1, but INTERNAL | |
515 | buffers are always aligned correctly. */ | |
516 | #define POSSIBLY_UNALIGNED \ | |
517 | (!defined _STRING_ARCH_unaligned \ | |
518 | && (((FROM_LOOP_MIN_NEEDED_FROM != 1 \ | |
519 | && FROM_LOOP_MAX_NEEDED_FROM % FROM_LOOP_MIN_NEEDED_FROM == 0) \ | |
520 | && (FROM_LOOP_MIN_NEEDED_TO != 1 \ | |
521 | && FROM_LOOP_MAX_NEEDED_TO % FROM_LOOP_MIN_NEEDED_TO == 0)) \ | |
522 | || ((TO_LOOP_MIN_NEEDED_FROM != 1 \ | |
523 | && TO_LOOP_MAX_NEEDED_FROM % TO_LOOP_MIN_NEEDED_FROM == 0) \ | |
524 | && (TO_LOOP_MIN_NEEDED_TO != 1 \ | |
525 | && TO_LOOP_MAX_NEEDED_TO % TO_LOOP_MIN_NEEDED_TO == 0)))) | |
526 | #if POSSIBLY_UNALIGNED | |
3aaad0b6 | 527 | int unaligned; |
77e1d15a UD |
528 | # define GEN_unaligned(name) GEN_unaligned2 (name) |
529 | # define GEN_unaligned2(name) name##_unaligned | |
faaa6f62 UD |
530 | #else |
531 | # define unaligned 0 | |
77e1d15a | 532 | #endif |
8619129f | 533 | |
28f1c862 UD |
534 | #ifdef PREPARE_LOOP |
535 | PREPARE_LOOP | |
536 | #endif | |
537 | ||
faaa6f62 | 538 | #if FROM_LOOP_MAX_NEEDED_FROM > 1 || TO_LOOP_MAX_NEEDED_FROM > 1 |
fd1b5c0f UD |
539 | /* If the function is used to implement the mb*towc*() or wc*tomb*() |
540 | functions we must test whether any bytes from the last call are | |
541 | stored in the `state' object. */ | |
faaa6f62 UD |
542 | if (((FROM_LOOP_MAX_NEEDED_FROM > 1 && TO_LOOP_MAX_NEEDED_FROM > 1) |
543 | || (FROM_LOOP_MAX_NEEDED_FROM > 1 && FROM_DIRECTION) | |
544 | || (TO_LOOP_MAX_NEEDED_FROM > 1 && !FROM_DIRECTION)) | |
fd1b5c0f UD |
545 | && consume_incomplete && (data->__statep->__count & 7) != 0) |
546 | { | |
f1d5c60d UD |
547 | /* Yep, we have some bytes left over. Process them now. |
548 | But this must not happen while we are called from an | |
549 | error handler. */ | |
550 | assert (outbufstart == NULL); | |
fd1b5c0f | 551 | |
faaa6f62 UD |
552 | # if FROM_LOOP_MAX_NEEDED_FROM > 1 |
553 | if (TO_LOOP_MAX_NEEDED_FROM == 1 || FROM_DIRECTION) | |
55985355 | 554 | status = SINGLE(FROM_LOOP) (step, data, inptrp, inend, &outbuf, |
b572c2da | 555 | outend, lirreversiblep |
85830c4c | 556 | EXTRA_LOOP_ARGS); |
fd1b5c0f | 557 | # endif |
faaa6f62 UD |
558 | # if !ONE_DIRECTION |
559 | # if FROM_LOOP_MAX_NEEDED_FROM > 1 && TO_LOOP_MAX_NEEDED_FROM > 1 | |
fd1b5c0f | 560 | else |
faaa6f62 UD |
561 | # endif |
562 | # if TO_LOOP_MAX_NEEDED_FROM > 1 | |
55985355 | 563 | status = SINGLE(TO_LOOP) (step, data, inptrp, inend, &outbuf, |
b572c2da | 564 | outend, lirreversiblep EXTRA_LOOP_ARGS); |
faaa6f62 | 565 | # endif |
fd1b5c0f UD |
566 | # endif |
567 | ||
365afefc | 568 | if (__builtin_expect (status, __GCONV_OK) != __GCONV_OK) |
fd1b5c0f UD |
569 | return status; |
570 | } | |
571 | #endif | |
572 | ||
faaa6f62 UD |
573 | #if POSSIBLY_UNALIGNED |
574 | unaligned = | |
575 | ((FROM_DIRECTION | |
576 | && ((uintptr_t) inptr % FROM_LOOP_MIN_NEEDED_FROM != 0 | |
577 | || ((data->__flags & __GCONV_IS_LAST) | |
578 | && (uintptr_t) outbuf % FROM_LOOP_MIN_NEEDED_TO != 0))) | |
579 | || (!FROM_DIRECTION | |
580 | && (((data->__flags & __GCONV_IS_LAST) | |
581 | && (uintptr_t) outbuf % TO_LOOP_MIN_NEEDED_TO != 0) | |
582 | || (uintptr_t) inptr % TO_LOOP_MIN_NEEDED_FROM != 0))); | |
3aaad0b6 UD |
583 | #endif |
584 | ||
f1813b56 | 585 | while (1) |
8619129f | 586 | { |
d6204268 UD |
587 | struct __gconv_trans_data *trans; |
588 | ||
8619129f | 589 | /* Remember the start value for this round. */ |
fd1b5c0f | 590 | inptr = *inptrp; |
8619129f | 591 | /* The outbuf buffer is empty. */ |
b117f744 | 592 | outstart = outbuf; |
8619129f | 593 | |
8619129f UD |
594 | #ifdef SAVE_RESET_STATE |
595 | SAVE_RESET_STATE (1); | |
596 | #endif | |
597 | ||
365afefc | 598 | if (__builtin_expect (!unaligned, 1)) |
77e1d15a UD |
599 | { |
600 | if (FROM_DIRECTION) | |
601 | /* Run the conversion loop. */ | |
55985355 | 602 | status = FROM_LOOP (step, data, inptrp, inend, &outbuf, outend, |
b572c2da | 603 | lirreversiblep EXTRA_LOOP_ARGS); |
77e1d15a UD |
604 | else |
605 | /* Run the conversion loop. */ | |
55985355 | 606 | status = TO_LOOP (step, data, inptrp, inend, &outbuf, outend, |
b572c2da | 607 | lirreversiblep EXTRA_LOOP_ARGS); |
77e1d15a | 608 | } |
faaa6f62 | 609 | #if POSSIBLY_UNALIGNED |
8619129f | 610 | else |
77e1d15a UD |
611 | { |
612 | if (FROM_DIRECTION) | |
613 | /* Run the conversion loop. */ | |
55985355 UD |
614 | status = GEN_unaligned (FROM_LOOP) (step, data, inptrp, inend, |
615 | &outbuf, outend, | |
b572c2da | 616 | lirreversiblep |
77e1d15a UD |
617 | EXTRA_LOOP_ARGS); |
618 | else | |
619 | /* Run the conversion loop. */ | |
55985355 UD |
620 | status = GEN_unaligned (TO_LOOP) (step, data, inptrp, inend, |
621 | &outbuf, outend, | |
b572c2da | 622 | lirreversiblep |
77e1d15a UD |
623 | EXTRA_LOOP_ARGS); |
624 | } | |
625 | #endif | |
8619129f | 626 | |
f1d5c60d UD |
627 | /* If we were called as part of an error handling module we |
628 | don't do anything else here. */ | |
629 | if (__builtin_expect (outbufstart != NULL, 0)) | |
630 | { | |
631 | *outbufstart = outbuf; | |
632 | return status; | |
633 | } | |
634 | ||
4d37e19c UD |
635 | /* Give the transliteration module the chance to store the |
636 | original text and the result in case it needs a context. */ | |
d6204268 UD |
637 | for (trans = data->__trans; trans != NULL; trans = trans->__next) |
638 | if (trans->__trans_context_fct != NULL) | |
639 | DL_CALL_FCT (trans->__trans_context_fct, | |
640 | (trans->__data, inptr, *inptrp, outstart, outbuf)); | |
4d37e19c | 641 | |
0aece08d UD |
642 | /* We finished one use of the loops. */ |
643 | ++data->__invocation_counter; | |
644 | ||
390955cb | 645 | /* If this is the last step leave the loop, there is nothing |
8619129f | 646 | we can do. */ |
365afefc | 647 | if (__builtin_expect (data->__flags & __GCONV_IS_LAST, 0)) |
8619129f UD |
648 | { |
649 | /* Store information about how many bytes are available. */ | |
d64b6ad0 | 650 | data->__outbuf = outbuf; |
bf979eef | 651 | |
38677ace UD |
652 | /* Remember how many non-identical characters we |
653 | converted in a irreversible way. */ | |
654 | *irreversible += lirreversible; | |
bf979eef | 655 | |
8619129f UD |
656 | break; |
657 | } | |
658 | ||
659 | /* Write out all output which was produced. */ | |
365afefc | 660 | if (__builtin_expect (outbuf > outstart, 1)) |
8619129f | 661 | { |
d64b6ad0 | 662 | const unsigned char *outerr = data->__outbuf; |
8619129f UD |
663 | int result; |
664 | ||
4bae5567 | 665 | result = DL_CALL_FCT (fct, (next_step, next_data, &outerr, |
f1d5c60d | 666 | outbuf, NULL, irreversible, 0, |
fd1b5c0f | 667 | consume_incomplete)); |
8619129f | 668 | |
d64b6ad0 | 669 | if (result != __GCONV_EMPTY_INPUT) |
8619129f | 670 | { |
978ce92b | 671 | if (__builtin_expect (outerr != outbuf, 0)) |
8619129f UD |
672 | { |
673 | #ifdef RESET_INPUT_BUFFER | |
674 | RESET_INPUT_BUFFER; | |
675 | #else | |
dbc92dc6 UD |
676 | /* We have a problem in one of the functions below. |
677 | Undo the conversion upto the error point. */ | |
8619129f UD |
678 | size_t nstatus; |
679 | ||
680 | /* Reload the pointers. */ | |
fd1b5c0f | 681 | *inptrp = inptr; |
b117f744 | 682 | outbuf = outstart; |
8619129f | 683 | |
2373b30e | 684 | /* Restore the state. */ |
8619129f UD |
685 | # ifdef SAVE_RESET_STATE |
686 | SAVE_RESET_STATE (0); | |
687 | # endif | |
688 | ||
f6ad4726 UD |
689 | if (__builtin_expect (!unaligned, 1)) |
690 | { | |
691 | if (FROM_DIRECTION) | |
692 | /* Run the conversion loop. */ | |
693 | nstatus = FROM_LOOP (step, data, inptrp, inend, | |
17427edd | 694 | &outbuf, outerr, |
f6ad4726 UD |
695 | lirreversiblep |
696 | EXTRA_LOOP_ARGS); | |
697 | else | |
698 | /* Run the conversion loop. */ | |
699 | nstatus = TO_LOOP (step, data, inptrp, inend, | |
17427edd | 700 | &outbuf, outerr, |
f6ad4726 UD |
701 | lirreversiblep |
702 | EXTRA_LOOP_ARGS); | |
703 | } | |
faaa6f62 | 704 | # if POSSIBLY_UNALIGNED |
8619129f | 705 | else |
f6ad4726 UD |
706 | { |
707 | if (FROM_DIRECTION) | |
708 | /* Run the conversion loop. */ | |
709 | nstatus = GEN_unaligned (FROM_LOOP) (step, data, | |
710 | inptrp, inend, | |
d1700665 | 711 | &outbuf, |
17427edd | 712 | outerr, |
f6ad4726 UD |
713 | lirreversiblep |
714 | EXTRA_LOOP_ARGS); | |
715 | else | |
716 | /* Run the conversion loop. */ | |
717 | nstatus = GEN_unaligned (TO_LOOP) (step, data, | |
718 | inptrp, inend, | |
17427edd | 719 | &outbuf, outerr, |
f6ad4726 UD |
720 | lirreversiblep |
721 | EXTRA_LOOP_ARGS); | |
722 | } | |
723 | # endif | |
8619129f UD |
724 | |
725 | /* We must run out of output buffer space in this | |
726 | rerun. */ | |
5aa8ff62 | 727 | assert (outbuf == outerr); |
d64b6ad0 | 728 | assert (nstatus == __GCONV_FULL_OUTPUT); |
0aece08d UD |
729 | |
730 | /* If we haven't consumed a single byte decrement | |
731 | the invocation counter. */ | |
365afefc | 732 | if (__builtin_expect (outbuf == outstart, 0)) |
0aece08d | 733 | --data->__invocation_counter; |
8619129f UD |
734 | #endif /* reset input buffer */ |
735 | } | |
736 | ||
737 | /* Change the status. */ | |
738 | status = result; | |
739 | } | |
740 | else | |
741 | /* All the output is consumed, we can make another run | |
742 | if everything was ok. */ | |
d64b6ad0 | 743 | if (status == __GCONV_FULL_OUTPUT) |
f1813b56 UD |
744 | { |
745 | status = __GCONV_OK; | |
746 | outbuf = data->__outbuf; | |
747 | } | |
8619129f | 748 | } |
f1813b56 UD |
749 | |
750 | if (status != __GCONV_OK) | |
751 | break; | |
752 | ||
753 | /* Reset the output buffer pointer for the next round. */ | |
754 | outbuf = data->__outbuf; | |
8619129f | 755 | } |
8619129f | 756 | |
918b9d72 UD |
757 | #ifdef END_LOOP |
758 | END_LOOP | |
759 | #endif | |
fd1b5c0f UD |
760 | |
761 | /* If we are supposed to consume all character store now all of the | |
762 | remaining characters in the `state' object. */ | |
faaa6f62 UD |
763 | #if FROM_LOOP_MAX_NEEDED_FROM > 1 || TO_LOOP_MAX_NEEDED_FROM > 1 |
764 | if (((FROM_LOOP_MAX_NEEDED_FROM > 1 && TO_LOOP_MAX_NEEDED_FROM > 1) | |
765 | || (FROM_LOOP_MAX_NEEDED_FROM > 1 && FROM_DIRECTION) | |
766 | || (TO_LOOP_MAX_NEEDED_FROM > 1 && !FROM_DIRECTION)) | |
365afefc UD |
767 | && __builtin_expect (consume_incomplete, 0) |
768 | && status == __GCONV_INCOMPLETE_INPUT) | |
fd1b5c0f UD |
769 | { |
770 | # ifdef STORE_REST | |
771 | mbstate_t *state = data->__statep; | |
772 | ||
773 | STORE_REST | |
774 | # else | |
fd1b5c0f UD |
775 | /* Make sure the remaining bytes fit into the state objects |
776 | buffer. */ | |
777 | assert (inend - *inptrp < 4); | |
778 | ||
dbc92dc6 | 779 | size_t cnt; |
fd1b5c0f UD |
780 | for (cnt = 0; *inptrp < inend; ++cnt) |
781 | data->__statep->__value.__wchb[cnt] = *(*inptrp)++; | |
782 | data->__statep->__count &= ~7; | |
783 | data->__statep->__count |= cnt; | |
784 | # endif | |
785 | } | |
786 | #endif | |
faaa6f62 UD |
787 | #undef unaligned |
788 | #undef POSSIBLY_UNALIGNED | |
8619129f UD |
789 | } |
790 | ||
791 | return status; | |
792 | } | |
793 | ||
794 | #undef DEFINE_INIT | |
795 | #undef CHARSET_NAME | |
796 | #undef DEFINE_FINI | |
797 | #undef MIN_NEEDED_FROM | |
798 | #undef MIN_NEEDED_TO | |
799 | #undef MAX_NEEDED_FROM | |
800 | #undef MAX_NEEDED_TO | |
faaa6f62 UD |
801 | #undef FROM_LOOP_MIN_NEEDED_FROM |
802 | #undef FROM_LOOP_MAX_NEEDED_FROM | |
803 | #undef FROM_LOOP_MIN_NEEDED_TO | |
804 | #undef FROM_LOOP_MAX_NEEDED_TO | |
805 | #undef TO_LOOP_MIN_NEEDED_FROM | |
806 | #undef TO_LOOP_MAX_NEEDED_FROM | |
807 | #undef TO_LOOP_MIN_NEEDED_TO | |
808 | #undef TO_LOOP_MAX_NEEDED_TO | |
8619129f UD |
809 | #undef FROM_DIRECTION |
810 | #undef EMIT_SHIFT_TO_INIT | |
811 | #undef FROM_LOOP | |
812 | #undef TO_LOOP | |
f9ad060c | 813 | #undef ONE_DIRECTION |
f6ad4726 | 814 | #undef SAVE_RESET_STATE |
8619129f UD |
815 | #undef RESET_INPUT_BUFFER |
816 | #undef FUNCTION_NAME | |
918b9d72 UD |
817 | #undef PREPARE_LOOP |
818 | #undef END_LOOP | |
f9ad060c | 819 | #undef EXTRA_LOOP_ARGS |
fd1b5c0f | 820 | #undef STORE_REST |
f9ad060c | 821 | #undef FROM_ONEBYTE |