]>
Commit | Line | Data |
---|---|---|
09376451 | 1 | /* Skeleton for a conversion module. |
d4697bc9 | 2 | Copyright (C) 1998-2014 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 | 16 | You should have received a copy of the GNU Lesser General Public |
59ba27a6 PE |
17 | License along with the GNU C Library; if not, see |
18 | <http://www.gnu.org/licenses/>. */ | |
8619129f UD |
19 | |
20 | /* This file can be included to provide definitions of several things | |
21 | many modules have in common. It can be customized using the following | |
22 | macros: | |
23 | ||
24 | DEFINE_INIT define the default initializer. This requires the | |
25 | following symbol to be defined. | |
26 | ||
27 | CHARSET_NAME string with official name of the coded character | |
28 | set (in all-caps) | |
29 | ||
30 | DEFINE_FINI define the default destructor function. | |
31 | ||
32 | MIN_NEEDED_FROM minimal number of bytes needed for the from-charset. | |
33 | MIN_NEEDED_TO likewise for the to-charset. | |
34 | ||
35 | MAX_NEEDED_FROM maximal number of bytes needed for the from-charset. | |
36 | This macro is optional, it defaults to MIN_NEEDED_FROM. | |
37 | MAX_NEEDED_TO likewise for the to-charset. | |
38 | ||
faaa6f62 UD |
39 | FROM_LOOP_MIN_NEEDED_FROM |
40 | FROM_LOOP_MAX_NEEDED_FROM | |
a784e502 UD |
41 | minimal/maximal number of bytes needed on input |
42 | of one round through the FROM_LOOP. Defaults | |
43 | to MIN_NEEDED_FROM and MAX_NEEDED_FROM, respectively. | |
faaa6f62 UD |
44 | FROM_LOOP_MIN_NEEDED_TO |
45 | FROM_LOOP_MAX_NEEDED_TO | |
a784e502 UD |
46 | minimal/maximal number of bytes needed on output |
47 | of one round through the FROM_LOOP. Defaults | |
48 | to MIN_NEEDED_TO and MAX_NEEDED_TO, respectively. | |
faaa6f62 UD |
49 | TO_LOOP_MIN_NEEDED_FROM |
50 | TO_LOOP_MAX_NEEDED_FROM | |
a784e502 UD |
51 | minimal/maximal number of bytes needed on input |
52 | of one round through the TO_LOOP. Defaults | |
53 | to MIN_NEEDED_TO and MAX_NEEDED_TO, respectively. | |
faaa6f62 UD |
54 | TO_LOOP_MIN_NEEDED_TO |
55 | TO_LOOP_MAX_NEEDED_TO | |
a784e502 UD |
56 | minimal/maximal number of bytes needed on output |
57 | of one round through the TO_LOOP. Defaults | |
58 | to MIN_NEEDED_FROM and MAX_NEEDED_FROM, respectively. | |
faaa6f62 | 59 | |
8619129f UD |
60 | FROM_DIRECTION this macro is supposed to return a value != 0 |
61 | if we convert from the current character set, | |
62 | otherwise it return 0. | |
63 | ||
64 | EMIT_SHIFT_TO_INIT this symbol is optional. If it is defined it | |
65 | defines some code which writes out a sequence | |
faaa6f62 | 66 | of bytes which bring the current state into |
8619129f UD |
67 | the initial state. |
68 | ||
69 | FROM_LOOP name of the function implementing the conversion | |
faaa6f62 | 70 | from the current character set. |
8619129f UD |
71 | TO_LOOP likewise for the other direction |
72 | ||
f6ad4726 UD |
73 | ONE_DIRECTION optional. If defined to 1, only one conversion |
74 | direction is defined instead of two. In this | |
75 | case, FROM_DIRECTION should be defined to 1, and | |
76 | FROM_LOOP and TO_LOOP should have the same value. | |
77 | ||
78 | SAVE_RESET_STATE in case of an error we must reset the state for | |
8619129f UD |
79 | the rerun so this macro must be defined for |
80 | stateful encodings. It takes an argument which | |
81 | is nonzero when saving. | |
82 | ||
83 | RESET_INPUT_BUFFER If the input character sets allow this the macro | |
84 | can be defined to reset the input buffer pointers | |
85 | to cover only those characters up to the error. | |
86 | ||
87 | FUNCTION_NAME if not set the conversion function is named `gconv'. | |
28f1c862 UD |
88 | |
89 | PREPARE_LOOP optional code preparing the conversion loop. Can | |
90 | contain variable definitions. | |
918b9d72 | 91 | END_LOOP also optional, may be used to store information |
28f1c862 UD |
92 | |
93 | EXTRA_LOOP_ARGS optional macro specifying extra arguments passed | |
94 | to loop function. | |
faaa6f62 | 95 | |
f9ad060c UD |
96 | STORE_REST optional, needed only when MAX_NEEDED_FROM > 4. |
97 | This macro stores the seen but unconverted input bytes | |
98 | in the state. | |
99 | ||
100 | FROM_ONEBYTE optional. If defined, should be the name of a | |
101 | specialized conversion function for a single byte | |
102 | from the current character set to INTERNAL. This | |
103 | function has prototype | |
104 | wint_t | |
105 | FROM_ONEBYTE (struct __gconv_step *, unsigned char); | |
106 | and does a special conversion: | |
107 | - The input is a single byte. | |
108 | - The output is a single uint32_t. | |
109 | - The state before the conversion is the initial state; | |
110 | the state after the conversion is irrelevant. | |
111 | - No transliteration. | |
112 | - __invocation_counter = 0. | |
113 | - __internal_use = 1. | |
114 | - do_flush = 0. | |
115 | ||
faaa6f62 UD |
116 | Modules can use mbstate_t to store conversion state as follows: |
117 | ||
118 | * Bits 2..0 of '__count' contain the number of lookahead input bytes | |
119 | stored in __value.__wchb. Always zero if the converter never | |
120 | returns __GCONV_INCOMPLETE_INPUT. | |
121 | ||
122 | * Bits 31..3 of '__count' are module dependent shift state. | |
123 | ||
124 | * __value: When STORE_REST/UNPACK_BYTES aren't defined and when the | |
125 | converter has returned __GCONV_INCOMPLETE_INPUT, this contains | |
126 | at most 4 lookahead bytes. Converters with an mb_cur_max > 4 | |
127 | (currently only UTF-8) must find a way to store their state | |
128 | in __value.__wch and define STORE_REST/UNPACK_BYTES appropriately. | |
129 | ||
130 | When __value contains lookahead, __count must not be zero, because | |
131 | the converter is not in the initial state then, and mbsinit() -- | |
132 | defined as a (__count == 0) test -- must reflect this. | |
8619129f UD |
133 | */ |
134 | ||
135 | #include <assert.h> | |
136 | #include <gconv.h> | |
137 | #include <string.h> | |
138 | #define __need_size_t | |
139 | #define __need_NULL | |
140 | #include <stddef.h> | |
c66dbe00 UD |
141 | |
142 | #ifndef STATIC_GCONV | |
143 | # include <dlfcn.h> | |
c66dbe00 | 144 | #endif |
8619129f | 145 | |
915a6c51 | 146 | #include <sysdep.h> |
e054f494 | 147 | #include <stdint.h> |
915a6c51 | 148 | |
a808d541 UD |
149 | #ifndef DL_CALL_FCT |
150 | # define DL_CALL_FCT(fct, args) fct args | |
151 | #endif | |
152 | ||
8619129f | 153 | /* The direction objects. */ |
45541bdf | 154 | #if DEFINE_INIT |
8619129f | 155 | # ifndef FROM_DIRECTION |
45541bdf UD |
156 | # define FROM_DIRECTION_VAL NULL |
157 | # define TO_DIRECTION_VAL ((void *) ~((uintptr_t) 0)) | |
158 | # define FROM_DIRECTION (step->__data == FROM_DIRECTION_VAL) | |
8619129f UD |
159 | # endif |
160 | #else | |
161 | # ifndef FROM_DIRECTION | |
dbc92dc6 | 162 | # error "FROM_DIRECTION must be provided if non-default init is used" |
8619129f UD |
163 | # endif |
164 | #endif | |
165 | ||
166 | ||
167 | /* How many bytes are needed at most for the from-charset. */ | |
168 | #ifndef MAX_NEEDED_FROM | |
169 | # define MAX_NEEDED_FROM MIN_NEEDED_FROM | |
170 | #endif | |
171 | ||
172 | /* Same for the to-charset. */ | |
173 | #ifndef MAX_NEEDED_TO | |
174 | # define MAX_NEEDED_TO MIN_NEEDED_TO | |
175 | #endif | |
176 | ||
faaa6f62 UD |
177 | /* Defaults for the per-direction min/max constants. */ |
178 | #ifndef FROM_LOOP_MIN_NEEDED_FROM | |
179 | # define FROM_LOOP_MIN_NEEDED_FROM MIN_NEEDED_FROM | |
180 | #endif | |
181 | #ifndef FROM_LOOP_MAX_NEEDED_FROM | |
182 | # define FROM_LOOP_MAX_NEEDED_FROM MAX_NEEDED_FROM | |
183 | #endif | |
184 | #ifndef FROM_LOOP_MIN_NEEDED_TO | |
185 | # define FROM_LOOP_MIN_NEEDED_TO MIN_NEEDED_TO | |
186 | #endif | |
187 | #ifndef FROM_LOOP_MAX_NEEDED_TO | |
188 | # define FROM_LOOP_MAX_NEEDED_TO MAX_NEEDED_TO | |
189 | #endif | |
190 | #ifndef TO_LOOP_MIN_NEEDED_FROM | |
191 | # define TO_LOOP_MIN_NEEDED_FROM MIN_NEEDED_TO | |
192 | #endif | |
193 | #ifndef TO_LOOP_MAX_NEEDED_FROM | |
194 | # define TO_LOOP_MAX_NEEDED_FROM MAX_NEEDED_TO | |
195 | #endif | |
196 | #ifndef TO_LOOP_MIN_NEEDED_TO | |
197 | # define TO_LOOP_MIN_NEEDED_TO MIN_NEEDED_FROM | |
198 | #endif | |
199 | #ifndef TO_LOOP_MAX_NEEDED_TO | |
200 | # define TO_LOOP_MAX_NEEDED_TO MAX_NEEDED_FROM | |
201 | #endif | |
202 | ||
8619129f | 203 | |
77e1d15a UD |
204 | /* Define macros which can access unaligned buffers. These macros are |
205 | supposed to be used only in code outside the inner loops. For the inner | |
206 | loops we have other definitions which allow optimized access. */ | |
207 | #ifdef _STRING_ARCH_unaligned | |
208 | /* We can handle unaligned memory access. */ | |
a784e502 UD |
209 | # define get16u(addr) *((const uint16_t *) (addr)) |
210 | # define get32u(addr) *((const uint32_t *) (addr)) | |
77e1d15a UD |
211 | |
212 | /* We need no special support for writing values either. */ | |
c1db8b0d UD |
213 | # define put16u(addr, val) *((uint16_t *) (addr)) = (val) |
214 | # define put32u(addr, val) *((uint32_t *) (addr)) = (val) | |
77e1d15a UD |
215 | #else |
216 | /* Distinguish between big endian and little endian. */ | |
217 | # if __BYTE_ORDER == __LITTLE_ENDIAN | |
c1db8b0d | 218 | # define get16u(addr) \ |
a784e502 UD |
219 | (((const unsigned char *) (addr))[1] << 8 \ |
220 | | ((const unsigned char *) (addr))[0]) | |
c1db8b0d | 221 | # define get32u(addr) \ |
a784e502 UD |
222 | (((((const unsigned char *) (addr))[3] << 8 \ |
223 | | ((const unsigned char *) (addr))[2]) << 8 \ | |
224 | | ((const unsigned char *) (addr))[1]) << 8 \ | |
225 | | ((const unsigned char *) (addr))[0]) | |
77e1d15a | 226 | |
c1db8b0d | 227 | # define put16u(addr, val) \ |
77e1d15a | 228 | ({ uint16_t __val = (val); \ |
cb2c5501 UD |
229 | ((unsigned char *) (addr))[0] = __val; \ |
230 | ((unsigned char *) (addr))[1] = __val >> 8; \ | |
77e1d15a | 231 | (void) 0; }) |
c1db8b0d | 232 | # define put32u(addr, val) \ |
cb2c5501 UD |
233 | ({ uint32_t __val = (val); \ |
234 | ((unsigned char *) (addr))[0] = __val; \ | |
77e1d15a | 235 | __val >>= 8; \ |
cb2c5501 | 236 | ((unsigned char *) (addr))[1] = __val; \ |
77e1d15a | 237 | __val >>= 8; \ |
cb2c5501 | 238 | ((unsigned char *) (addr))[2] = __val; \ |
77e1d15a | 239 | __val >>= 8; \ |
cb2c5501 | 240 | ((unsigned char *) (addr))[3] = __val; \ |
77e1d15a UD |
241 | (void) 0; }) |
242 | # else | |
c1db8b0d | 243 | # define get16u(addr) \ |
a784e502 UD |
244 | (((const unsigned char *) (addr))[0] << 8 \ |
245 | | ((const unsigned char *) (addr))[1]) | |
c1db8b0d | 246 | # define get32u(addr) \ |
a784e502 UD |
247 | (((((const unsigned char *) (addr))[0] << 8 \ |
248 | | ((const unsigned char *) (addr))[1]) << 8 \ | |
249 | | ((const unsigned char *) (addr))[2]) << 8 \ | |
250 | | ((const unsigned char *) (addr))[3]) | |
77e1d15a | 251 | |
c1db8b0d | 252 | # define put16u(addr, val) \ |
77e1d15a | 253 | ({ uint16_t __val = (val); \ |
cb2c5501 | 254 | ((unsigned char *) (addr))[1] = __val; \ |
a1303dc8 | 255 | ((unsigned char *) (addr))[0] = __val >> 8; \ |
77e1d15a | 256 | (void) 0; }) |
c1db8b0d | 257 | # define put32u(addr, val) \ |
cb2c5501 UD |
258 | ({ uint32_t __val = (val); \ |
259 | ((unsigned char *) (addr))[3] = __val; \ | |
77e1d15a | 260 | __val >>= 8; \ |
cb2c5501 | 261 | ((unsigned char *) (addr))[2] = __val; \ |
77e1d15a | 262 | __val >>= 8; \ |
cb2c5501 | 263 | ((unsigned char *) (addr))[1] = __val; \ |
77e1d15a | 264 | __val >>= 8; \ |
cb2c5501 | 265 | ((unsigned char *) (addr))[0] = __val; \ |
77e1d15a UD |
266 | (void) 0; }) |
267 | # endif | |
268 | #endif | |
269 | ||
270 | ||
f6ad4726 UD |
271 | /* For conversions from a fixed width character set to another fixed width |
272 | character set we can define RESET_INPUT_BUFFER in a very fast way. */ | |
8619129f | 273 | #if !defined RESET_INPUT_BUFFER && !defined SAVE_RESET_STATE |
faaa6f62 UD |
274 | # if FROM_LOOP_MIN_NEEDED_FROM == FROM_LOOP_MAX_NEEDED_FROM \ |
275 | && FROM_LOOP_MIN_NEEDED_TO == FROM_LOOP_MAX_NEEDED_TO \ | |
276 | && TO_LOOP_MIN_NEEDED_FROM == TO_LOOP_MAX_NEEDED_FROM \ | |
277 | && TO_LOOP_MIN_NEEDED_TO == TO_LOOP_MAX_NEEDED_TO | |
278 | /* We have to use these `if's here since the compiler cannot know that | |
279 | (outbuf - outerr) is always divisible by FROM/TO_LOOP_MIN_NEEDED_TO. | |
280 | The ?:1 avoids division by zero warnings that gcc 3.2 emits even for | |
281 | obviously unreachable code. */ | |
282 | # define RESET_INPUT_BUFFER \ | |
283 | if (FROM_DIRECTION) \ | |
284 | { \ | |
285 | if (FROM_LOOP_MIN_NEEDED_FROM % FROM_LOOP_MIN_NEEDED_TO == 0) \ | |
286 | *inptrp -= (outbuf - outerr) \ | |
287 | * (FROM_LOOP_MIN_NEEDED_FROM / FROM_LOOP_MIN_NEEDED_TO); \ | |
288 | else if (FROM_LOOP_MIN_NEEDED_TO % FROM_LOOP_MIN_NEEDED_FROM == 0) \ | |
289 | *inptrp -= (outbuf - outerr) \ | |
290 | / (FROM_LOOP_MIN_NEEDED_TO / FROM_LOOP_MIN_NEEDED_FROM \ | |
291 | ? : 1); \ | |
292 | else \ | |
293 | *inptrp -= ((outbuf - outerr) / FROM_LOOP_MIN_NEEDED_TO) \ | |
294 | * FROM_LOOP_MIN_NEEDED_FROM; \ | |
295 | } \ | |
296 | else \ | |
297 | { \ | |
298 | if (TO_LOOP_MIN_NEEDED_FROM % TO_LOOP_MIN_NEEDED_TO == 0) \ | |
299 | *inptrp -= (outbuf - outerr) \ | |
300 | * (TO_LOOP_MIN_NEEDED_FROM / TO_LOOP_MIN_NEEDED_TO); \ | |
301 | else if (TO_LOOP_MIN_NEEDED_TO % TO_LOOP_MIN_NEEDED_FROM == 0) \ | |
302 | *inptrp -= (outbuf - outerr) \ | |
303 | / (TO_LOOP_MIN_NEEDED_TO / TO_LOOP_MIN_NEEDED_FROM ? : 1); \ | |
304 | else \ | |
305 | *inptrp -= ((outbuf - outerr) / TO_LOOP_MIN_NEEDED_TO) \ | |
306 | * TO_LOOP_MIN_NEEDED_FROM; \ | |
307 | } | |
8619129f UD |
308 | # endif |
309 | #endif | |
310 | ||
311 | ||
312 | /* The default init function. It simply matches the name and initializes | |
313 | the step data to point to one of the objects above. */ | |
314 | #if DEFINE_INIT | |
315 | # ifndef CHARSET_NAME | |
316 | # error "CHARSET_NAME not defined" | |
317 | # endif | |
318 | ||
a1620a4c | 319 | extern int gconv_init (struct __gconv_step *step); |
8619129f | 320 | int |
d64b6ad0 | 321 | gconv_init (struct __gconv_step *step) |
8619129f UD |
322 | { |
323 | /* Determine which direction. */ | |
d64b6ad0 | 324 | if (strcmp (step->__from_name, CHARSET_NAME) == 0) |
918b9d72 | 325 | { |
45541bdf | 326 | step->__data = FROM_DIRECTION_VAL; |
390955cb | 327 | |
faaa6f62 UD |
328 | step->__min_needed_from = FROM_LOOP_MIN_NEEDED_FROM; |
329 | step->__max_needed_from = FROM_LOOP_MAX_NEEDED_FROM; | |
330 | step->__min_needed_to = FROM_LOOP_MIN_NEEDED_TO; | |
331 | step->__max_needed_to = FROM_LOOP_MAX_NEEDED_TO; | |
f9ad060c UD |
332 | |
333 | #ifdef FROM_ONEBYTE | |
334 | step->__btowc_fct = FROM_ONEBYTE; | |
335 | #endif | |
918b9d72 | 336 | } |
365afefc | 337 | else if (__builtin_expect (strcmp (step->__to_name, CHARSET_NAME), 0) == 0) |
918b9d72 | 338 | { |
45541bdf | 339 | step->__data = TO_DIRECTION_VAL; |
390955cb | 340 | |
faaa6f62 UD |
341 | step->__min_needed_from = TO_LOOP_MIN_NEEDED_FROM; |
342 | step->__max_needed_from = TO_LOOP_MAX_NEEDED_FROM; | |
343 | step->__min_needed_to = TO_LOOP_MIN_NEEDED_TO; | |
344 | step->__max_needed_to = TO_LOOP_MAX_NEEDED_TO; | |
918b9d72 | 345 | } |
390955cb | 346 | else |
d64b6ad0 | 347 | return __GCONV_NOCONV; |
8619129f | 348 | |
f6ad4726 | 349 | #ifdef SAVE_RESET_STATE |
d64b6ad0 | 350 | step->__stateful = 1; |
9ce5071a | 351 | #else |
d64b6ad0 | 352 | step->__stateful = 0; |
9ce5071a UD |
353 | #endif |
354 | ||
d64b6ad0 | 355 | return __GCONV_OK; |
8619129f UD |
356 | } |
357 | #endif | |
358 | ||
359 | ||
360 | /* The default destructor function does nothing in the moment and so | |
f6ad4726 UD |
361 | we don't define it at all. But we still provide the macro just in |
362 | case we need it some day. */ | |
8619129f UD |
363 | #if DEFINE_FINI |
364 | #endif | |
365 | ||
366 | ||
28f1c862 UD |
367 | /* If no arguments have to passed to the loop function define the macro |
368 | as empty. */ | |
369 | #ifndef EXTRA_LOOP_ARGS | |
370 | # define EXTRA_LOOP_ARGS | |
371 | #endif | |
372 | ||
373 | ||
8619129f UD |
374 | /* This is the actual conversion function. */ |
375 | #ifndef FUNCTION_NAME | |
376 | # define FUNCTION_NAME gconv | |
377 | #endif | |
378 | ||
fd1b5c0f UD |
379 | /* The macros are used to access the function to convert single characters. */ |
380 | #define SINGLE(fct) SINGLE2 (fct) | |
381 | #define SINGLE2(fct) fct##_single | |
382 | ||
383 | ||
a1620a4c UD |
384 | extern int FUNCTION_NAME (struct __gconv_step *step, |
385 | struct __gconv_step_data *data, | |
386 | const unsigned char **inptrp, | |
387 | const unsigned char *inend, | |
388 | unsigned char **outbufstart, size_t *irreversible, | |
389 | int do_flush, int consume_incomplete); | |
8619129f | 390 | int |
d64b6ad0 | 391 | FUNCTION_NAME (struct __gconv_step *step, struct __gconv_step_data *data, |
fd1b5c0f | 392 | const unsigned char **inptrp, const unsigned char *inend, |
f1d5c60d | 393 | unsigned char **outbufstart, size_t *irreversible, int do_flush, |
55985355 | 394 | int consume_incomplete) |
8619129f | 395 | { |
d64b6ad0 UD |
396 | struct __gconv_step *next_step = step + 1; |
397 | struct __gconv_step_data *next_data = data + 1; | |
915a6c51 | 398 | __gconv_fct fct = NULL; |
8619129f UD |
399 | int status; |
400 | ||
915a6c51 UD |
401 | if ((data->__flags & __GCONV_IS_LAST) == 0) |
402 | { | |
403 | fct = next_step->__fct; | |
404 | #ifdef PTR_DEMANGLE | |
405 | if (next_step->__shlib_handle != NULL) | |
406 | PTR_DEMANGLE (fct); | |
407 | #endif | |
408 | } | |
85830c4c | 409 | |
8619129f UD |
410 | /* If the function is called with no input this means we have to reset |
411 | to the initial state. The possibly partly converted input is | |
412 | dropped. */ | |
a1ffb40e | 413 | if (__glibc_unlikely (do_flush)) |
8619129f | 414 | { |
f1d5c60d UD |
415 | /* This should never happen during error handling. */ |
416 | assert (outbufstart == NULL); | |
417 | ||
2373b30e UD |
418 | status = __GCONV_OK; |
419 | ||
390955cb | 420 | #ifdef EMIT_SHIFT_TO_INIT |
2373b30e UD |
421 | if (do_flush == 1) |
422 | { | |
423 | /* We preserve the initial values of the pointer variables. */ | |
424 | unsigned char *outbuf = data->__outbuf; | |
425 | unsigned char *outstart = outbuf; | |
426 | unsigned char *outend = data->__outbufend; | |
427 | ||
428 | # ifdef PREPARE_LOOP | |
429 | PREPARE_LOOP | |
430 | # endif | |
431 | ||
432 | # ifdef SAVE_RESET_STATE | |
433 | SAVE_RESET_STATE (1); | |
434 | # endif | |
435 | ||
436 | /* Emit the escape sequence to reset the state. */ | |
437 | EMIT_SHIFT_TO_INIT; | |
438 | ||
439 | /* Call the steps down the chain if there are any but only if we | |
440 | successfully emitted the escape sequence. This should only | |
441 | fail if the output buffer is full. If the input is invalid | |
442 | it should be discarded since the user wants to start from a | |
443 | clean state. */ | |
444 | if (status == __GCONV_OK) | |
445 | { | |
446 | if (data->__flags & __GCONV_IS_LAST) | |
447 | /* Store information about how many bytes are available. */ | |
448 | data->__outbuf = outbuf; | |
449 | else | |
450 | { | |
451 | /* Write out all output which was produced. */ | |
452 | if (outbuf > outstart) | |
453 | { | |
454 | const unsigned char *outerr = outstart; | |
455 | int result; | |
456 | ||
457 | result = DL_CALL_FCT (fct, (next_step, next_data, | |
458 | &outerr, outbuf, NULL, | |
459 | irreversible, 0, | |
460 | consume_incomplete)); | |
461 | ||
462 | if (result != __GCONV_EMPTY_INPUT) | |
463 | { | |
a1ffb40e | 464 | if (__glibc_unlikely (outerr != outbuf)) |
2373b30e UD |
465 | { |
466 | /* We have a problem. Undo the conversion. */ | |
467 | outbuf = outstart; | |
468 | ||
469 | /* Restore the state. */ | |
470 | # ifdef SAVE_RESET_STATE | |
471 | SAVE_RESET_STATE (0); | |
472 | # endif | |
473 | } | |
474 | ||
475 | /* Change the status. */ | |
476 | status = result; | |
477 | } | |
478 | } | |
479 | ||
480 | if (status == __GCONV_OK) | |
481 | /* Now flush the remaining steps. */ | |
482 | status = DL_CALL_FCT (fct, (next_step, next_data, NULL, | |
483 | NULL, NULL, irreversible, 1, | |
484 | consume_incomplete)); | |
485 | } | |
486 | } | |
487 | } | |
488 | else | |
8619129f | 489 | #endif |
2373b30e UD |
490 | { |
491 | /* Clear the state object. There might be bytes in there from | |
492 | previous calls with CONSUME_INCOMPLETE == 1. But don't emit | |
493 | escape sequences. */ | |
494 | memset (data->__statep, '\0', sizeof (*data->__statep)); | |
495 | ||
496 | if (! (data->__flags & __GCONV_IS_LAST)) | |
497 | /* Now flush the remaining steps. */ | |
498 | status = DL_CALL_FCT (fct, (next_step, next_data, NULL, NULL, | |
499 | NULL, irreversible, do_flush, | |
500 | consume_incomplete)); | |
501 | } | |
8619129f UD |
502 | } |
503 | else | |
504 | { | |
8619129f | 505 | /* We preserve the initial values of the pointer variables. */ |
fd1b5c0f | 506 | const unsigned char *inptr = *inptrp; |
f1d5c60d UD |
507 | unsigned char *outbuf = (__builtin_expect (outbufstart == NULL, 1) |
508 | ? data->__outbuf : *outbufstart); | |
d64b6ad0 | 509 | unsigned char *outend = data->__outbufend; |
b117f744 | 510 | unsigned char *outstart; |
3aaad0b6 UD |
511 | /* This variable is used to count the number of characters we |
512 | actually converted. */ | |
38677ace | 513 | size_t lirreversible = 0; |
b572c2da | 514 | size_t *lirreversiblep = irreversible ? &lirreversible : NULL; |
faaa6f62 UD |
515 | |
516 | /* The following assumes that encodings, which have a variable length | |
c0c3f78a | 517 | what might unalign a buffer even though it is an aligned in the |
faaa6f62 UD |
518 | beginning, either don't have the minimal number of bytes as a divisor |
519 | of the maximum length or have a minimum length of 1. This is true | |
520 | for all known and supported encodings. | |
521 | We use && instead of || to combine the subexpression for the FROM | |
522 | encoding and for the TO encoding, because usually one of them is | |
523 | INTERNAL, for which the subexpression evaluates to 1, but INTERNAL | |
524 | buffers are always aligned correctly. */ | |
525 | #define POSSIBLY_UNALIGNED \ | |
526 | (!defined _STRING_ARCH_unaligned \ | |
527 | && (((FROM_LOOP_MIN_NEEDED_FROM != 1 \ | |
528 | && FROM_LOOP_MAX_NEEDED_FROM % FROM_LOOP_MIN_NEEDED_FROM == 0) \ | |
529 | && (FROM_LOOP_MIN_NEEDED_TO != 1 \ | |
530 | && FROM_LOOP_MAX_NEEDED_TO % FROM_LOOP_MIN_NEEDED_TO == 0)) \ | |
531 | || ((TO_LOOP_MIN_NEEDED_FROM != 1 \ | |
532 | && TO_LOOP_MAX_NEEDED_FROM % TO_LOOP_MIN_NEEDED_FROM == 0) \ | |
533 | && (TO_LOOP_MIN_NEEDED_TO != 1 \ | |
534 | && TO_LOOP_MAX_NEEDED_TO % TO_LOOP_MIN_NEEDED_TO == 0)))) | |
535 | #if POSSIBLY_UNALIGNED | |
3aaad0b6 | 536 | int unaligned; |
77e1d15a UD |
537 | # define GEN_unaligned(name) GEN_unaligned2 (name) |
538 | # define GEN_unaligned2(name) name##_unaligned | |
faaa6f62 UD |
539 | #else |
540 | # define unaligned 0 | |
77e1d15a | 541 | #endif |
8619129f | 542 | |
28f1c862 UD |
543 | #ifdef PREPARE_LOOP |
544 | PREPARE_LOOP | |
545 | #endif | |
546 | ||
faaa6f62 | 547 | #if FROM_LOOP_MAX_NEEDED_FROM > 1 || TO_LOOP_MAX_NEEDED_FROM > 1 |
fd1b5c0f UD |
548 | /* If the function is used to implement the mb*towc*() or wc*tomb*() |
549 | functions we must test whether any bytes from the last call are | |
550 | stored in the `state' object. */ | |
faaa6f62 UD |
551 | if (((FROM_LOOP_MAX_NEEDED_FROM > 1 && TO_LOOP_MAX_NEEDED_FROM > 1) |
552 | || (FROM_LOOP_MAX_NEEDED_FROM > 1 && FROM_DIRECTION) | |
553 | || (TO_LOOP_MAX_NEEDED_FROM > 1 && !FROM_DIRECTION)) | |
fd1b5c0f UD |
554 | && consume_incomplete && (data->__statep->__count & 7) != 0) |
555 | { | |
f1d5c60d | 556 | /* Yep, we have some bytes left over. Process them now. |
a784e502 UD |
557 | But this must not happen while we are called from an |
558 | error handler. */ | |
f1d5c60d | 559 | assert (outbufstart == NULL); |
fd1b5c0f | 560 | |
faaa6f62 UD |
561 | # if FROM_LOOP_MAX_NEEDED_FROM > 1 |
562 | if (TO_LOOP_MAX_NEEDED_FROM == 1 || FROM_DIRECTION) | |
55985355 | 563 | status = SINGLE(FROM_LOOP) (step, data, inptrp, inend, &outbuf, |
b572c2da | 564 | outend, lirreversiblep |
85830c4c | 565 | EXTRA_LOOP_ARGS); |
fd1b5c0f | 566 | # endif |
faaa6f62 UD |
567 | # if !ONE_DIRECTION |
568 | # if FROM_LOOP_MAX_NEEDED_FROM > 1 && TO_LOOP_MAX_NEEDED_FROM > 1 | |
fd1b5c0f | 569 | else |
faaa6f62 UD |
570 | # endif |
571 | # if TO_LOOP_MAX_NEEDED_FROM > 1 | |
55985355 | 572 | status = SINGLE(TO_LOOP) (step, data, inptrp, inend, &outbuf, |
b572c2da | 573 | outend, lirreversiblep EXTRA_LOOP_ARGS); |
faaa6f62 | 574 | # endif |
fd1b5c0f UD |
575 | # endif |
576 | ||
365afefc | 577 | if (__builtin_expect (status, __GCONV_OK) != __GCONV_OK) |
fd1b5c0f UD |
578 | return status; |
579 | } | |
580 | #endif | |
581 | ||
faaa6f62 UD |
582 | #if POSSIBLY_UNALIGNED |
583 | unaligned = | |
584 | ((FROM_DIRECTION | |
585 | && ((uintptr_t) inptr % FROM_LOOP_MIN_NEEDED_FROM != 0 | |
586 | || ((data->__flags & __GCONV_IS_LAST) | |
587 | && (uintptr_t) outbuf % FROM_LOOP_MIN_NEEDED_TO != 0))) | |
588 | || (!FROM_DIRECTION | |
589 | && (((data->__flags & __GCONV_IS_LAST) | |
590 | && (uintptr_t) outbuf % TO_LOOP_MIN_NEEDED_TO != 0) | |
591 | || (uintptr_t) inptr % TO_LOOP_MIN_NEEDED_FROM != 0))); | |
3aaad0b6 UD |
592 | #endif |
593 | ||
f1813b56 | 594 | while (1) |
8619129f | 595 | { |
d6204268 UD |
596 | struct __gconv_trans_data *trans; |
597 | ||
8619129f | 598 | /* Remember the start value for this round. */ |
fd1b5c0f | 599 | inptr = *inptrp; |
8619129f | 600 | /* The outbuf buffer is empty. */ |
b117f744 | 601 | outstart = outbuf; |
8619129f | 602 | |
8619129f UD |
603 | #ifdef SAVE_RESET_STATE |
604 | SAVE_RESET_STATE (1); | |
605 | #endif | |
606 | ||
a1ffb40e | 607 | if (__glibc_likely (!unaligned)) |
77e1d15a UD |
608 | { |
609 | if (FROM_DIRECTION) | |
610 | /* Run the conversion loop. */ | |
55985355 | 611 | status = FROM_LOOP (step, data, inptrp, inend, &outbuf, outend, |
b572c2da | 612 | lirreversiblep EXTRA_LOOP_ARGS); |
77e1d15a UD |
613 | else |
614 | /* Run the conversion loop. */ | |
55985355 | 615 | status = TO_LOOP (step, data, inptrp, inend, &outbuf, outend, |
b572c2da | 616 | lirreversiblep EXTRA_LOOP_ARGS); |
77e1d15a | 617 | } |
faaa6f62 | 618 | #if POSSIBLY_UNALIGNED |
8619129f | 619 | else |
77e1d15a UD |
620 | { |
621 | if (FROM_DIRECTION) | |
622 | /* Run the conversion loop. */ | |
55985355 UD |
623 | status = GEN_unaligned (FROM_LOOP) (step, data, inptrp, inend, |
624 | &outbuf, outend, | |
b572c2da | 625 | lirreversiblep |
77e1d15a UD |
626 | EXTRA_LOOP_ARGS); |
627 | else | |
628 | /* Run the conversion loop. */ | |
55985355 UD |
629 | status = GEN_unaligned (TO_LOOP) (step, data, inptrp, inend, |
630 | &outbuf, outend, | |
b572c2da | 631 | lirreversiblep |
77e1d15a UD |
632 | EXTRA_LOOP_ARGS); |
633 | } | |
634 | #endif | |
8619129f | 635 | |
f1d5c60d UD |
636 | /* If we were called as part of an error handling module we |
637 | don't do anything else here. */ | |
a1ffb40e | 638 | if (__glibc_unlikely (outbufstart != NULL)) |
f1d5c60d UD |
639 | { |
640 | *outbufstart = outbuf; | |
641 | return status; | |
642 | } | |
643 | ||
4d37e19c UD |
644 | /* Give the transliteration module the chance to store the |
645 | original text and the result in case it needs a context. */ | |
d6204268 UD |
646 | for (trans = data->__trans; trans != NULL; trans = trans->__next) |
647 | if (trans->__trans_context_fct != NULL) | |
648 | DL_CALL_FCT (trans->__trans_context_fct, | |
649 | (trans->__data, inptr, *inptrp, outstart, outbuf)); | |
4d37e19c | 650 | |
0aece08d UD |
651 | /* We finished one use of the loops. */ |
652 | ++data->__invocation_counter; | |
653 | ||
390955cb | 654 | /* If this is the last step leave the loop, there is nothing |
a784e502 | 655 | we can do. */ |
a1ffb40e | 656 | if (__glibc_unlikely (data->__flags & __GCONV_IS_LAST)) |
8619129f UD |
657 | { |
658 | /* Store information about how many bytes are available. */ | |
d64b6ad0 | 659 | data->__outbuf = outbuf; |
bf979eef | 660 | |
38677ace | 661 | /* Remember how many non-identical characters we |
c0c3f78a | 662 | converted in an irreversible way. */ |
38677ace | 663 | *irreversible += lirreversible; |
bf979eef | 664 | |
8619129f UD |
665 | break; |
666 | } | |
667 | ||
668 | /* Write out all output which was produced. */ | |
a1ffb40e | 669 | if (__glibc_likely (outbuf > outstart)) |
8619129f | 670 | { |
d64b6ad0 | 671 | const unsigned char *outerr = data->__outbuf; |
8619129f UD |
672 | int result; |
673 | ||
4bae5567 | 674 | result = DL_CALL_FCT (fct, (next_step, next_data, &outerr, |
f1d5c60d | 675 | outbuf, NULL, irreversible, 0, |
fd1b5c0f | 676 | consume_incomplete)); |
8619129f | 677 | |
d64b6ad0 | 678 | if (result != __GCONV_EMPTY_INPUT) |
8619129f | 679 | { |
a1ffb40e | 680 | if (__glibc_unlikely (outerr != outbuf)) |
8619129f UD |
681 | { |
682 | #ifdef RESET_INPUT_BUFFER | |
683 | RESET_INPUT_BUFFER; | |
684 | #else | |
dbc92dc6 UD |
685 | /* We have a problem in one of the functions below. |
686 | Undo the conversion upto the error point. */ | |
8619129f UD |
687 | size_t nstatus; |
688 | ||
689 | /* Reload the pointers. */ | |
fd1b5c0f | 690 | *inptrp = inptr; |
b117f744 | 691 | outbuf = outstart; |
8619129f | 692 | |
2373b30e | 693 | /* Restore the state. */ |
8619129f UD |
694 | # ifdef SAVE_RESET_STATE |
695 | SAVE_RESET_STATE (0); | |
696 | # endif | |
697 | ||
a1ffb40e | 698 | if (__glibc_likely (!unaligned)) |
f6ad4726 UD |
699 | { |
700 | if (FROM_DIRECTION) | |
701 | /* Run the conversion loop. */ | |
702 | nstatus = FROM_LOOP (step, data, inptrp, inend, | |
17427edd | 703 | &outbuf, outerr, |
f6ad4726 UD |
704 | lirreversiblep |
705 | EXTRA_LOOP_ARGS); | |
706 | else | |
707 | /* Run the conversion loop. */ | |
708 | nstatus = TO_LOOP (step, data, inptrp, inend, | |
17427edd | 709 | &outbuf, outerr, |
f6ad4726 UD |
710 | lirreversiblep |
711 | EXTRA_LOOP_ARGS); | |
712 | } | |
faaa6f62 | 713 | # if POSSIBLY_UNALIGNED |
8619129f | 714 | else |
f6ad4726 UD |
715 | { |
716 | if (FROM_DIRECTION) | |
717 | /* Run the conversion loop. */ | |
718 | nstatus = GEN_unaligned (FROM_LOOP) (step, data, | |
719 | inptrp, inend, | |
d1700665 | 720 | &outbuf, |
17427edd | 721 | outerr, |
f6ad4726 UD |
722 | lirreversiblep |
723 | EXTRA_LOOP_ARGS); | |
724 | else | |
725 | /* Run the conversion loop. */ | |
726 | nstatus = GEN_unaligned (TO_LOOP) (step, data, | |
727 | inptrp, inend, | |
17427edd | 728 | &outbuf, outerr, |
f6ad4726 UD |
729 | lirreversiblep |
730 | EXTRA_LOOP_ARGS); | |
731 | } | |
732 | # endif | |
8619129f UD |
733 | |
734 | /* We must run out of output buffer space in this | |
735 | rerun. */ | |
5aa8ff62 | 736 | assert (outbuf == outerr); |
d64b6ad0 | 737 | assert (nstatus == __GCONV_FULL_OUTPUT); |
0aece08d UD |
738 | |
739 | /* If we haven't consumed a single byte decrement | |
740 | the invocation counter. */ | |
a1ffb40e | 741 | if (__glibc_unlikely (outbuf == outstart)) |
0aece08d | 742 | --data->__invocation_counter; |
8619129f UD |
743 | #endif /* reset input buffer */ |
744 | } | |
745 | ||
746 | /* Change the status. */ | |
747 | status = result; | |
748 | } | |
749 | else | |
750 | /* All the output is consumed, we can make another run | |
751 | if everything was ok. */ | |
d64b6ad0 | 752 | if (status == __GCONV_FULL_OUTPUT) |
f1813b56 UD |
753 | { |
754 | status = __GCONV_OK; | |
755 | outbuf = data->__outbuf; | |
756 | } | |
8619129f | 757 | } |
f1813b56 UD |
758 | |
759 | if (status != __GCONV_OK) | |
760 | break; | |
761 | ||
762 | /* Reset the output buffer pointer for the next round. */ | |
763 | outbuf = data->__outbuf; | |
8619129f | 764 | } |
8619129f | 765 | |
918b9d72 UD |
766 | #ifdef END_LOOP |
767 | END_LOOP | |
768 | #endif | |
fd1b5c0f UD |
769 | |
770 | /* If we are supposed to consume all character store now all of the | |
771 | remaining characters in the `state' object. */ | |
faaa6f62 UD |
772 | #if FROM_LOOP_MAX_NEEDED_FROM > 1 || TO_LOOP_MAX_NEEDED_FROM > 1 |
773 | if (((FROM_LOOP_MAX_NEEDED_FROM > 1 && TO_LOOP_MAX_NEEDED_FROM > 1) | |
774 | || (FROM_LOOP_MAX_NEEDED_FROM > 1 && FROM_DIRECTION) | |
775 | || (TO_LOOP_MAX_NEEDED_FROM > 1 && !FROM_DIRECTION)) | |
365afefc UD |
776 | && __builtin_expect (consume_incomplete, 0) |
777 | && status == __GCONV_INCOMPLETE_INPUT) | |
fd1b5c0f UD |
778 | { |
779 | # ifdef STORE_REST | |
780 | mbstate_t *state = data->__statep; | |
781 | ||
782 | STORE_REST | |
783 | # else | |
fd1b5c0f | 784 | /* Make sure the remaining bytes fit into the state objects |
a784e502 | 785 | buffer. */ |
fd1b5c0f UD |
786 | assert (inend - *inptrp < 4); |
787 | ||
dbc92dc6 | 788 | size_t cnt; |
fd1b5c0f UD |
789 | for (cnt = 0; *inptrp < inend; ++cnt) |
790 | data->__statep->__value.__wchb[cnt] = *(*inptrp)++; | |
791 | data->__statep->__count &= ~7; | |
792 | data->__statep->__count |= cnt; | |
793 | # endif | |
794 | } | |
795 | #endif | |
faaa6f62 UD |
796 | #undef unaligned |
797 | #undef POSSIBLY_UNALIGNED | |
8619129f UD |
798 | } |
799 | ||
800 | return status; | |
801 | } | |
802 | ||
803 | #undef DEFINE_INIT | |
804 | #undef CHARSET_NAME | |
805 | #undef DEFINE_FINI | |
806 | #undef MIN_NEEDED_FROM | |
807 | #undef MIN_NEEDED_TO | |
808 | #undef MAX_NEEDED_FROM | |
809 | #undef MAX_NEEDED_TO | |
faaa6f62 UD |
810 | #undef FROM_LOOP_MIN_NEEDED_FROM |
811 | #undef FROM_LOOP_MAX_NEEDED_FROM | |
812 | #undef FROM_LOOP_MIN_NEEDED_TO | |
813 | #undef FROM_LOOP_MAX_NEEDED_TO | |
814 | #undef TO_LOOP_MIN_NEEDED_FROM | |
815 | #undef TO_LOOP_MAX_NEEDED_FROM | |
816 | #undef TO_LOOP_MIN_NEEDED_TO | |
817 | #undef TO_LOOP_MAX_NEEDED_TO | |
8619129f UD |
818 | #undef FROM_DIRECTION |
819 | #undef EMIT_SHIFT_TO_INIT | |
820 | #undef FROM_LOOP | |
821 | #undef TO_LOOP | |
f9ad060c | 822 | #undef ONE_DIRECTION |
f6ad4726 | 823 | #undef SAVE_RESET_STATE |
8619129f UD |
824 | #undef RESET_INPUT_BUFFER |
825 | #undef FUNCTION_NAME | |
918b9d72 UD |
826 | #undef PREPARE_LOOP |
827 | #undef END_LOOP | |
f9ad060c | 828 | #undef EXTRA_LOOP_ARGS |
fd1b5c0f | 829 | #undef STORE_REST |
f9ad060c | 830 | #undef FROM_ONEBYTE |