1 /* Conversion loop frame work.
2 Copyright (C) 1998-2003, 2005, 2008, 2011 Free Software Foundation, Inc.
3 This file is part of the GNU C Library.
4 Contributed by Ulrich Drepper <drepper@cygnus.com>, 1998.
6 The GNU C Library is free software; you can redistribute it and/or
7 modify it under the terms of the GNU Lesser General Public
8 License as published by the Free Software Foundation; either
9 version 2.1 of the License, or (at your option) any later version.
11 The GNU C Library is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 Lesser General Public License for more details.
16 You should have received a copy of the GNU Lesser General Public
17 License along with the GNU C Library; if not, write to the Free
18 Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
21 /* This file provides a frame for the reader loop in all conversion modules.
22 The actual code must (of course) be provided in the actual module source
23 code but certain actions can be written down generically, with some
24 customization options which are these:
26 MIN_NEEDED_INPUT minimal number of input bytes needed for the next
28 MIN_NEEDED_OUTPUT minimal number of bytes produced by the next round
31 MAX_NEEDED_INPUT you guess it, this is the maximal number of input
32 bytes needed. It defaults to MIN_NEEDED_INPUT
33 MAX_NEEDED_OUTPUT likewise for output bytes.
35 LOOPFCT name of the function created. If not specified
36 the name is `loop' but this prevents the use
37 of multiple functions in the same file.
39 BODY this is supposed to expand to the body of the loop.
40 The user must provide this.
42 EXTRA_LOOP_DECLS extra arguments passed from converion loop call.
44 INIT_PARAMS code to define and initialize variables from params.
45 UPDATE_PARAMS code to store result in params.
47 ONEBYTE_BODY body of the specialized conversion function for a
48 single byte from the current character set to INTERNAL.
57 #include <sys/param.h> /* For MIN. */
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. */
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))
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)
76 # define FCTNAME2(name) name
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])
89 # define put16(addr, val) \
90 ({ uint16_t __val = (val); \
91 ((unsigned char *) (addr))[0] = __val; \
92 ((unsigned char *) (addr))[1] = __val >> 8; \
94 # define put32(addr, val) \
95 ({ uint32_t __val = (val); \
96 ((unsigned char *) (addr))[0] = __val; \
98 ((unsigned char *) (addr))[1] = __val; \
100 ((unsigned char *) (addr))[2] = __val; \
102 ((unsigned char *) (addr))[3] = __val; \
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])
114 # define put16(addr, val) \
115 ({ uint16_t __val = (val); \
116 ((unsigned char *) (addr))[1] = __val; \
117 ((unsigned char *) (addr))[0] = __val >> 8; \
119 # define put32(addr, val) \
120 ({ uint32_t __val = (val); \
121 ((unsigned char *) (addr))[3] = __val; \
123 ((unsigned char *) (addr))[2] = __val; \
125 ((unsigned char *) (addr))[1] = __val; \
127 ((unsigned char *) (addr))[0] = __val; \
131 # define FCTNAME2(name) name##_unaligned
133 #define FCTNAME(name) FCTNAME2(name)
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"
143 /* Let's see how many bytes we produce. */
144 #ifndef MAX_NEEDED_INPUT
145 # define MAX_NEEDED_INPUT MIN_NEEDED_INPUT
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"
155 /* Let's see how many bytes we produce. */
156 #ifndef MAX_NEEDED_OUTPUT
157 # define MAX_NEEDED_OUTPUT MIN_NEEDED_OUTPUT
160 /* Default name for the function. */
162 # define LOOPFCT loop
165 /* Make sure we have a loop body. */
167 # error "Definition of BODY missing for function" LOOPFCT
171 /* If no arguments have to passed to the loop function define the macro
173 #ifndef EXTRA_LOOP_DECLS
174 # define EXTRA_LOOP_DECLS
177 /* Allow using UPDATE_PARAMS in macros where #ifdef UPDATE_PARAMS test
179 #ifndef UPDATE_PARAMS
180 # define UPDATE_PARAMS do { } while (0)
182 #ifndef REINIT_PARAMS
183 # define REINIT_PARAMS do { } while (0)
187 /* To make it easier for the writers of the modules, we define a macro
188 to test whether we have to ignore errors. */
189 #define ignore_errors_p() \
190 (irreversible != NULL && (flags & __GCONV_IGNORE_ERRORS))
193 /* Error handling for the FROM_LOOP direction, with ignoring of errors.
194 Note that we cannot use the do while (0) trick since `break' and
195 `continue' must reach certain points. */
196 #define STANDARD_FROM_LOOP_ERR_HANDLER(Incr) \
198 result = __GCONV_ILLEGAL_INPUT; \
200 if (! ignore_errors_p ()) \
203 /* We ignore the invalid input byte sequence. */ \
206 /* But we keep result == __GCONV_ILLEGAL_INPUT, because of the constraint \
207 that "iconv -c" must give the same exitcode as "iconv". */ \
211 /* Error handling for the TO_LOOP direction, with use of transliteration/
212 transcription functions and ignoring of errors. Note that we cannot use
213 the do while (0) trick since `break' and `continue' must reach certain
215 #define STANDARD_TO_LOOP_ERR_HANDLER(Incr) \
217 struct __gconv_trans_data *trans; \
219 result = __GCONV_ILLEGAL_INPUT; \
221 if (irreversible == NULL) \
222 /* This means we are in call from __gconv_transliterate. In this \
223 case we are not doing any error recovery outself. */ \
226 /* If needed, flush any conversion state, so that __gconv_transliterate \
227 starts with current shift state. */ \
230 /* First try the transliteration methods. */ \
231 for (trans = step_data->__trans; trans != NULL; trans = trans->__next) \
233 result = DL_CALL_FCT (trans->__trans_fct, \
234 (step, step_data, trans->__data, *inptrp, \
235 &inptr, inend, &outptr, irreversible)); \
236 if (result != __GCONV_ILLEGAL_INPUT) \
242 /* If any of them recognized the input continue with the loop. */ \
243 if (result != __GCONV_ILLEGAL_INPUT) \
245 if (__builtin_expect (result == __GCONV_FULL_OUTPUT, 0)) \
251 /* Next see whether we have to ignore the error. If not, stop. */ \
252 if (! ignore_errors_p ()) \
255 /* When we come here it means we ignore the character. */ \
258 /* But we keep result == __GCONV_ILLEGAL_INPUT, because of the constraint \
259 that "iconv -c" must give the same exitcode as "iconv". */ \
264 /* Handling of Unicode 3.1 TAG characters. Unicode recommends
265 "If language codes are not relevant to the particular processing
266 operation, then they should be ignored." This macro is usually
267 called right before STANDARD_TO_LOOP_ERR_HANDLER (Incr). */
268 #define UNICODE_TAG_HANDLER(Character, Incr) \
270 /* TAG characters are those in the range U+E0000..U+E007F. */ \
271 if (((Character) >> 7) == (0xe0000 >> 7)) \
279 /* The function returns the status, as defined in gconv.h. */
281 __attribute ((always_inline
))
282 FCTNAME (LOOPFCT
) (struct __gconv_step
*step
,
283 struct __gconv_step_data
*step_data
,
284 const unsigned char **inptrp
, const unsigned char *inend
,
285 unsigned char **outptrp
, const unsigned char *outend
,
286 size_t *irreversible EXTRA_LOOP_DECLS
)
288 #ifdef LOOP_NEED_STATE
289 mbstate_t *state
= step_data
->__statep
;
291 #ifdef LOOP_NEED_FLAGS
292 int flags
= step_data
->__flags
;
294 #ifdef LOOP_NEED_DATA
295 void *data
= step
->__data
;
297 int result
= __GCONV_EMPTY_INPUT
;
298 const unsigned char *inptr
= *inptrp
;
299 unsigned char *outptr
= *outptrp
;
305 while (inptr
!= inend
)
307 /* `if' cases for MIN_NEEDED_OUTPUT ==/!= 1 is made to help the
308 compiler generating better code. They will be optimized away
309 since MIN_NEEDED_OUTPUT is always a constant. */
310 if (MIN_NEEDED_INPUT
> 1
311 && __builtin_expect (inptr
+ MIN_NEEDED_INPUT
> inend
, 0))
313 /* We don't have enough input for another complete input
315 result
= __GCONV_INCOMPLETE_INPUT
;
318 if ((MIN_NEEDED_OUTPUT
!= 1
319 && __builtin_expect (outptr
+ MIN_NEEDED_OUTPUT
> outend
, 0))
320 || (MIN_NEEDED_OUTPUT
== 1
321 && __builtin_expect (outptr
>= outend
, 0)))
323 /* Overflow in the output buffer. */
324 result
= __GCONV_FULL_OUTPUT
;
328 /* Here comes the body the user provides. It can stop with
329 RESULT set to GCONV_INCOMPLETE_INPUT (if the size of the
330 input characters vary in size), GCONV_ILLEGAL_INPUT, or
331 GCONV_FULL_OUTPUT (if the output characters vary in size). */
335 /* Update the pointers pointed to by the parameters. */
344 /* Include the file a second time to define the function to handle
346 #if !defined DEFINE_UNALIGNED && !defined _STRING_ARCH_unaligned \
347 && MIN_NEEDED_INPUT != 1 && MAX_NEEDED_INPUT % MIN_NEEDED_INPUT == 0 \
348 && MIN_NEEDED_OUTPUT != 1 && MAX_NEEDED_OUTPUT % MIN_NEEDED_OUTPUT == 0
355 # define DEFINE_UNALIGNED
357 # undef DEFINE_UNALIGNED
361 #if MAX_NEEDED_INPUT > 1
362 # define SINGLE(fct) SINGLE2 (fct)
363 # define SINGLE2(fct) fct##_single
365 __attribute ((always_inline
))
366 SINGLE(LOOPFCT
) (struct __gconv_step
*step
,
367 struct __gconv_step_data
*step_data
,
368 const unsigned char **inptrp
, const unsigned char *inend
,
369 unsigned char **outptrp
, unsigned char *outend
,
370 size_t *irreversible EXTRA_LOOP_DECLS
)
372 mbstate_t *state
= step_data
->__statep
;
373 #ifdef LOOP_NEED_FLAGS
374 int flags
= step_data
->__flags
;
376 #ifdef LOOP_NEED_DATA
377 void *data
= step
->__data
;
379 int result
= __GCONV_OK
;
380 unsigned char bytebuf
[MAX_NEEDED_INPUT
];
381 const unsigned char *inptr
= *inptrp
;
382 unsigned char *outptr
= *outptrp
;
392 /* Add the bytes from the state to the input buffer. */
393 for (inlen
= 0; inlen
< (size_t) (state
->__count
& 7); ++inlen
)
394 bytebuf
[inlen
] = state
->__value
.__wchb
[inlen
];
397 /* Are there enough bytes in the input buffer? */
398 if (MIN_NEEDED_INPUT
> 1
399 && __builtin_expect (inptr
+ (MIN_NEEDED_INPUT
- inlen
) > inend
, 0))
403 while (inptr
< inend
)
404 bytebuf
[inlen
++] = *inptr
++;
408 inend
= &bytebuf
[inlen
];
412 /* We don't have enough input for another complete input
414 while (inptr
< inend
)
415 state
->__value
.__wchb
[inlen
++] = *inptr
++;
418 return __GCONV_INCOMPLETE_INPUT
;
421 /* Enough space in output buffer. */
422 if ((MIN_NEEDED_OUTPUT
!= 1 && outptr
+ MIN_NEEDED_OUTPUT
> outend
)
423 || (MIN_NEEDED_OUTPUT
== 1 && outptr
>= outend
))
424 /* Overflow in the output buffer. */
425 return __GCONV_FULL_OUTPUT
;
427 /* Now add characters from the normal input buffer. */
429 bytebuf
[inlen
++] = *inptr
++;
430 while (inlen
< MAX_NEEDED_INPUT
&& inptr
< inend
);
433 inend
= &bytebuf
[inlen
];
441 /* Now we either have produced an output character and consumed all the
442 bytes from the state and at least one more, or the character is still
443 incomplete, or we have some other error (like illegal input character,
444 no space in output buffer). */
445 if (__builtin_expect (inptr
!= bytebuf
, 1))
447 /* We found a new character. */
448 assert (inptr
- bytebuf
> (state
->__count
& 7));
450 *inptrp
+= inptr
- bytebuf
- (state
->__count
& 7);
455 /* Clear the state buffer. */
459 state
->__count
&= ~7;
462 else if (result
== __GCONV_INCOMPLETE_INPUT
)
464 /* This can only happen if we have less than MAX_NEEDED_INPUT bytes
466 assert (inend
!= &bytebuf
[MAX_NEEDED_INPUT
]);
468 *inptrp
+= inend
- bytebuf
- (state
->__count
& 7);
474 /* We don't have enough input for another complete input
476 assert (inend
- inptr
> (state
->__count
& ~7));
477 assert (inend
- inptr
<= 7);
478 state
->__count
= (state
->__count
& ~7) | (inend
- inptr
);
480 while (inptr
< inend
)
481 state
->__value
.__wchb
[inlen
++] = *inptr
++;
493 /* Define the shortcut function for btowc. */
495 gconv_btowc (struct __gconv_step
*step
, unsigned char c
)
497 # define FROM_ONEBYTE gconv_btowc
501 /* We remove the macro definitions so that we can include this file again
502 for the definition of another function. */
503 #undef MIN_NEEDED_INPUT
504 #undef MAX_NEEDED_INPUT
505 #undef MIN_NEEDED_OUTPUT
506 #undef MAX_NEEDED_OUTPUT
510 #undef EXTRA_LOOP_DECLS
517 #undef LOOP_NEED_STATE
518 #undef LOOP_NEED_FLAGS
519 #undef LOOP_NEED_DATA