1 /* Conversion loop frame work.
2 Copyright (C) 1998-2002, 2003, 2005 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
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))
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) \
189 result = __GCONV_ILLEGAL_INPUT; \
191 if (! ignore_errors_p ()) \
194 /* We ignore the invalid input byte sequence. */ \
197 /* But we keep result == __GCONV_ILLEGAL_INPUT, because of the constraint \
198 that "iconv -c" must give the same exitcode as "iconv". */ \
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
206 #define STANDARD_TO_LOOP_ERR_HANDLER(Incr) \
208 struct __gconv_trans_data *trans; \
210 result = __GCONV_ILLEGAL_INPUT; \
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. */ \
217 /* First try the transliteration methods. */ \
218 for (trans = step_data->__trans; trans != NULL; trans = trans->__next) \
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) \
226 /* If any of them recognized the input continue with the loop. */ \
227 if (result != __GCONV_ILLEGAL_INPUT) \
229 if (__builtin_expect (result == __GCONV_FULL_OUTPUT, 0)) \
235 /* Next see whether we have to ignore the error. If not, stop. */ \
236 if (! ignore_errors_p ()) \
239 /* When we come here it means we ignore the character. */ \
242 /* But we keep result == __GCONV_ILLEGAL_INPUT, because of the constraint \
243 that "iconv -c" must give the same exitcode as "iconv". */ \
248 /* Handling of Unicode 3.1 TAG characters. Unicode recommends
249 "If language codes are not relevant to the particular processing
250 operation, then they should be ignored." This macro is usually
251 called right before STANDARD_TO_LOOP_ERR_HANDLER (Incr). */
252 #define UNICODE_TAG_HANDLER(Character, Incr) \
254 /* TAG characters are those in the range U+E0000..U+E007F. */ \
255 if (((Character) >> 7) == (0xe0000 >> 7)) \
263 /* The function returns the status, as defined in gconv.h. */
265 __attribute ((always_inline
))
266 FCTNAME (LOOPFCT
) (struct __gconv_step
*step
,
267 struct __gconv_step_data
*step_data
,
268 const unsigned char **inptrp
, const unsigned char *inend
,
269 unsigned char **outptrp
, const unsigned char *outend
,
270 size_t *irreversible EXTRA_LOOP_DECLS
)
272 #ifdef LOOP_NEED_STATE
273 mbstate_t *state
= step_data
->__statep
;
275 #ifdef LOOP_NEED_FLAGS
276 int flags
= step_data
->__flags
;
278 #ifdef LOOP_NEED_DATA
279 void *data
= step
->__data
;
281 int result
= __GCONV_EMPTY_INPUT
;
282 const unsigned char *inptr
= *inptrp
;
283 unsigned char *outptr
= *outptrp
;
289 while (inptr
!= inend
)
291 /* `if' cases for MIN_NEEDED_OUTPUT ==/!= 1 is made to help the
292 compiler generating better code. They will be optimized away
293 since MIN_NEEDED_OUTPUT is always a constant. */
294 if (MIN_NEEDED_INPUT
> 1
295 && __builtin_expect (inptr
+ MIN_NEEDED_INPUT
> inend
, 0))
297 /* We don't have enough input for another complete input
299 result
= __GCONV_INCOMPLETE_INPUT
;
302 if ((MIN_NEEDED_OUTPUT
!= 1
303 && __builtin_expect (outptr
+ MIN_NEEDED_OUTPUT
> outend
, 0))
304 || (MIN_NEEDED_OUTPUT
== 1
305 && __builtin_expect (outptr
>= outend
, 0)))
307 /* Overflow in the output buffer. */
308 result
= __GCONV_FULL_OUTPUT
;
312 /* Here comes the body the user provides. It can stop with
313 RESULT set to GCONV_INCOMPLETE_INPUT (if the size of the
314 input characters vary in size), GCONV_ILLEGAL_INPUT, or
315 GCONV_FULL_OUTPUT (if the output characters vary in size). */
319 /* Update the pointers pointed to by the parameters. */
330 /* Include the file a second time to define the function to handle
332 #if !defined DEFINE_UNALIGNED && !defined _STRING_ARCH_unaligned \
333 && MIN_NEEDED_INPUT != 1 && MAX_NEEDED_INPUT % MIN_NEEDED_INPUT == 0 \
334 && MIN_NEEDED_OUTPUT != 1 && MAX_NEEDED_OUTPUT % MIN_NEEDED_OUTPUT == 0
341 # define DEFINE_UNALIGNED
343 # undef DEFINE_UNALIGNED
347 #if MAX_NEEDED_INPUT > 1
348 # define SINGLE(fct) SINGLE2 (fct)
349 # define SINGLE2(fct) fct##_single
351 __attribute ((always_inline
))
352 SINGLE(LOOPFCT
) (struct __gconv_step
*step
,
353 struct __gconv_step_data
*step_data
,
354 const unsigned char **inptrp
, const unsigned char *inend
,
355 unsigned char **outptrp
, unsigned char *outend
,
356 size_t *irreversible EXTRA_LOOP_DECLS
)
358 mbstate_t *state
= step_data
->__statep
;
359 #ifdef LOOP_NEED_FLAGS
360 int flags
= step_data
->__flags
;
362 #ifdef LOOP_NEED_DATA
363 void *data
= step
->__data
;
365 int result
= __GCONV_OK
;
366 unsigned char bytebuf
[MAX_NEEDED_INPUT
];
367 const unsigned char *inptr
= *inptrp
;
368 unsigned char *outptr
= *outptrp
;
378 /* Add the bytes from the state to the input buffer. */
379 for (inlen
= 0; inlen
< (size_t) (state
->__count
& 7); ++inlen
)
380 bytebuf
[inlen
] = state
->__value
.__wchb
[inlen
];
383 /* Are there enough bytes in the input buffer? */
384 if (__builtin_expect (inptr
+ (MIN_NEEDED_INPUT
- inlen
) > inend
, 0))
390 inend
= &bytebuf
[inlen
];
394 /* We don't have enough input for another complete input
396 while (inptr
< inend
)
397 state
->__value
.__wchb
[inlen
++] = *inptr
++;
400 return __GCONV_INCOMPLETE_INPUT
;
403 /* Enough space in output buffer. */
404 if ((MIN_NEEDED_OUTPUT
!= 1 && outptr
+ MIN_NEEDED_OUTPUT
> outend
)
405 || (MIN_NEEDED_OUTPUT
== 1 && outptr
>= outend
))
406 /* Overflow in the output buffer. */
407 return __GCONV_FULL_OUTPUT
;
409 /* Now add characters from the normal input buffer. */
411 bytebuf
[inlen
++] = *inptr
++;
412 while (inlen
< MAX_NEEDED_INPUT
&& inptr
< inend
);
415 inend
= &bytebuf
[inlen
];
423 /* Now we either have produced an output character and consumed all the
424 bytes from the state and at least one more, or the character is still
425 incomplete, or we have some other error (like illegal input character,
426 no space in output buffer). */
427 if (__builtin_expect (inptr
!= bytebuf
, 1))
429 /* We found a new character. */
430 assert (inptr
- bytebuf
> (state
->__count
& 7));
432 *inptrp
+= inptr
- bytebuf
- (state
->__count
& 7);
437 /* Clear the state buffer. */
441 state
->__count
&= ~7;
444 else if (result
== __GCONV_INCOMPLETE_INPUT
)
446 /* This can only happen if we have less than MAX_NEEDED_INPUT bytes
448 assert (inend
!= &bytebuf
[MAX_NEEDED_INPUT
]);
450 *inptrp
+= inend
- bytebuf
- (state
->__count
& 7);
456 /* We don't have enough input for another complete input
458 assert (inend
- inptr
> (state
->__count
& ~7));
459 assert (inend
- inptr
<= 7);
460 state
->__count
= (state
->__count
& ~7) | (inend
- inptr
);
462 while (inptr
< inend
)
463 state
->__value
.__wchb
[inlen
++] = *inptr
++;
475 /* Define the shortcut function for btowc. */
477 gconv_btowc (struct __gconv_step
*step
, unsigned char c
)
479 # define FROM_ONEBYTE gconv_btowc
483 /* We remove the macro definitions so that we can include this file again
484 for the definition of another function. */
485 #undef MIN_NEEDED_INPUT
486 #undef MAX_NEEDED_INPUT
487 #undef MIN_NEEDED_OUTPUT
488 #undef MAX_NEEDED_OUTPUT
492 #undef EXTRA_LOOP_DECLS
498 #undef LOOP_NEED_STATE
499 #undef LOOP_NEED_FLAGS
500 #undef LOOP_NEED_DATA