1 /* Skeleton for a conversion module.
2 Copyright (C) 1998, 1999, 2000 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 Library General Public License as
8 published by the Free Software Foundation; either version 2 of the
9 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 Library General Public License for more details.
16 You should have received a copy of the GNU Library General Public
17 License along with the GNU C Library; see the file COPYING.LIB. If not,
18 write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
19 Boston, MA 02111-1307, USA. */
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
25 DEFINE_INIT define the default initializer. This requires the
26 following symbol to be defined.
28 CHARSET_NAME string with official name of the coded character
31 DEFINE_FINI define the default destructor function.
33 MIN_NEEDED_FROM minimal number of bytes needed for the from-charset.
34 MIN_NEEDED_TO likewise for the to-charset.
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.
40 DEFINE_DIRECTION_OBJECTS
41 two objects will be defined to be used when the
42 `gconv' function must only distinguish two
43 directions. This is implied by DEFINE_INIT.
44 If this macro is not defined the following
45 macro must be available.
47 FROM_DIRECTION this macro is supposed to return a value != 0
48 if we convert from the current character set,
49 otherwise it return 0.
51 EMIT_SHIFT_TO_INIT this symbol is optional. If it is defined it
52 defines some code which writes out a sequence
53 of characters which bring the current state into
56 FROM_LOOP name of the function implementing the conversion
57 from the current characters.
58 TO_LOOP likewise for the other direction
60 RESET_STATE in case of an error we must reset the state for
61 the rerun so this macro must be defined for
62 stateful encodings. It takes an argument which
63 is nonzero when saving.
65 RESET_INPUT_BUFFER If the input character sets allow this the macro
66 can be defined to reset the input buffer pointers
67 to cover only those characters up to the error.
69 FUNCTION_NAME if not set the conversion function is named `gconv'.
71 PREPARE_LOOP optional code preparing the conversion loop. Can
72 contain variable definitions.
73 END_LOOP also optional, may be used to store information
75 EXTRA_LOOP_ARGS optional macro specifying extra arguments passed
91 # define DL_CALL_FCT(fct, args) fct args
94 /* The direction objects. */
95 #if DEFINE_DIRECTION_OBJECTS || DEFINE_INIT
96 static int from_object
;
99 # ifndef FROM_DIRECTION
100 # define FROM_DIRECTION (step->__data == &from_object)
103 # ifndef FROM_DIRECTION
104 # error "FROM_DIRECTION must be provided if direction objects are not used"
109 /* How many bytes are needed at most for the from-charset. */
110 #ifndef MAX_NEEDED_FROM
111 # define MAX_NEEDED_FROM MIN_NEEDED_FROM
114 /* Same for the to-charset. */
115 #ifndef MAX_NEEDED_TO
116 # define MAX_NEEDED_TO MIN_NEEDED_TO
120 /* Define macros which can access unaligned buffers. These macros are
121 supposed to be used only in code outside the inner loops. For the inner
122 loops we have other definitions which allow optimized access. */
123 #ifdef _STRING_ARCH_unaligned
124 /* We can handle unaligned memory access. */
125 # define get16u(addr) *((uint16_t *) (addr))
126 # define get32u(addr) *((uint32_t *) (addr))
128 /* We need no special support for writing values either. */
129 # define put16u(addr, val) *((uint16_t *) (addr)) = (val)
130 # define put32u(addr, val) *((uint32_t *) (addr)) = (val)
132 /* Distinguish between big endian and little endian. */
133 # if __BYTE_ORDER == __LITTLE_ENDIAN
134 # define get16u(addr) \
135 (((__const unsigned char *) (addr))[1] << 8 \
136 | ((__const unsigned char *) (addr))[0])
137 # define get32u(addr) \
138 (((((__const unsigned char *) (addr))[3] << 8 \
139 | ((__const unsigned char *) (addr))[2]) << 8 \
140 | ((__const unsigned char *) (addr))[1]) << 8 \
141 | ((__const unsigned char *) (addr))[0])
143 # define put16u(addr, val) \
144 ({ uint16_t __val = (val); \
145 ((unsigned char *) (addr))[0] = __val; \
146 ((unsigned char *) (addr))[1] = __val >> 8; \
148 # define put32u(addr, val) \
149 ({ uint32_t __val = (val); \
150 ((unsigned char *) (addr))[0] = __val; \
152 ((unsigned char *) (addr))[1] = __val; \
154 ((unsigned char *) (addr))[2] = __val; \
156 ((unsigned char *) (addr))[3] = __val; \
159 # define get16u(addr) \
160 (((__const unsigned char *) (addr))[0] << 8 \
161 | ((__const unsigned char *) (addr))[1])
162 # define get32u(addr) \
163 (((((__const unsigned char *) (addr))[0] << 8 \
164 | ((__const unsigned char *) (addr))[1]) << 8 \
165 | ((__const unsigned char *) (addr))[2]) << 8 \
166 | ((__const unsigned char *) (addr))[3])
168 # define put16u(addr, val) \
169 ({ uint16_t __val = (val); \
170 ((unsigned char *) (addr))[1] = __val; \
171 ((unsigned char *) (addr))[0] = __val >> 8; \
173 # define put32u(addr, val) \
174 ({ uint32_t __val = (val); \
175 ((unsigned char *) (addr))[3] = __val; \
177 ((unsigned char *) (addr))[2] = __val; \
179 ((unsigned char *) (addr))[1] = __val; \
181 ((unsigned char *) (addr))[0] = __val; \
187 /* For conversions from a fixed width character sets to another fixed width
188 character set we we can define RESET_INPUT_BUFFER is necessary. */
189 #if !defined RESET_INPUT_BUFFER && !defined SAVE_RESET_STATE
190 # if MIN_NEEDED_FROM == MAX_NEEDED_FROM && MIN_NEEDED_TO == MAX_NEEDED_TO
191 /* We have to use these `if's here since the compiler cannot know that
192 (outbuf - outerr) is always divisible by MIN_NEEDED_TO. */
193 # define RESET_INPUT_BUFFER \
194 if (MIN_NEEDED_FROM % MIN_NEEDED_TO == 0) \
195 *inptrp -= (outbuf - outerr) * (MIN_NEEDED_FROM / MIN_NEEDED_TO); \
196 else if (MIN_NEEDED_TO % MIN_NEEDED_FROM == 0) \
197 *inptrp -= (outbuf - outerr) / (MIN_NEEDED_TO / MIN_NEEDED_FROM); \
199 *inptrp -= ((outbuf - outerr) / MIN_NEEDED_TO) * MIN_NEEDED_FROM
204 /* The default init function. It simply matches the name and initializes
205 the step data to point to one of the objects above. */
207 # ifndef CHARSET_NAME
208 # error "CHARSET_NAME not defined"
212 gconv_init (struct __gconv_step
*step
)
214 /* Determine which direction. */
215 if (strcmp (step
->__from_name
, CHARSET_NAME
) == 0)
217 step
->__data
= &from_object
;
219 step
->__min_needed_from
= MIN_NEEDED_FROM
;
220 step
->__max_needed_from
= MAX_NEEDED_FROM
;
221 step
->__min_needed_to
= MIN_NEEDED_TO
;
222 step
->__max_needed_to
= MAX_NEEDED_TO
;
224 else if (strcmp (step
->__to_name
, CHARSET_NAME
) == 0)
226 step
->__data
= &to_object
;
228 step
->__min_needed_from
= MIN_NEEDED_TO
;
229 step
->__max_needed_from
= MAX_NEEDED_TO
;
230 step
->__min_needed_to
= MIN_NEEDED_FROM
;
231 step
->__max_needed_to
= MAX_NEEDED_FROM
;
234 return __GCONV_NOCONV
;
237 step
->__stateful
= 1;
239 step
->__stateful
= 0;
247 /* The default destructor function does nothing in the moment and so
248 be define it at all. But we still provide the macro just in case
249 we need it some day. */
254 /* If no arguments have to passed to the loop function define the macro
256 #ifndef EXTRA_LOOP_ARGS
257 # define EXTRA_LOOP_ARGS
261 /* This is the actual conversion function. */
262 #ifndef FUNCTION_NAME
263 # define FUNCTION_NAME gconv
266 /* The macros are used to access the function to convert single characters. */
267 #define SINGLE(fct) SINGLE2 (fct)
268 #define SINGLE2(fct) fct##_single
272 FUNCTION_NAME (struct __gconv_step
*step
, struct __gconv_step_data
*data
,
273 const unsigned char **inptrp
, const unsigned char *inend
,
274 size_t *written
, int do_flush
, int consume_incomplete
)
276 struct __gconv_step
*next_step
= step
+ 1;
277 struct __gconv_step_data
*next_data
= data
+ 1;
278 __gconv_fct fct
= data
->__is_last
? NULL
: next_step
->__fct
;
281 /* If the function is called with no input this means we have to reset
282 to the initial state. The possibly partly converted input is
284 if (__builtin_expect (do_flush
, 0))
288 #ifdef EMIT_SHIFT_TO_INIT
289 /* Emit the escape sequence to reset the state. */
292 /* Call the steps down the chain if there are any but only if we
293 successfully emitted the escape sequence. */
294 if (status
== __GCONV_OK
&& ! data
->__is_last
)
295 status
= DL_CALL_FCT (fct
, (next_step
, next_data
, NULL
, NULL
,
296 written
, 1, consume_incomplete
));
300 /* We preserve the initial values of the pointer variables. */
301 const unsigned char *inptr
= *inptrp
;
302 unsigned char *outbuf
= data
->__outbuf
;
303 unsigned char *outend
= data
->__outbufend
;
304 unsigned char *outstart
;
305 /* This variable is used to count the number of characters we
306 actually converted. */
307 size_t converted
= 0;
308 #if defined _STRING_ARCH_unaligned \
309 || MIN_NEEDED_FROM == 1 || MAX_NEEDED_FROM % MIN_NEEDED_FROM != 0 \
310 || MIN_NEEDED_TO == 1 || MAX_NEEDED_TO % MIN_NEEDED_TO != 0
314 # define GEN_unaligned(name) GEN_unaligned2 (name)
315 # define GEN_unaligned2(name) name##_unaligned
322 #if MAX_NEEDED_FROM > 1 || MAX_NEEDED_TO > 1
323 /* If the function is used to implement the mb*towc*() or wc*tomb*()
324 functions we must test whether any bytes from the last call are
325 stored in the `state' object. */
326 if (((MAX_NEEDED_FROM
> 1 && FROM_DIRECTION
)
327 || (MAX_NEEDED_TO
> 1 && !FROM_DIRECTION
))
328 && consume_incomplete
&& (data
->__statep
->__count
& 7) != 0)
330 /* Yep, we have some bytes left over. Process them now. */
332 # if MAX_NEEDED_FROM > 1
333 if (MAX_NEEDED_TO
== 1 || FROM_DIRECTION
)
334 status
= SINGLE(FROM_LOOP
) (inptrp
, inend
, &outbuf
, outend
,
335 data
->__statep
, step
->__data
,
336 &converted EXTRA_LOOP_ARGS
);
338 # if MAX_NEEDED_FROM > 1 && MAX_NEEDED_TO > 1 && !ONE_DIRECTION
341 # if MAX_NEEDED_TO > 1 && !ONE_DIRECTION
342 status
= SINGLE(TO_LOOP
) (inptrp
, inend
, &outbuf
, outend
,
343 data
->__statep
, step
->__data
,
344 &converted EXTRA_LOOP_ARGS
);
347 if (status
!= __GCONV_OK
)
352 #if !defined _STRING_ARCH_unaligned \
353 && MIN_NEEDED_FROM != 1 && MAX_NEEDED_FROM % MIN_NEEDED_FROM == 0 \
354 && MIN_NEEDED_TO != 1 && MAX_NEEDED_TO % MIN_NEEDED_TO == 0
355 /* The following assumes that encodings, which have a variable length
356 what might unalign a buffer even though it is a aligned in the
357 beginning, either don't have the minimal number of bytes as a divisor
358 of the maximum length or have a minimum length of 1. This is true
359 for all known and supported encodings. */
360 unaligned
= ((FROM_DIRECTION
361 && ((uintptr_t) inptr
% MIN_NEEDED_FROM
!= 0
363 && (uintptr_t) outbuf
% MIN_NEEDED_TO
!= 0)))
366 && (uintptr_t) outbuf
% MIN_NEEDED_FROM
!= 0)
367 || (uintptr_t) inptr
% MIN_NEEDED_TO
!= 0)));
372 /* Remember the start value for this round. */
374 /* The outbuf buffer is empty. */
377 #ifdef SAVE_RESET_STATE
378 SAVE_RESET_STATE (1);
384 /* Run the conversion loop. */
385 status
= FROM_LOOP (inptrp
, inend
, &outbuf
, outend
,
386 data
->__statep
, step
->__data
, &converted
389 /* Run the conversion loop. */
390 status
= TO_LOOP (inptrp
, inend
, &outbuf
, outend
,
391 data
->__statep
, step
->__data
, &converted
394 #if !defined _STRING_ARCH_unaligned \
395 && MIN_NEEDED_FROM != 1 && MAX_NEEDED_FROM % MIN_NEEDED_FROM == 0 \
396 && MIN_NEEDED_TO != 1 && MAX_NEEDED_TO % MIN_NEEDED_TO == 0
400 /* Run the conversion loop. */
401 status
= GEN_unaligned (FROM_LOOP
) (inptrp
, inend
, &outbuf
,
402 outend
, data
->__statep
,
403 step
->__data
, &converted
406 /* Run the conversion loop. */
407 status
= GEN_unaligned (TO_LOOP
) (inptrp
, inend
, &outbuf
,
408 outend
, data
->__statep
,
409 step
->__data
, &converted
414 /* We finished one use of the loops. */
415 ++data
->__invocation_counter
;
417 /* If this is the last step leave the loop, there is nothing
421 /* Store information about how many bytes are available. */
422 data
->__outbuf
= outbuf
;
424 /* Remember how many non-identical characters we converted. */
425 *written
+= converted
;
430 /* Write out all output which was produced. */
431 if (outbuf
> outstart
)
433 const unsigned char *outerr
= data
->__outbuf
;
436 result
= DL_CALL_FCT (fct
, (next_step
, next_data
, &outerr
,
438 consume_incomplete
));
440 if (result
!= __GCONV_EMPTY_INPUT
)
442 if (__builtin_expect (outerr
!= outbuf
, 0))
444 #ifdef RESET_INPUT_BUFFER
447 /* We have a problem with the in on of the functions
448 below. Undo the conversion upto the error point. */
451 /* Reload the pointers. */
455 /* Reset the state. */
456 # ifdef SAVE_RESET_STATE
457 SAVE_RESET_STATE (0);
461 /* Run the conversion loop. */
462 nstatus
= FROM_LOOP ((const unsigned char **) inptrp
,
463 (const unsigned char *) inend
,
464 (unsigned char **) &outbuf
,
465 (unsigned char *) outerr
,
466 data
->__statep
, step
->__data
,
467 &converted EXTRA_LOOP_ARGS
);
469 /* Run the conversion loop. */
470 nstatus
= TO_LOOP ((const unsigned char **) inptrp
,
471 (const unsigned char *) inend
,
472 (unsigned char **) &outbuf
,
473 (unsigned char *) outerr
,
474 data
->__statep
, step
->__data
,
475 &converted EXTRA_LOOP_ARGS
);
477 /* We must run out of output buffer space in this
479 assert (outbuf
== outerr
);
480 assert (nstatus
== __GCONV_FULL_OUTPUT
);
482 /* If we haven't consumed a single byte decrement
483 the invocation counter. */
484 if (outbuf
== outstart
)
485 --data
->__invocation_counter
;
486 #endif /* reset input buffer */
489 /* Change the status. */
493 /* All the output is consumed, we can make another run
494 if everything was ok. */
495 if (status
== __GCONV_FULL_OUTPUT
)
499 while (status
== __GCONV_OK
);
505 /* If we are supposed to consume all character store now all of the
506 remaining characters in the `state' object. */
507 #if MAX_NEEDED_FROM > 1 || MAX_NEEDED_TO > 1
508 if (((MAX_NEEDED_FROM
> 1 && FROM_DIRECTION
)
509 || (MAX_NEEDED_TO
> 1 && !FROM_DIRECTION
))
510 && consume_incomplete
&& status
== __GCONV_INCOMPLETE_INPUT
)
513 mbstate_t *state
= data
->__statep
;
519 /* Make sure the remaining bytes fit into the state objects
521 assert (inend
- *inptrp
< 4);
523 for (cnt
= 0; *inptrp
< inend
; ++cnt
)
524 data
->__statep
->__value
.__wchb
[cnt
] = *(*inptrp
)++;
525 data
->__statep
->__count
&= ~7;
526 data
->__statep
->__count
|= cnt
;
538 #undef MIN_NEEDED_FROM
540 #undef MAX_NEEDED_FROM
542 #undef DEFINE_DIRECTION_OBJECTS
543 #undef FROM_DIRECTION
544 #undef EMIT_SHIFT_TO_INIT
548 #undef RESET_INPUT_BUFFER