1 /* Conversion loop frame work.
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 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.
54 #include <sys/param.h> /* For MIN. */
59 /* We have to provide support for machines which are not able to handled
60 unaligned memory accesses. Some of the character encodings have
61 representations with a fixed width of 2 or 4 bytes. But if we cannot
62 access unaligned memory we still have to read byte-wise. */
64 #if defined _STRING_ARCH_unaligned || !defined DEFINE_UNALIGNED
65 /* We can handle unaligned memory access. */
66 # define get16(addr) *((uint16_t *) (addr))
67 # define get32(addr) *((uint32_t *) (addr))
69 /* We need no special support for writing values either. */
70 # define put16(addr, val) *((uint16_t *) (addr)) = (val)
71 # define put32(addr, val) *((uint32_t *) (addr)) = (val)
73 # define FCTNAME2(name) name
75 /* Distinguish between big endian and little endian. */
76 # if __BYTE_ORDER == __LITTLE_ENDIAN
77 # define get16(addr) \
78 (((__const unsigned char *) (addr))[1] << 8 \
79 | ((__const unsigned char *) (addr))[0])
80 # define get32(addr) \
81 (((((__const unsigned char *) (addr))[3] << 8 \
82 | ((__const unsigned char *) (addr))[2]) << 8 \
83 | ((__const unsigned char *) (addr))[1]) << 8 \
84 | ((__const unsigned char *) (addr))[0])
86 # define put16(addr, val) \
87 ({ uint16_t __val = (val); \
88 ((unsigned char *) (addr))[0] = __val; \
89 ((unsigned char *) (addr))[1] = __val >> 8; \
91 # define put32(addr, val) \
92 ({ uint32_t __val = (val); \
93 ((unsigned char *) (addr))[0] = __val; \
95 ((unsigned char *) (addr))[1] = __val; \
97 ((unsigned char *) (addr))[2] = __val; \
99 ((unsigned char *) (addr))[3] = __val; \
102 # define get16(addr) \
103 (((__const unsigned char *) (addr))[0] << 8 \
104 | ((__const unsigned char *) (addr))[1])
105 # define get32(addr) \
106 (((((__const unsigned char *) (addr))[0] << 8 \
107 | ((__const unsigned char *) (addr))[1]) << 8 \
108 | ((__const unsigned char *) (addr))[2]) << 8 \
109 | ((__const unsigned char *) (addr))[3])
111 # define put16(addr, val) \
112 ({ uint16_t __val = (val); \
113 ((unsigned char *) (addr))[1] = __val; \
114 ((unsigned char *) (addr))[2] = __val >> 8; \
116 # define put32(addr, val) \
117 ({ uint32_t __val = (val); \
118 ((unsigned char *) (addr))[3] = __val; \
120 ((unsigned char *) (addr))[2] = __val; \
122 ((unsigned char *) (addr))[1] = __val; \
124 ((unsigned char *) (addr))[0] = __val; \
128 # define FCTNAME2(name) name##_unaligned
130 #define FCTNAME(name) FCTNAME2(name)
133 /* We need at least one byte for the next round. */
134 #ifndef MIN_NEEDED_INPUT
135 # error "MIN_NEEDED_INPUT definition missing"
138 /* Let's see how many bytes we produce. */
139 #ifndef MAX_NEEDED_INPUT
140 # define MAX_NEEDED_INPUT MIN_NEEDED_INPUT
143 /* We produce at least one byte in the next round. */
144 #ifndef MIN_NEEDED_OUTPUT
145 # error "MIN_NEEDED_OUTPUT definition missing"
148 /* Let's see how many bytes we produce. */
149 #ifndef MAX_NEEDED_OUTPUT
150 # define MAX_NEEDED_OUTPUT MIN_NEEDED_OUTPUT
153 /* Default name for the function. */
155 # define LOOPFCT loop
158 /* Make sure we have a loop body. */
160 # error "Definition of BODY missing for function" LOOPFCT
164 /* If no arguments have to passed to the loop function define the macro
166 #ifndef EXTRA_LOOP_DECLS
167 # define EXTRA_LOOP_DECLS
171 /* To make it easier for the writers of the modules, we define a macro
172 to test whether we have to ignore errors. */
173 #define ignore_errors_p() (flags & __GCONV_IGNORE_ERRORS)
176 /* The function returns the status, as defined in gconv.h. */
178 FCTNAME (LOOPFCT) (struct __gconv_step *step,
179 struct __gconv_step_data *step_data,
180 const unsigned char **inptrp, const unsigned char *inend,
181 unsigned char **outptrp, unsigned char *outend,
182 size_t *irreversible EXTRA_LOOP_DECLS)
184 #ifdef LOOP_NEED_STATE
185 mbstate_t *state = step_data->__statep;
187 #ifdef LOOP_NEED_FLAGS
188 int flags = step_data->__flags;
190 #ifdef LOOP_NEED_DATA
191 void *data = step->__data;
193 int result = __GCONV_EMPTY_INPUT;
194 const unsigned char *inptr = *inptrp;
195 unsigned char *outptr = *outptrp;
201 while (inptr != inend)
203 /* `if' cases for MIN_NEEDED_OUTPUT ==/!= 1 is made to help the
204 compiler generating better code. It will optimized away
205 since MIN_NEEDED_OUTPUT is always a constant. */
206 if ((MIN_NEEDED_OUTPUT != 1
207 && __builtin_expect (outptr + MIN_NEEDED_OUTPUT > outend, 0))
208 || (MIN_NEEDED_OUTPUT == 1
209 && __builtin_expect (outptr >= outend, 0)))
211 /* Overflow in the output buffer. */
212 result = __GCONV_FULL_OUTPUT;
215 if (MIN_NEEDED_INPUT > 1
216 && __builtin_expect (inptr + MIN_NEEDED_INPUT > inend, 0))
218 /* We don't have enough input for another complete input
220 result = __GCONV_INCOMPLETE_INPUT;
224 /* Here comes the body the user provides. It can stop with
225 RESULT set to GCONV_INCOMPLETE_INPUT (if the size of the
226 input characters vary in size), GCONV_ILLEGAL_INPUT, or
227 GCONV_FULL_OUTPUT (if the output characters vary in size). */
231 /* Update the pointers pointed to by the parameters. */
242 /* Include the file a second time to define the function to handle
244 #if !defined DEFINE_UNALIGNED && !defined _STRING_ARCH_unaligned \
245 && MIN_NEEDED_FROM != 1 && MAX_NEEDED_FROM % MIN_NEEDED_FROM == 0 \
246 && MIN_NEEDED_TO != 1 && MAX_NEEDED_TO % MIN_NEEDED_TO == 0
253 # define DEFINE_UNALIGNED
255 # undef DEFINE_UNALIGNED
259 #if MAX_NEEDED_INPUT > 1
260 # define SINGLE(fct) SINGLE2 (fct)
261 # define SINGLE2(fct) fct##_single
263 SINGLE(LOOPFCT) (struct __gconv_step *step,
264 struct __gconv_step_data *step_data,
265 const unsigned char **inptrp, const unsigned char *inend,
266 unsigned char **outptrp, unsigned char *outend,
267 size_t *irreversible EXTRA_LOOP_DECLS)
269 mbstate_t *state = step_data->__statep;
270 #ifdef LOOP_NEED_FLAGS
271 int flags = step_data->__flags;
273 #ifdef LOOP_NEED_DATA
274 void *data = step->__data;
276 int result = __GCONV_OK;
277 unsigned char bytebuf[MAX_NEEDED_INPUT];
278 const unsigned char *inptr = *inptrp;
279 unsigned char *outptr = *outptrp;
289 /* Add the bytes from the state to the input buffer. */
290 for (inlen = 0; inlen < (state->__count & 7); ++ inlen)
291 bytebuf[inlen] = state->__value.__wchb[inlen];
294 /* Are there enough bytes in the input buffer? */
295 if (__builtin_expect (inptr + (MIN_NEEDED_INPUT - inlen) > inend, 0))
301 inend = &bytebuf[inlen];
305 /* We don't have enough input for another complete input
307 while (inptr < inend)
308 state->__value.__wchb[inlen++] = *inptr++;
311 return __GCONV_INCOMPLETE_INPUT;
314 /* Enough space in output buffer. */
315 if ((MIN_NEEDED_OUTPUT != 1 && outptr + MIN_NEEDED_OUTPUT > outend)
316 || (MIN_NEEDED_OUTPUT == 1 && outptr >= outend))
317 /* Overflow in the output buffer. */
318 return __GCONV_FULL_OUTPUT;
320 /* Now add characters from the normal input buffer. */
322 bytebuf[inlen++] = *inptr++;
323 while (inlen < MAX_NEEDED_INPUT && inptr < inend);
326 inend = &bytebuf[inlen];
334 /* Now we either have produced an output character and consumed all the
335 bytes from the state and at least one more, or the character is still
336 incomplete, or we have some other error (like illegal input character,
337 no space in output buffer). */
338 if (__builtin_expect (inptr != bytebuf, 1))
340 /* We found a new character. */
341 assert (inptr - bytebuf > (state->__count & 7));
343 *inptrp += inptr - bytebuf - (state->__count & 7);
348 /* Clear the state buffer. */
349 state->__count &= ~7;
351 else if (result == __GCONV_INCOMPLETE_INPUT)
353 /* This can only happen if we have less than MAX_NEEDED_INPUT bytes
355 assert (inend != &bytebuf[MAX_NEEDED_INPUT]);
357 *inptrp += inend - bytebuf - (state->__count & 7);
363 /* We don't have enough input for another complete input
365 while (inptr < inend)
366 state->__value.__wchb[inlen++] = *inptr++;
377 /* We remove the macro definitions so that we can include this file again
378 for the definition of another function. */
379 #undef MIN_NEEDED_INPUT
380 #undef MAX_NEEDED_INPUT
381 #undef MIN_NEEDED_OUTPUT
382 #undef MAX_NEEDED_OUTPUT
386 #undef EXTRA_LOOP_DECLS
390 #undef LOOP_NEED_STATE
391 #undef LOOP_NEED_FLAGS
392 #undef LOOP_NEED_DATA