/* Conversion loop frame work.
- Copyright (C) 1998, 1999, 2000 Free Software Foundation, Inc.
+ Copyright (C) 1998-2019 Free Software Foundation, Inc.
This file is part of the GNU C Library.
Contributed by Ulrich Drepper <drepper@cygnus.com>, 1998.
The GNU C Library is free software; you can redistribute it and/or
- modify it under the terms of the GNU Library General Public License as
- published by the Free Software Foundation; either version 2 of the
- License, or (at your option) any later version.
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
The GNU C Library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
- Library General Public License for more details.
+ Lesser General Public License for more details.
- You should have received a copy of the GNU Library General Public
- License along with the GNU C Library; see the file COPYING.LIB. If not,
- write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
- Boston, MA 02111-1307, USA. */
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <https://www.gnu.org/licenses/>. */
/* This file provides a frame for the reader loop in all conversion modules.
The actual code must (of course) be provided in the actual module source
BODY this is supposed to expand to the body of the loop.
The user must provide this.
- EXTRA_LOOP_DECLS extra arguments passed from converion loop call.
+ EXTRA_LOOP_DECLS extra arguments passed from conversion loop call.
INIT_PARAMS code to define and initialize variables from params.
UPDATE_PARAMS code to store result in params.
+
+ ONEBYTE_BODY body of the specialized conversion function for a
+ single byte from the current character set to INTERNAL.
*/
#include <assert.h>
#include <endian.h>
-#include <gconv.h>
+#include <iconv/gconv_int.h>
#include <stdint.h>
#include <string.h>
#include <wchar.h>
#include <sys/param.h> /* For MIN. */
#define __need_size_t
#include <stddef.h>
-
+#include <libc-diag.h>
/* We have to provide support for machines which are not able to handled
unaligned memory accesses. Some of the character encodings have
representations with a fixed width of 2 or 4 bytes. But if we cannot
access unaligned memory we still have to read byte-wise. */
#undef FCTNAME2
-#if defined _STRING_ARCH_unaligned || !defined DEFINE_UNALIGNED
+#if _STRING_ARCH_unaligned || !defined DEFINE_UNALIGNED
/* We can handle unaligned memory access. */
-# define get16(addr) *((uint16_t *) (addr))
-# define get32(addr) *((uint32_t *) (addr))
+# define get16(addr) *((const uint16_t *) (addr))
+# define get32(addr) *((const uint32_t *) (addr))
/* We need no special support for writing values either. */
# define put16(addr, val) *((uint16_t *) (addr)) = (val)
/* Distinguish between big endian and little endian. */
# if __BYTE_ORDER == __LITTLE_ENDIAN
# define get16(addr) \
- (((__const unsigned char *) (addr))[1] << 8 \
- | ((__const unsigned char *) (addr))[0])
+ (((const unsigned char *) (addr))[1] << 8 \
+ | ((const unsigned char *) (addr))[0])
# define get32(addr) \
- (((((__const unsigned char *) (addr))[3] << 8 \
- | ((__const unsigned char *) (addr))[2]) << 8 \
- | ((__const unsigned char *) (addr))[1]) << 8 \
- | ((__const unsigned char *) (addr))[0])
+ (((((const unsigned char *) (addr))[3] << 8 \
+ | ((const unsigned char *) (addr))[2]) << 8 \
+ | ((const unsigned char *) (addr))[1]) << 8 \
+ | ((const unsigned char *) (addr))[0])
# define put16(addr, val) \
({ uint16_t __val = (val); \
(void) 0; })
# else
# define get16(addr) \
- (((__const unsigned char *) (addr))[0] << 8 \
- | ((__const unsigned char *) (addr))[1])
+ (((const unsigned char *) (addr))[0] << 8 \
+ | ((const unsigned char *) (addr))[1])
# define get32(addr) \
- (((((__const unsigned char *) (addr))[0] << 8 \
- | ((__const unsigned char *) (addr))[1]) << 8 \
- | ((__const unsigned char *) (addr))[2]) << 8 \
- | ((__const unsigned char *) (addr))[3])
+ (((((const unsigned char *) (addr))[0] << 8 \
+ | ((const unsigned char *) (addr))[1]) << 8 \
+ | ((const unsigned char *) (addr))[2]) << 8 \
+ | ((const unsigned char *) (addr))[3])
# define put16(addr, val) \
({ uint16_t __val = (val); \
((unsigned char *) (addr))[1] = __val; \
- ((unsigned char *) (addr))[2] = __val >> 8; \
+ ((unsigned char *) (addr))[0] = __val >> 8; \
(void) 0; })
# define put32(addr, val) \
({ uint32_t __val = (val); \
/* We need at least one byte for the next round. */
#ifndef MIN_NEEDED_INPUT
# error "MIN_NEEDED_INPUT definition missing"
+#elif MIN_NEEDED_INPUT < 1
+# error "MIN_NEEDED_INPUT must be >= 1"
#endif
/* Let's see how many bytes we produce. */
/* We produce at least one byte in the next round. */
#ifndef MIN_NEEDED_OUTPUT
# error "MIN_NEEDED_OUTPUT definition missing"
+#elif MIN_NEEDED_OUTPUT < 1
+# error "MIN_NEEDED_OUTPUT must be >= 1"
#endif
/* Let's see how many bytes we produce. */
# define EXTRA_LOOP_DECLS
#endif
+/* Allow using UPDATE_PARAMS in macros where #ifdef UPDATE_PARAMS test
+ isn't possible. */
+#ifndef UPDATE_PARAMS
+# define UPDATE_PARAMS do { } while (0)
+#endif
+#ifndef REINIT_PARAMS
+# define REINIT_PARAMS do { } while (0)
+#endif
+
/* To make it easier for the writers of the modules, we define a macro
to test whether we have to ignore errors. */
-#define ignore_errors_p() (flags & __GCONV_IGNORE_ERRORS)
+#define ignore_errors_p() \
+ (irreversible != NULL && (flags & __GCONV_IGNORE_ERRORS))
+
+
+/* Error handling for the FROM_LOOP direction, with ignoring of errors.
+ Note that we cannot use the do while (0) trick since `break' and
+ `continue' must reach certain points. */
+#define STANDARD_FROM_LOOP_ERR_HANDLER(Incr) \
+ { \
+ result = __GCONV_ILLEGAL_INPUT; \
+ \
+ if (! ignore_errors_p ()) \
+ break; \
+ \
+ /* We ignore the invalid input byte sequence. */ \
+ inptr += (Incr); \
+ ++*irreversible; \
+ /* But we keep result == __GCONV_ILLEGAL_INPUT, because of the constraint \
+ that "iconv -c" must give the same exitcode as "iconv". */ \
+ continue; \
+ }
+
+/* Error handling for the TO_LOOP direction, with use of transliteration/
+ transcription functions and ignoring of errors. Note that we cannot use
+ the do while (0) trick since `break' and `continue' must reach certain
+ points. */
+#define STANDARD_TO_LOOP_ERR_HANDLER(Incr) \
+ { \
+ result = __GCONV_ILLEGAL_INPUT; \
+ \
+ if (irreversible == NULL) \
+ /* This means we are in call from __gconv_transliterate. In this \
+ case we are not doing any error recovery outself. */ \
+ break; \
+ \
+ /* If needed, flush any conversion state, so that __gconv_transliterate \
+ starts with current shift state. */ \
+ UPDATE_PARAMS; \
+ \
+ /* First try the transliteration methods. */ \
+ if ((step_data->__flags & __GCONV_TRANSLIT) != 0) \
+ result = __gconv_transliterate \
+ (step, step_data, *inptrp, \
+ &inptr, inend, &outptr, irreversible); \
+ \
+ REINIT_PARAMS; \
+ \
+ /* If any of them recognized the input continue with the loop. */ \
+ if (result != __GCONV_ILLEGAL_INPUT) \
+ { \
+ if (__glibc_unlikely (result == __GCONV_FULL_OUTPUT)) \
+ break; \
+ \
+ continue; \
+ } \
+ \
+ /* Next see whether we have to ignore the error. If not, stop. */ \
+ if (! ignore_errors_p ()) \
+ break; \
+ \
+ /* When we come here it means we ignore the character. */ \
+ ++*irreversible; \
+ inptr += Incr; \
+ /* But we keep result == __GCONV_ILLEGAL_INPUT, because of the constraint \
+ that "iconv -c" must give the same exitcode as "iconv". */ \
+ continue; \
+ }
+
+
+/* With GCC 7 when compiling with -Os for 32-bit s390 the compiler
+ warns that the variable 'ch', in the definition of BODY in
+ sysdeps/s390/multiarch/8bit-generic.c, may be used uninitialized in
+ the call to UNICODE_TAG_HANDLER in that macro. This variable is
+ actually always initialized before use, in the prior loop if INDEX
+ is nonzero and in the following 'if' if INDEX is zero. That code
+ has a comment referencing this diagnostic disabling; updates in one
+ place may require updates in the other. */
+DIAG_PUSH_NEEDS_COMMENT;
+DIAG_IGNORE_Os_NEEDS_COMMENT (7, "-Wmaybe-uninitialized");
+/* Handling of Unicode 3.1 TAG characters. Unicode recommends
+ "If language codes are not relevant to the particular processing
+ operation, then they should be ignored." This macro is usually
+ called right before STANDARD_TO_LOOP_ERR_HANDLER (Incr). */
+#define UNICODE_TAG_HANDLER(Character, Incr) \
+ { \
+ /* TAG characters are those in the range U+E0000..U+E007F. */ \
+ if (((Character) >> 7) == (0xe0000 >> 7)) \
+ { \
+ inptr += Incr; \
+ continue; \
+ } \
+ }
+DIAG_POP_NEEDS_COMMENT;
/* The function returns the status, as defined in gconv.h. */
static inline int
-FCTNAME (LOOPFCT) (const unsigned char **inptrp, const unsigned char *inend,
- unsigned char **outptrp, unsigned char *outend,
- mbstate_t *state, int flags, void *data, size_t *converted
- EXTRA_LOOP_DECLS)
+__attribute ((always_inline))
+FCTNAME (LOOPFCT) (struct __gconv_step *step,
+ struct __gconv_step_data *step_data,
+ const unsigned char **inptrp, const unsigned char *inend,
+ unsigned char **outptrp, const unsigned char *outend,
+ size_t *irreversible EXTRA_LOOP_DECLS)
{
- int result = __GCONV_OK;
+#ifdef LOOP_NEED_STATE
+ mbstate_t *state = step_data->__statep;
+#endif
+#ifdef LOOP_NEED_FLAGS
+ int flags = step_data->__flags;
+#endif
+#ifdef LOOP_NEED_DATA
+ void *data = step->__data;
+#endif
+ int result = __GCONV_EMPTY_INPUT;
const unsigned char *inptr = *inptrp;
unsigned char *outptr = *outptrp;
- /* We run one loop where we avoid checks for underflow/overflow of the
- buffers to speed up the conversion a bit. */
- size_t min_in_rounds = (inend - inptr) / MAX_NEEDED_INPUT;
- size_t min_out_rounds = (outend - outptr) / MAX_NEEDED_OUTPUT;
- size_t min_rounds = MIN (min_in_rounds, min_out_rounds);
-
#ifdef INIT_PARAMS
INIT_PARAMS;
#endif
-#undef NEED_LENGTH_TEST
-#define NEED_LENGTH_TEST 0
- while (min_rounds-- > 0)
+ while (inptr != inend)
{
- /* Here comes the body the user provides. It can stop with RESULT
- set to GCONV_INCOMPLETE_INPUT (if the size of the input characters
- vary in size), GCONV_ILLEGAL_INPUT, or GCONV_FULL_OUTPUT (if the
- output characters vary in size. */
- BODY
- }
-
- if (result == __GCONV_OK)
- {
-#if MIN_NEEDED_INPUT == MAX_NEEDED_INPUT \
- && MIN_NEEDED_OUTPUT == MAX_NEEDED_OUTPUT
- /* We don't need to start another loop since we were able to determine
- the maximal number of characters to copy in advance. What remains
- to be determined is the status. */
- if (inptr == inend)
- /* No more input. */
- result = __GCONV_EMPTY_INPUT;
- else if ((MIN_NEEDED_OUTPUT != 1 && outptr + MIN_NEEDED_OUTPUT > outend)
- || (MIN_NEEDED_OUTPUT == 1 && outptr >= outend))
- /* Overflow in the output buffer. */
- result = __GCONV_FULL_OUTPUT;
- else
- /* We have something left in the input buffer. */
- result = __GCONV_INCOMPLETE_INPUT;
-#else
- result = __GCONV_EMPTY_INPUT;
-
-# undef NEED_LENGTH_TEST
-# define NEED_LENGTH_TEST 1
- while (inptr != inend)
+ /* `if' cases for MIN_NEEDED_OUTPUT ==/!= 1 is made to help the
+ compiler generating better code. They will be optimized away
+ since MIN_NEEDED_OUTPUT is always a constant. */
+ if (MIN_NEEDED_INPUT > 1
+ && __builtin_expect (inptr + MIN_NEEDED_INPUT > inend, 0))
+ {
+ /* We don't have enough input for another complete input
+ character. */
+ result = __GCONV_INCOMPLETE_INPUT;
+ break;
+ }
+ if ((MIN_NEEDED_OUTPUT != 1
+ && __builtin_expect (outptr + MIN_NEEDED_OUTPUT > outend, 0))
+ || (MIN_NEEDED_OUTPUT == 1
+ && __builtin_expect (outptr >= outend, 0)))
{
- /* `if' cases for MIN_NEEDED_OUTPUT ==/!= 1 is made to help the
- compiler generating better code. It will optimized away
- since MIN_NEEDED_OUTPUT is always a constant. */
- if ((MIN_NEEDED_OUTPUT != 1
- && __builtin_expect (outptr + MIN_NEEDED_OUTPUT > outend, 0))
- || (MIN_NEEDED_OUTPUT == 1
- && __builtin_expect (outptr >= outend, 0)))
- {
- /* Overflow in the output buffer. */
- result = __GCONV_FULL_OUTPUT;
- break;
- }
- if (MIN_NEEDED_INPUT > 1
- && __builtin_expect (inptr + MIN_NEEDED_INPUT > inend, 0))
- {
- /* We don't have enough input for another complete input
- character. */
- result = __GCONV_INCOMPLETE_INPUT;
- break;
- }
-
- /* Here comes the body the user provides. It can stop with
- RESULT set to GCONV_INCOMPLETE_INPUT (if the size of the
- input characters vary in size), GCONV_ILLEGAL_INPUT, or
- GCONV_FULL_OUTPUT (if the output characters vary in size). */
- BODY
+ /* Overflow in the output buffer. */
+ result = __GCONV_FULL_OUTPUT;
+ break;
}
-#endif /* Input and output charset are not both fixed width. */
+
+ /* Here comes the body the user provides. It can stop with
+ RESULT set to GCONV_INCOMPLETE_INPUT (if the size of the
+ input characters vary in size), GCONV_ILLEGAL_INPUT, or
+ GCONV_FULL_OUTPUT (if the output characters vary in size). */
+ BODY
}
/* Update the pointers pointed to by the parameters. */
*inptrp = inptr;
*outptrp = outptr;
-#ifdef UPDATE_PARAMS
UPDATE_PARAMS;
-#endif
return result;
}
-/* Include the file a second time to define the function to define the
- function to handle unaligned access. */
-#if !defined DEFINE_UNALIGNED && !defined _STRING_ARCH_unaligned \
- && MIN_NEEDED_FROM != 1 && MAX_NEEDED_FROM % MIN_NEEDED_FROM == 0 \
- && MIN_NEEDED_TO != 1 && MAX_NEEDED_TO % MIN_NEEDED_TO == 0
+/* Include the file a second time to define the function to handle
+ unaligned access. */
+#if !defined DEFINE_UNALIGNED && !_STRING_ARCH_unaligned \
+ && MIN_NEEDED_INPUT != 1 && MAX_NEEDED_INPUT % MIN_NEEDED_INPUT == 0 \
+ && MIN_NEEDED_OUTPUT != 1 && MAX_NEEDED_OUTPUT % MIN_NEEDED_OUTPUT == 0
# undef get16
# undef get32
# undef put16
# define DEFINE_UNALIGNED
# include "loop.c"
# undef DEFINE_UNALIGNED
-#endif
-
-
-#if MAX_NEEDED_INPUT > 1
-# define SINGLE(fct) SINGLE2 (fct)
-# define SINGLE2(fct) fct##_single
+#else
+# if MAX_NEEDED_INPUT > 1
+# define SINGLE(fct) SINGLE2 (fct)
+# define SINGLE2(fct) fct##_single
static inline int
-SINGLE(LOOPFCT) (const unsigned char **inptrp, const unsigned char *inend,
+__attribute ((always_inline))
+SINGLE(LOOPFCT) (struct __gconv_step *step,
+ struct __gconv_step_data *step_data,
+ const unsigned char **inptrp, const unsigned char *inend,
unsigned char **outptrp, unsigned char *outend,
- mbstate_t *state, int flags, void *data, size_t *converted
- EXTRA_LOOP_DECLS)
+ size_t *irreversible EXTRA_LOOP_DECLS)
{
+ mbstate_t *state = step_data->__statep;
+# ifdef LOOP_NEED_FLAGS
+ int flags = step_data->__flags;
+# endif
+# ifdef LOOP_NEED_DATA
+ void *data = step->__data;
+# endif
int result = __GCONV_OK;
unsigned char bytebuf[MAX_NEEDED_INPUT];
const unsigned char *inptr = *inptrp;
unsigned char *outptr = *outptrp;
size_t inlen;
-#ifdef INIT_PARAMS
+# ifdef INIT_PARAMS
INIT_PARAMS;
-#endif
+# endif
-#ifdef UNPACK_BYTES
+# ifdef UNPACK_BYTES
UNPACK_BYTES
-#else
+# else
/* Add the bytes from the state to the input buffer. */
- for (inlen = 0; inlen < (state->__count & 7); ++ inlen)
+ assert ((state->__count & 7) <= sizeof (state->__value));
+ for (inlen = 0; inlen < (size_t) (state->__count & 7); ++inlen)
bytebuf[inlen] = state->__value.__wchb[inlen];
-#endif
+# endif
/* Are there enough bytes in the input buffer? */
- if (__builtin_expect (inptr + (MIN_NEEDED_INPUT - inlen) > inend, 0))
+ if (MIN_NEEDED_INPUT > 1
+ && __builtin_expect (inptr + (MIN_NEEDED_INPUT - inlen) > inend, 0))
{
*inptrp = inend;
-#ifdef STORE_REST
+# ifdef STORE_REST
+
+ /* Building with -O3 GCC emits a `array subscript is above array
+ bounds' warning. GCC BZ #64739 has been opened for this. */
+ DIAG_PUSH_NEEDS_COMMENT;
+ DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Warray-bounds");
+ while (inptr < inend)
+ bytebuf[inlen++] = *inptr++;
+ DIAG_POP_NEEDS_COMMENT;
+
inptr = bytebuf;
inptrp = &inptr;
inend = &bytebuf[inlen];
STORE_REST
-#else
+# else
/* We don't have enough input for another complete input
character. */
while (inptr < inend)
state->__value.__wchb[inlen++] = *inptr++;
-#endif
+# endif
return __GCONV_INCOMPLETE_INPUT;
}
inptr = bytebuf;
inend = &bytebuf[inlen];
-#undef NEED_LENGTH_TEST
-#define NEED_LENGTH_TEST 1
+
do
{
BODY
bytes from the state and at least one more, or the character is still
incomplete, or we have some other error (like illegal input character,
no space in output buffer). */
- if (__builtin_expect (inptr != bytebuf, 1))
+ if (__glibc_likely (inptr != bytebuf))
{
/* We found a new character. */
assert (inptr - bytebuf > (state->__count & 7));
result = __GCONV_OK;
/* Clear the state buffer. */
+# ifdef CLEAR_STATE
+ CLEAR_STATE;
+# else
state->__count &= ~7;
+# endif
}
else if (result == __GCONV_INCOMPLETE_INPUT)
{
assert (inend != &bytebuf[MAX_NEEDED_INPUT]);
*inptrp += inend - bytebuf - (state->__count & 7);
-#ifdef STORE_REST
+# ifdef STORE_REST
inptrp = &inptr;
STORE_REST
-#else
+# else
/* We don't have enough input for another complete input
character. */
+ assert (inend - inptr > (state->__count & ~7));
+ assert (inend - inptr <= sizeof (state->__value));
+ state->__count = (state->__count & ~7) | (inend - inptr);
+ inlen = 0;
while (inptr < inend)
state->__value.__wchb[inlen++] = *inptr++;
-#endif
+# endif
}
return result;
}
-# undef SINGLE
-# undef SINGLE2
-#endif
+# undef SINGLE
+# undef SINGLE2
+# endif
+# ifdef ONEBYTE_BODY
+/* Define the shortcut function for btowc. */
+static wint_t
+gconv_btowc (struct __gconv_step *step, unsigned char c)
+ ONEBYTE_BODY
+# define FROM_ONEBYTE gconv_btowc
+# endif
+
+#endif
+
/* We remove the macro definitions so that we can include this file again
for the definition of another function. */
#undef MIN_NEEDED_INPUT
#undef EXTRA_LOOP_DECLS
#undef INIT_PARAMS
#undef UPDATE_PARAMS
+#undef REINIT_PARAMS
+#undef ONEBYTE_BODY
+#undef UNPACK_BYTES
+#undef CLEAR_STATE
+#undef LOOP_NEED_STATE
+#undef LOOP_NEED_FLAGS
+#undef LOOP_NEED_DATA
#undef get16
#undef get32
#undef put16
#undef put32
#undef unaligned
-#undef UNPACK_BYTES
projects / thirdparty / glibc.git / blobdiff