[thirdparty/glibc.git] / iconv / loop.c

/* Conversion loop frame work.
   Copyright (C) 1998-2022 Free Software Foundation, Inc.
   This file is part of the GNU C Library.

   The GNU C Library is free software; you can redistribute it and/or
   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
   Lesser General Public License for more details.

   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
   code but certain actions can be written down generically, with some
   customization options which are these:

     MIN_NEEDED_INPUT	minimal number of input bytes needed for the next
			conversion.
     MIN_NEEDED_OUTPUT	minimal number of bytes produced by the next round
			of conversion.

     MAX_NEEDED_INPUT	you guess it, this is the maximal number of input
			bytes needed.  It defaults to MIN_NEEDED_INPUT
     MAX_NEEDED_OUTPUT	likewise for output bytes.

     LOOPFCT		name of the function created.  If not specified
			the name is `loop' but this prevents the use
			of multiple functions in the same file.

     BODY		this is supposed to expand to the body of the loop.
			The user must provide this.

     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 <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 _STRING_ARCH_unaligned || !defined DEFINE_UNALIGNED
/* We can handle unaligned memory access.  */
# 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)
# define put32(addr, val) *((uint32_t *) (addr)) = (val)

# define FCTNAME2(name) name
#else
/* 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])
#  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])

#  define put16(addr, val) \
     ({ uint16_t __val = (val);						      \
	((unsigned char *) (addr))[0] = __val;				      \
	((unsigned char *) (addr))[1] = __val >> 8;			      \
	(void) 0; })
#  define put32(addr, val) \
     ({ uint32_t __val = (val);						      \
	((unsigned char *) (addr))[0] = __val;				      \
	__val >>= 8;							      \
	((unsigned char *) (addr))[1] = __val;				      \
	__val >>= 8;							      \
	((unsigned char *) (addr))[2] = __val;				      \
	__val >>= 8;							      \
	((unsigned char *) (addr))[3] = __val;				      \
	(void) 0; })
# else
#  define get16(addr) \
     (((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])

#  define put16(addr, val) \
     ({ uint16_t __val = (val);						      \
	((unsigned char *) (addr))[1] = __val;				      \
	((unsigned char *) (addr))[0] = __val >> 8;			      \
	(void) 0; })
#  define put32(addr, val) \
     ({ uint32_t __val = (val);						      \
	((unsigned char *) (addr))[3] = __val;				      \
	__val >>= 8;							      \
	((unsigned char *) (addr))[2] = __val;				      \
	__val >>= 8;							      \
	((unsigned char *) (addr))[1] = __val;				      \
	__val >>= 8;							      \
	((unsigned char *) (addr))[0] = __val;				      \
	(void) 0; })
# endif

# define FCTNAME2(name) name##_unaligned
#endif
#define FCTNAME(name) FCTNAME2(name)


/* 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.  */
#ifndef MAX_NEEDED_INPUT
# define MAX_NEEDED_INPUT	MIN_NEEDED_INPUT
#endif

/* 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.  */
#ifndef MAX_NEEDED_OUTPUT
# define MAX_NEEDED_OUTPUT	MIN_NEEDED_OUTPUT
#endif

/* Default name for the function.  */
#ifndef LOOPFCT
# define LOOPFCT		loop
#endif

/* Make sure we have a loop body.  */
#ifndef BODY
# error "Definition of BODY missing for function" LOOPFCT
#endif


/* If no arguments have to passed to the loop function define the macro
   as empty.  */
#ifndef EXTRA_LOOP_DECLS
# 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() \
  (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
__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)
{
#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;

#ifdef INIT_PARAMS
  INIT_PARAMS;
#endif

  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)))
	{
	  /* Overflow in the output buffer.  */
	  result = __GCONV_FULL_OUTPUT;
	  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
    }

  /* Update the pointers pointed to by the parameters.  */
  *inptrp = inptr;
  *outptrp = outptr;
  UPDATE_PARAMS;

  return result;
}


/* 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
# undef put32
# undef unaligned

# define DEFINE_UNALIGNED
# include "loop.c"
# undef DEFINE_UNALIGNED
#else
# if MAX_NEEDED_INPUT > 1
#  define SINGLE(fct) SINGLE2 (fct)
#  define SINGLE2(fct) fct##_single
static inline int
__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,
		 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
  INIT_PARAMS;
#  endif

#  ifdef UNPACK_BYTES
  UNPACK_BYTES
#  else
  /* Add the bytes from the state to the input buffer.  */
  assert ((state->__count & 7) <= sizeof (state->__value));
  for (inlen = 0; inlen < (size_t) (state->__count & 7); ++inlen)
    bytebuf[inlen] = state->__value.__wchb[inlen];
#  endif

  /* Are there enough bytes in the input buffer?  */
  if (MIN_NEEDED_INPUT > 1
      && __builtin_expect (inptr + (MIN_NEEDED_INPUT - inlen) > inend, 0))
    {
      *inptrp = inend;
#  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
      /* We don't have enough input for another complete input
	 character.  */
      size_t inlen_after = inlen + (inend - inptr);
      assert (inlen_after <= sizeof (state->__value.__wchb));
      for (; inlen < inlen_after; inlen++)
	state->__value.__wchb[inlen] = *inptr++;
#  endif

      return __GCONV_INCOMPLETE_INPUT;
    }

  /* Enough space in output buffer.  */
  if ((MIN_NEEDED_OUTPUT != 1 && outptr + MIN_NEEDED_OUTPUT > outend)
      || (MIN_NEEDED_OUTPUT == 1 && outptr >= outend))
    /* Overflow in the output buffer.  */
    return __GCONV_FULL_OUTPUT;

  /*  Now add characters from the normal input buffer.  */
  if (inlen >= MAX_NEEDED_INPUT)
    /* Avoid a -Wstringop-overflow= warning when this loop is
       unrolled.  The compiler cannot otherwise see that this is
       unreachable because it depends on (state->__count & 7) not
       being too large after a previous conversion step.  */
    __builtin_unreachable ();
  do
    bytebuf[inlen++] = *inptr++;
  while (inlen < MAX_NEEDED_INPUT && inptr < inend);

  inptr = bytebuf;
  inend = &bytebuf[inlen];

  do
    {
      BODY
    }
  while (0);

  /* Now we either have produced an output character and consumed all the
     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 (__glibc_likely (inptr != bytebuf))
    {
      /* We found a new character.  */
      assert (inptr - bytebuf > (state->__count & 7));

      *inptrp += inptr - bytebuf - (state->__count & 7);
      *outptrp = outptr;

      result = __GCONV_OK;

      /* Clear the state buffer.  */
#  ifdef CLEAR_STATE
      CLEAR_STATE;
#  else
      state->__count &= ~7;
#  endif
    }
  else if (result == __GCONV_INCOMPLETE_INPUT)
    {
      /* This can only happen if we have less than MAX_NEEDED_INPUT bytes
	 available.  */
      assert (inend != &bytebuf[MAX_NEEDED_INPUT]);

      *inptrp += inend - bytebuf - (state->__count & 7);
#  ifdef STORE_REST
      inptrp = &inptr;

      STORE_REST
#  else
      /* We don't have enough input for another complete input
	 character.  */
      assert (inend - inptr > (state->__count & ~7));
      assert (inend - inptr <= sizeof (state->__value.__wchb));
      state->__count = (state->__count & ~7) | (inend - inptr);
      for (inlen = 0; inlen < inend - inptr; inlen++)
	state->__value.__wchb[inlen] = inptr[inlen];
      inptr = inend;
#  endif
    }

  return result;
}
#  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 MAX_NEEDED_INPUT
#undef MIN_NEEDED_OUTPUT
#undef MAX_NEEDED_OUTPUT
#undef LOOPFCT
#undef BODY
#undef LOOPFCT
#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
Commit	Line	Data
8619129f	1	/* Conversion loop frame work.
581c785b	2	Copyright (C) 1998-2022 Free Software Foundation, Inc.
8619129f	3	This file is part of the GNU C Library.
8619129f UD	4
8619129f UD	5	The GNU C Library is free software; you can redistribute it and/or
41bdb6e2 AJ	6	modify it under the terms of the GNU Lesser General Public
	7	License as published by the Free Software Foundation; either
	8	version 2.1 of the License, or (at your option) any later version.
8619129f UD	9
	10	The GNU C Library is distributed in the hope that it will be useful,
	11	but WITHOUT ANY WARRANTY; without even the implied warranty of
	12	MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
41bdb6e2	13	Lesser General Public License for more details.
8619129f	14
41bdb6e2	15	You should have received a copy of the GNU Lesser General Public
59ba27a6	16	License along with the GNU C Library; if not, see
5a82c748	17	<https://www.gnu.org/licenses/>. */
8619129f UD	18
	19	/* This file provides a frame for the reader loop in all conversion modules.
	20	The actual code must (of course) be provided in the actual module source
	21	code but certain actions can be written down generically, with some
	22	customization options which are these:
	23
	24	MIN_NEEDED_INPUT minimal number of input bytes needed for the next
	25	conversion.
	26	MIN_NEEDED_OUTPUT minimal number of bytes produced by the next round
	27	of conversion.
	28
	29	MAX_NEEDED_INPUT you guess it, this is the maximal number of input
	30	bytes needed. It defaults to MIN_NEEDED_INPUT
	31	MAX_NEEDED_OUTPUT likewise for output bytes.
	32
8619129f UD	33	LOOPFCT name of the function created. If not specified
	34	the name is `loop' but this prevents the use
	35	of multiple functions in the same file.
	36
8619129f UD	37	BODY this is supposed to expand to the body of the loop.
8619129f UD	38	The user must provide this.
28f1c862	39
382466e0	40	EXTRA_LOOP_DECLS extra arguments passed from conversion loop call.
66175fa8 UD	41
	42	INIT_PARAMS code to define and initialize variables from params.
	43	UPDATE_PARAMS code to store result in params.
f9ad060c UD	44
	45	ONEBYTE_BODY body of the specialized conversion function for a
	46	single byte from the current character set to INTERNAL.
8619129f UD	47	*/
8619129f UD	48
fd1b5c0f	49	#include <assert.h>
b35e58e4	50	#include <endian.h>
7ac6fad9	51	#include <iconv/gconv_int.h>
b35e58e4 UD	52	#include <stdint.h>
b35e58e4 UD	53	#include <string.h>
d64b6ad0	54	#include <wchar.h>
8619129f UD	55	#include <sys/param.h> /* For MIN. */
	56	#define __need_size_t
	57	#include <stddef.h>
9090848d	58	#include <libc-diag.h>
8619129f	59
b35e58e4 UD	60	/* We have to provide support for machines which are not able to handled
	61	unaligned memory accesses. Some of the character encodings have
	62	representations with a fixed width of 2 or 4 bytes. But if we cannot
	63	access unaligned memory we still have to read byte-wise. */
	64	#undef FCTNAME2
27822ce6	65	#if _STRING_ARCH_unaligned \|\| !defined DEFINE_UNALIGNED
b35e58e4	66	/* We can handle unaligned memory access. */
a784e502 UD	67	# define get16(addr) ((const uint16_t ) (addr))
a784e502 UD	68	# define get32(addr) ((const uint32_t ) (addr))
b35e58e4 UD	69
	70	/* We need no special support for writing values either. */
	71	# define put16(addr, val) ((uint16_t ) (addr)) = (val)
	72	# define put32(addr, val) ((uint32_t ) (addr)) = (val)
	73
	74	# define FCTNAME2(name) name
	75	#else
	76	/* Distinguish between big endian and little endian. */
	77	# if __BYTE_ORDER == __LITTLE_ENDIAN
	78	# define get16(addr) \
a784e502 UD	79	(((const unsigned char *) (addr))[1] << 8 \
a784e502 UD	80	\| ((const unsigned char *) (addr))[0])
b35e58e4	81	# define get32(addr) \
a784e502 UD	82	(((((const unsigned char *) (addr))[3] << 8 \
	83	\| ((const unsigned char *) (addr))[2]) << 8 \
	84	\| ((const unsigned char *) (addr))[1]) << 8 \
	85	\| ((const unsigned char *) (addr))[0])
b35e58e4	86
cb2c5501	87	# define put16(addr, val) \
b35e58e4	88	({ uint16_t __val = (val); \
cb2c5501 UD	89	((unsigned char *) (addr))[0] = __val; \
cb2c5501 UD	90	((unsigned char *) (addr))[1] = __val >> 8; \
b35e58e4	91	(void) 0; })
cb2c5501 UD	92	# define put32(addr, val) \
	93	({ uint32_t __val = (val); \
	94	((unsigned char *) (addr))[0] = __val; \
b35e58e4	95	__val >>= 8; \
cb2c5501	96	((unsigned char *) (addr))[1] = __val; \
b35e58e4	97	__val >>= 8; \
cb2c5501	98	((unsigned char *) (addr))[2] = __val; \
b35e58e4	99	__val >>= 8; \
cb2c5501	100	((unsigned char *) (addr))[3] = __val; \
b35e58e4 UD	101	(void) 0; })
	102	# else
	103	# define get16(addr) \
a784e502 UD	104	(((const unsigned char *) (addr))[0] << 8 \
a784e502 UD	105	\| ((const unsigned char *) (addr))[1])
b35e58e4	106	# define get32(addr) \
a784e502 UD	107	(((((const unsigned char *) (addr))[0] << 8 \
	108	\| ((const unsigned char *) (addr))[1]) << 8 \
	109	\| ((const unsigned char *) (addr))[2]) << 8 \
	110	\| ((const unsigned char *) (addr))[3])
b35e58e4	111
cb2c5501	112	# define put16(addr, val) \
b35e58e4	113	({ uint16_t __val = (val); \
cb2c5501	114	((unsigned char *) (addr))[1] = __val; \
f7ccf2fc	115	((unsigned char *) (addr))[0] = __val >> 8; \
b35e58e4	116	(void) 0; })
cb2c5501 UD	117	# define put32(addr, val) \
	118	({ uint32_t __val = (val); \
	119	((unsigned char *) (addr))[3] = __val; \
b35e58e4	120	__val >>= 8; \
cb2c5501	121	((unsigned char *) (addr))[2] = __val; \
b35e58e4	122	__val >>= 8; \
cb2c5501	123	((unsigned char *) (addr))[1] = __val; \
b35e58e4	124	__val >>= 8; \
cb2c5501	125	((unsigned char *) (addr))[0] = __val; \
b35e58e4 UD	126	(void) 0; })
	127	# endif
	128
	129	# define FCTNAME2(name) name##_unaligned
	130	#endif
	131	#define FCTNAME(name) FCTNAME2(name)
	132
	133
8619129f UD	134	/* We need at least one byte for the next round. */
8619129f UD	135	#ifndef MIN_NEEDED_INPUT
5aa8ff62	136	# error "MIN_NEEDED_INPUT definition missing"
4a0de63b UD	137	#elif MIN_NEEDED_INPUT < 1
4a0de63b UD	138	# error "MIN_NEEDED_INPUT must be >= 1"
8619129f UD	139	#endif
	140
	141	/* Let's see how many bytes we produce. */
	142	#ifndef MAX_NEEDED_INPUT
	143	# define MAX_NEEDED_INPUT MIN_NEEDED_INPUT
	144	#endif
	145
	146	/* We produce at least one byte in the next round. */
	147	#ifndef MIN_NEEDED_OUTPUT
5aa8ff62	148	# error "MIN_NEEDED_OUTPUT definition missing"
c0a0f9a3 UD	149	#elif MIN_NEEDED_OUTPUT < 1
c0a0f9a3 UD	150	# error "MIN_NEEDED_OUTPUT must be >= 1"
8619129f UD	151	#endif
	152
	153	/* Let's see how many bytes we produce. */
	154	#ifndef MAX_NEEDED_OUTPUT
	155	# define MAX_NEEDED_OUTPUT MIN_NEEDED_OUTPUT
	156	#endif
	157
	158	/* Default name for the function. */
	159	#ifndef LOOPFCT
	160	# define LOOPFCT loop
	161	#endif
	162
	163	/* Make sure we have a loop body. */
	164	#ifndef BODY
	165	# error "Definition of BODY missing for function" LOOPFCT
	166	#endif
	167
8619129f	168
28f1c862 UD	169	/* If no arguments have to passed to the loop function define the macro
	170	as empty. */
	171	#ifndef EXTRA_LOOP_DECLS
	172	# define EXTRA_LOOP_DECLS
	173	#endif
	174
4b1b449d UD	175	/* Allow using UPDATE_PARAMS in macros where #ifdef UPDATE_PARAMS test
	176	isn't possible. */
	177	#ifndef UPDATE_PARAMS
	178	# define UPDATE_PARAMS do { } while (0)
	179	#endif
	180	#ifndef REINIT_PARAMS
	181	# define REINIT_PARAMS do { } while (0)
	182	#endif
	183
28f1c862	184
85830c4c UD	185	/* To make it easier for the writers of the modules, we define a macro
85830c4c UD	186	to test whether we have to ignore errors. */
b572c2da UD	187	#define ignore_errors_p() \
b572c2da UD	188	(irreversible != NULL && (flags & __GCONV_IGNORE_ERRORS))
85830c4c UD	189
85830c4c UD	190
e438a468 UD	191	/* Error handling for the FROM_LOOP direction, with ignoring of errors.
	192	Note that we cannot use the do while (0) trick since `break' and
	193	`continue' must reach certain points. */
	194	#define STANDARD_FROM_LOOP_ERR_HANDLER(Incr) \
	195	{ \
	196	result = __GCONV_ILLEGAL_INPUT; \
	197	\
	198	if (! ignore_errors_p ()) \
	199	break; \
	200	\
	201	/* We ignore the invalid input byte sequence. */ \
	202	inptr += (Incr); \
	203	++*irreversible; \
	204	/* But we keep result == __GCONV_ILLEGAL_INPUT, because of the constraint \
	205	that "iconv -c" must give the same exitcode as "iconv". */ \
	206	continue; \
	207	}
	208
	209	/* Error handling for the TO_LOOP direction, with use of transliteration/
	210	transcription functions and ignoring of errors. Note that we cannot use
	211	the do while (0) trick since `break' and `continue' must reach certain
	212	points. */
	213	#define STANDARD_TO_LOOP_ERR_HANDLER(Incr) \
d6204268	214	{ \
d6204268	215	result = __GCONV_ILLEGAL_INPUT; \
b572c2da UD	216	\
	217	if (irreversible == NULL) \
	218	/* This means we are in call from __gconv_transliterate. In this \
	219	case we are not doing any error recovery outself. */ \
	220	break; \
	221	\
4b1b449d UD	222	/* If needed, flush any conversion state, so that __gconv_transliterate \
	223	starts with current shift state. */ \
	224	UPDATE_PARAMS; \
	225	\
d6204268	226	/* First try the transliteration methods. */ \
ba7b4d29 FW	227	if ((step_data->__flags & __GCONV_TRANSLIT) != 0) \
	228	result = __gconv_transliterate \
	229	(step, step_data, *inptrp, \
	230	&inptr, inend, &outptr, irreversible); \
4b1b449d UD	231	\
	232	REINIT_PARAMS; \
	233	\
7888313d	234	/* If any of them recognized the input continue with the loop. */ \
d6204268	235	if (result != __GCONV_ILLEGAL_INPUT) \
f2a8406a	236	{ \
a1ffb40e	237	if (__glibc_unlikely (result == __GCONV_FULL_OUTPUT)) \
f2a8406a UD	238	break; \
	239	\
	240	continue; \
	241	} \
d6204268 UD	242	\
	243	/* Next see whether we have to ignore the error. If not, stop. */ \
	244	if (! ignore_errors_p ()) \
	245	break; \
b572c2da	246	\
d6204268 UD	247	/* When we come here it means we ignore the character. */ \
	248	++*irreversible; \
	249	inptr += Incr; \
e438a468 UD	250	/* But we keep result == __GCONV_ILLEGAL_INPUT, because of the constraint \
e438a468 UD	251	that "iconv -c" must give the same exitcode as "iconv". */ \
d6204268 UD	252	continue; \
	253	}
	254
	255
6900d2ca JM	256	/* With GCC 7 when compiling with -Os for 32-bit s390 the compiler
	257	warns that the variable 'ch', in the definition of BODY in
	258	sysdeps/s390/multiarch/8bit-generic.c, may be used uninitialized in
	259	the call to UNICODE_TAG_HANDLER in that macro. This variable is
	260	actually always initialized before use, in the prior loop if INDEX
	261	is nonzero and in the following 'if' if INDEX is zero. That code
	262	has a comment referencing this diagnostic disabling; updates in one
	263	place may require updates in the other. */
	264	DIAG_PUSH_NEEDS_COMMENT;
	265	DIAG_IGNORE_Os_NEEDS_COMMENT (7, "-Wmaybe-uninitialized");
9a1f6754 UD	266	/* Handling of Unicode 3.1 TAG characters. Unicode recommends
9a1f6754 UD	267	"If language codes are not relevant to the particular processing
e438a468 UD	268	operation, then they should be ignored." This macro is usually
e438a468 UD	269	called right before STANDARD_TO_LOOP_ERR_HANDLER (Incr). */
9a1f6754 UD	270	#define UNICODE_TAG_HANDLER(Character, Incr) \
	271	{ \
	272	/* TAG characters are those in the range U+E0000..U+E007F. */ \
	273	if (((Character) >> 7) == (0xe0000 >> 7)) \
	274	{ \
	275	inptr += Incr; \
	276	continue; \
	277	} \
	278	}
6900d2ca	279	DIAG_POP_NEEDS_COMMENT;
9a1f6754 UD	280
9a1f6754 UD	281
8619129f UD	282	/* The function returns the status, as defined in gconv.h. */
8619129f UD	283	static inline int
dd9423a6	284	__attribute ((always_inline))
55985355 UD	285	FCTNAME (LOOPFCT) (struct __gconv_step *step,
	286	struct __gconv_step_data *step_data,
	287	const unsigned char *inptrp, const unsigned char inend,
17427edd	288	unsigned char *outptrp, const unsigned char outend,
38677ace	289	size_t *irreversible EXTRA_LOOP_DECLS)
8619129f	290	{
55985355 UD	291	#ifdef LOOP_NEED_STATE
	292	mbstate_t *state = step_data->__statep;
	293	#endif
	294	#ifdef LOOP_NEED_FLAGS
	295	int flags = step_data->__flags;
	296	#endif
	297	#ifdef LOOP_NEED_DATA
	298	void *data = step->__data;
	299	#endif
	300	int result = __GCONV_EMPTY_INPUT;
8619129f UD	301	const unsigned char inptr = inptrp;
8619129f UD	302	unsigned char outptr = outptrp;
8619129f	303
66175fa8 UD	304	#ifdef INIT_PARAMS
	305	INIT_PARAMS;
	306	#endif
	307
55985355	308	while (inptr != inend)
8619129f	309	{
55985355	310	/* `if' cases for MIN_NEEDED_OUTPUT ==/!= 1 is made to help the
ca3c0135	311	compiler generating better code. They will be optimized away
55985355	312	since MIN_NEEDED_OUTPUT is always a constant. */
eb9dc2a2 UD	313	if (MIN_NEEDED_INPUT > 1
	314	&& __builtin_expect (inptr + MIN_NEEDED_INPUT > inend, 0))
	315	{
	316	/* We don't have enough input for another complete input
	317	character. */
	318	result = __GCONV_INCOMPLETE_INPUT;
	319	break;
	320	}
55985355 UD	321	if ((MIN_NEEDED_OUTPUT != 1
	322	&& __builtin_expect (outptr + MIN_NEEDED_OUTPUT > outend, 0))
	323	\|\| (MIN_NEEDED_OUTPUT == 1
	324	&& __builtin_expect (outptr >= outend, 0)))
	325	{
	326	/* Overflow in the output buffer. */
	327	result = __GCONV_FULL_OUTPUT;
	328	break;
	329	}
55985355 UD	330
	331	/* Here comes the body the user provides. It can stop with
	332	RESULT set to GCONV_INCOMPLETE_INPUT (if the size of the
	333	input characters vary in size), GCONV_ILLEGAL_INPUT, or
	334	GCONV_FULL_OUTPUT (if the output characters vary in size). */
	335	BODY
8619129f UD	336	}
8619129f UD	337
8619129f UD	338	/* Update the pointers pointed to by the parameters. */
	339	*inptrp = inptr;
	340	*outptrp = outptr;
66175fa8	341	UPDATE_PARAMS;
8619129f UD	342
	343	return result;
	344	}
	345
	346
b02b4774 UD	347	/* Include the file a second time to define the function to handle
b02b4774 UD	348	unaligned access. */
27822ce6	349	#if !defined DEFINE_UNALIGNED && !_STRING_ARCH_unaligned \
4a0de63b UD	350	&& MIN_NEEDED_INPUT != 1 && MAX_NEEDED_INPUT % MIN_NEEDED_INPUT == 0 \
4a0de63b UD	351	&& MIN_NEEDED_OUTPUT != 1 && MAX_NEEDED_OUTPUT % MIN_NEEDED_OUTPUT == 0
fd1b5c0f UD	352	# undef get16
	353	# undef get32
	354	# undef put16
	355	# undef put32
	356	# undef unaligned
	357
b35e58e4 UD	358	# define DEFINE_UNALIGNED
	359	# include "loop.c"
	360	# undef DEFINE_UNALIGNED
32bead5b WN	361	#else
	362	# if MAX_NEEDED_INPUT > 1
	363	# define SINGLE(fct) SINGLE2 (fct)
	364	# define SINGLE2(fct) fct##_single
fd1b5c0f	365	static inline int
dd9423a6	366	__attribute ((always_inline))
55985355 UD	367	SINGLE(LOOPFCT) (struct __gconv_step *step,
	368	struct __gconv_step_data *step_data,
	369	const unsigned char *inptrp, const unsigned char inend,
fd1b5c0f	370	unsigned char *outptrp, unsigned char outend,
55985355	371	size_t *irreversible EXTRA_LOOP_DECLS)
fd1b5c0f	372	{
55985355	373	mbstate_t *state = step_data->__statep;
32bead5b	374	# ifdef LOOP_NEED_FLAGS
55985355	375	int flags = step_data->__flags;
32bead5b WN	376	# endif
32bead5b WN	377	# ifdef LOOP_NEED_DATA
55985355	378	void *data = step->__data;
32bead5b	379	# endif
fd1b5c0f UD	380	int result = __GCONV_OK;
	381	unsigned char bytebuf[MAX_NEEDED_INPUT];
	382	const unsigned char inptr = inptrp;
	383	unsigned char outptr = outptrp;
	384	size_t inlen;
	385
32bead5b	386	# ifdef INIT_PARAMS
fd1b5c0f	387	INIT_PARAMS;
32bead5b	388	# endif
fd1b5c0f	389
32bead5b	390	# ifdef UNPACK_BYTES
fd1b5c0f	391	UNPACK_BYTES
32bead5b	392	# else
fd1b5c0f	393	/* Add the bytes from the state to the input buffer. */
5e0d0300	394	assert ((state->__count & 7) <= sizeof (state->__value));
17427edd	395	for (inlen = 0; inlen < (size_t) (state->__count & 7); ++inlen)
fd1b5c0f	396	bytebuf[inlen] = state->__value.__wchb[inlen];
32bead5b	397	# endif
fd1b5c0f UD	398
fd1b5c0f UD	399	/* Are there enough bytes in the input buffer? */
0656e90e UD	400	if (MIN_NEEDED_INPUT > 1
0656e90e UD	401	&& __builtin_expect (inptr + (MIN_NEEDED_INPUT - inlen) > inend, 0))
fd1b5c0f	402	{
fd1b5c0f	403	*inptrp = inend;
32bead5b	404	# ifdef STORE_REST
5fe8e359 AK	405
	406	/* Building with -O3 GCC emits a `array subscript is above array
	407	bounds' warning. GCC BZ #64739 has been opened for this. */
	408	DIAG_PUSH_NEEDS_COMMENT;
	409	DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Warray-bounds");
1af4e298 UD	410	while (inptr < inend)
1af4e298 UD	411	bytebuf[inlen++] = *inptr++;
5fe8e359	412	DIAG_POP_NEEDS_COMMENT;
1af4e298	413
fd1b5c0f UD	414	inptr = bytebuf;
	415	inptrp = &inptr;
	416	inend = &bytebuf[inlen];
	417
	418	STORE_REST
32bead5b	419	# else
fd1b5c0f UD	420	/* We don't have enough input for another complete input
fd1b5c0f UD	421	character. */
08538f36 SL	422	size_t inlen_after = inlen + (inend - inptr);
	423	assert (inlen_after <= sizeof (state->__value.__wchb));
	424	for (; inlen < inlen_after; inlen++)
	425	state->__value.__wchb[inlen] = *inptr++;
32bead5b	426	# endif
fd1b5c0f UD	427
	428	return __GCONV_INCOMPLETE_INPUT;
	429	}
	430
	431	/* Enough space in output buffer. */
	432	if ((MIN_NEEDED_OUTPUT != 1 && outptr + MIN_NEEDED_OUTPUT > outend)
	433	\|\| (MIN_NEEDED_OUTPUT == 1 && outptr >= outend))
	434	/* Overflow in the output buffer. */
	435	return __GCONV_FULL_OUTPUT;
	436
	437	/* Now add characters from the normal input buffer. */
c8126360 JM	438	if (inlen >= MAX_NEEDED_INPUT)
	439	/* Avoid a -Wstringop-overflow= warning when this loop is
	440	unrolled. The compiler cannot otherwise see that this is
	441	unreachable because it depends on (state->__count & 7) not
	442	being too large after a previous conversion step. */
	443	__builtin_unreachable ();
fd1b5c0f UD	444	do
fd1b5c0f UD	445	bytebuf[inlen++] = *inptr++;
316518d6	446	while (inlen < MAX_NEEDED_INPUT && inptr < inend);
fd1b5c0f UD	447
fd1b5c0f UD	448	inptr = bytebuf;
316518d6	449	inend = &bytebuf[inlen];
55985355	450
fd1b5c0f UD	451	do
	452	{
	453	BODY
	454	}
	455	while (0);
	456
316518d6 UD	457	/* Now we either have produced an output character and consumed all the
	458	bytes from the state and at least one more, or the character is still
	459	incomplete, or we have some other error (like illegal input character,
	460	no space in output buffer). */
a1ffb40e	461	if (__glibc_likely (inptr != bytebuf))
fd1b5c0f	462	{
316518d6	463	/* We found a new character. */
fd1b5c0f UD	464	assert (inptr - bytebuf > (state->__count & 7));
	465
	466	*inptrp += inptr - bytebuf - (state->__count & 7);
	467	*outptrp = outptr;
	468
316518d6 UD	469	result = __GCONV_OK;
316518d6 UD	470
fd1b5c0f	471	/* Clear the state buffer. */
32bead5b	472	# ifdef CLEAR_STATE
41f112ad	473	CLEAR_STATE;
32bead5b	474	# else
fd1b5c0f	475	state->__count &= ~7;
32bead5b	476	# endif
fd1b5c0f	477	}
316518d6 UD	478	else if (result == __GCONV_INCOMPLETE_INPUT)
	479	{
	480	/* This can only happen if we have less than MAX_NEEDED_INPUT bytes
	481	available. */
	482	assert (inend != &bytebuf[MAX_NEEDED_INPUT]);
	483
	484	*inptrp += inend - bytebuf - (state->__count & 7);
32bead5b	485	# ifdef STORE_REST
316518d6 UD	486	inptrp = &inptr;
	487
	488	STORE_REST
32bead5b	489	# else
316518d6 UD	490	/* We don't have enough input for another complete input
316518d6 UD	491	character. */
5512d89b	492	assert (inend - inptr > (state->__count & ~7));
08538f36	493	assert (inend - inptr <= sizeof (state->__value.__wchb));
5512d89b	494	state->__count = (state->__count & ~7) \| (inend - inptr);
08538f36 SL	495	for (inlen = 0; inlen < inend - inptr; inlen++)
	496	state->__value.__wchb[inlen] = inptr[inlen];
	497	inptr = inend;
32bead5b	498	# endif
316518d6	499	}
fd1b5c0f UD	500
	501	return result;
	502	}
32bead5b WN	503	# undef SINGLE
	504	# undef SINGLE2
	505	# endif
fd1b5c0f UD	506
fd1b5c0f UD	507
32bead5b	508	# ifdef ONEBYTE_BODY
f9ad060c UD	509	/* Define the shortcut function for btowc. */
	510	static wint_t
	511	gconv_btowc (struct __gconv_step *step, unsigned char c)
	512	ONEBYTE_BODY
32bead5b WN	513	# define FROM_ONEBYTE gconv_btowc
32bead5b WN	514	# endif
f9ad060c	515
32bead5b	516	#endif
f9ad060c	517
8619129f UD	518	/* We remove the macro definitions so that we can include this file again
	519	for the definition of another function. */
	520	#undef MIN_NEEDED_INPUT
	521	#undef MAX_NEEDED_INPUT
	522	#undef MIN_NEEDED_OUTPUT
	523	#undef MAX_NEEDED_OUTPUT
	524	#undef LOOPFCT
8619129f UD	525	#undef BODY
8619129f UD	526	#undef LOOPFCT
28f1c862	527	#undef EXTRA_LOOP_DECLS
66175fa8 UD	528	#undef INIT_PARAMS
66175fa8 UD	529	#undef UPDATE_PARAMS
4b1b449d	530	#undef REINIT_PARAMS
f9ad060c	531	#undef ONEBYTE_BODY
55985355	532	#undef UNPACK_BYTES
41f112ad	533	#undef CLEAR_STATE
55985355 UD	534	#undef LOOP_NEED_STATE
	535	#undef LOOP_NEED_FLAGS
	536	#undef LOOP_NEED_DATA
fd1b5c0f UD	537	#undef get16
	538	#undef get32
	539	#undef put16
	540	#undef put32
	541	#undef unaligned