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

/* Conversion loop frame work.
   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 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
   <http://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 <gconv.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.  */
      while (inptr < inend)
	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.  */
  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));
      state->__count = (state->__count & ~7) | (inend - inptr);
      inlen = 0;
      while (inptr < inend)
	state->__value.__wchb[inlen++] = *inptr++;
#  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.
04277e02	2	Copyright (C) 1998-2019 Free Software Foundation, Inc.
8619129f UD	3	This file is part of the GNU C Library.
	4	Contributed by Ulrich Drepper <drepper@cygnus.com>, 1998.
	5
	6	The GNU C Library is free software; you can redistribute it and/or
41bdb6e2 AJ	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.
8619129f UD	10
	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
41bdb6e2	14	Lesser General Public License for more details.
8619129f	15
41bdb6e2	16	You should have received a copy of the GNU Lesser General Public
59ba27a6 PE	17	License along with the GNU C Library; if not, see
59ba27a6 PE	18	<http://www.gnu.org/licenses/>. */
8619129f UD	19
	20	/* This file provides a frame for the reader loop in all conversion modules.
	21	The actual code must (of course) be provided in the actual module source
	22	code but certain actions can be written down generically, with some
	23	customization options which are these:
	24
	25	MIN_NEEDED_INPUT minimal number of input bytes needed for the next
	26	conversion.
	27	MIN_NEEDED_OUTPUT minimal number of bytes produced by the next round
	28	of conversion.
	29
	30	MAX_NEEDED_INPUT you guess it, this is the maximal number of input
	31	bytes needed. It defaults to MIN_NEEDED_INPUT
	32	MAX_NEEDED_OUTPUT likewise for output bytes.
	33
8619129f UD	34	LOOPFCT name of the function created. If not specified
	35	the name is `loop' but this prevents the use
	36	of multiple functions in the same file.
	37
8619129f UD	38	BODY this is supposed to expand to the body of the loop.
8619129f UD	39	The user must provide this.
28f1c862	40
382466e0	41	EXTRA_LOOP_DECLS extra arguments passed from conversion loop call.
66175fa8 UD	42
	43	INIT_PARAMS code to define and initialize variables from params.
	44	UPDATE_PARAMS code to store result in params.
f9ad060c UD	45
	46	ONEBYTE_BODY body of the specialized conversion function for a
	47	single byte from the current character set to INTERNAL.
8619129f UD	48	*/
8619129f UD	49
fd1b5c0f	50	#include <assert.h>
b35e58e4	51	#include <endian.h>
8619129f	52	#include <gconv.h>
b35e58e4 UD	53	#include <stdint.h>
b35e58e4 UD	54	#include <string.h>
d64b6ad0	55	#include <wchar.h>
8619129f UD	56	#include <sys/param.h> /* For MIN. */
	57	#define __need_size_t
	58	#include <stddef.h>
9090848d	59	#include <libc-diag.h>
8619129f	60
b35e58e4 UD	61	/* We have to provide support for machines which are not able to handled
	62	unaligned memory accesses. Some of the character encodings have
	63	representations with a fixed width of 2 or 4 bytes. But if we cannot
	64	access unaligned memory we still have to read byte-wise. */
	65	#undef FCTNAME2
27822ce6	66	#if _STRING_ARCH_unaligned \|\| !defined DEFINE_UNALIGNED
b35e58e4	67	/* We can handle unaligned memory access. */
a784e502 UD	68	# define get16(addr) ((const uint16_t ) (addr))
a784e502 UD	69	# define get32(addr) ((const uint32_t ) (addr))
b35e58e4 UD	70
	71	/* We need no special support for writing values either. */
	72	# define put16(addr, val) ((uint16_t ) (addr)) = (val)
	73	# define put32(addr, val) ((uint32_t ) (addr)) = (val)
	74
	75	# define FCTNAME2(name) name
	76	#else
	77	/* Distinguish between big endian and little endian. */
	78	# if __BYTE_ORDER == __LITTLE_ENDIAN
	79	# define get16(addr) \
a784e502 UD	80	(((const unsigned char *) (addr))[1] << 8 \
a784e502 UD	81	\| ((const unsigned char *) (addr))[0])
b35e58e4	82	# define get32(addr) \
a784e502 UD	83	(((((const unsigned char *) (addr))[3] << 8 \
	84	\| ((const unsigned char *) (addr))[2]) << 8 \
	85	\| ((const unsigned char *) (addr))[1]) << 8 \
	86	\| ((const unsigned char *) (addr))[0])
b35e58e4	87
cb2c5501	88	# define put16(addr, val) \
b35e58e4	89	({ uint16_t __val = (val); \
cb2c5501 UD	90	((unsigned char *) (addr))[0] = __val; \
cb2c5501 UD	91	((unsigned char *) (addr))[1] = __val >> 8; \
b35e58e4	92	(void) 0; })
cb2c5501 UD	93	# define put32(addr, val) \
	94	({ uint32_t __val = (val); \
	95	((unsigned char *) (addr))[0] = __val; \
b35e58e4	96	__val >>= 8; \
cb2c5501	97	((unsigned char *) (addr))[1] = __val; \
b35e58e4	98	__val >>= 8; \
cb2c5501	99	((unsigned char *) (addr))[2] = __val; \
b35e58e4	100	__val >>= 8; \
cb2c5501	101	((unsigned char *) (addr))[3] = __val; \
b35e58e4 UD	102	(void) 0; })
	103	# else
	104	# define get16(addr) \
a784e502 UD	105	(((const unsigned char *) (addr))[0] << 8 \
a784e502 UD	106	\| ((const unsigned char *) (addr))[1])
b35e58e4	107	# define get32(addr) \
a784e502 UD	108	(((((const unsigned char *) (addr))[0] << 8 \
	109	\| ((const unsigned char *) (addr))[1]) << 8 \
	110	\| ((const unsigned char *) (addr))[2]) << 8 \
	111	\| ((const unsigned char *) (addr))[3])
b35e58e4	112
cb2c5501	113	# define put16(addr, val) \
b35e58e4	114	({ uint16_t __val = (val); \
cb2c5501	115	((unsigned char *) (addr))[1] = __val; \
f7ccf2fc	116	((unsigned char *) (addr))[0] = __val >> 8; \
b35e58e4	117	(void) 0; })
cb2c5501 UD	118	# define put32(addr, val) \
	119	({ uint32_t __val = (val); \
	120	((unsigned char *) (addr))[3] = __val; \
b35e58e4	121	__val >>= 8; \
cb2c5501	122	((unsigned char *) (addr))[2] = __val; \
b35e58e4	123	__val >>= 8; \
cb2c5501	124	((unsigned char *) (addr))[1] = __val; \
b35e58e4	125	__val >>= 8; \
cb2c5501	126	((unsigned char *) (addr))[0] = __val; \
b35e58e4 UD	127	(void) 0; })
	128	# endif
	129
	130	# define FCTNAME2(name) name##_unaligned
	131	#endif
	132	#define FCTNAME(name) FCTNAME2(name)
	133
	134
8619129f UD	135	/* We need at least one byte for the next round. */
8619129f UD	136	#ifndef MIN_NEEDED_INPUT
5aa8ff62	137	# error "MIN_NEEDED_INPUT definition missing"
4a0de63b UD	138	#elif MIN_NEEDED_INPUT < 1
4a0de63b UD	139	# error "MIN_NEEDED_INPUT must be >= 1"
8619129f UD	140	#endif
	141
	142	/* Let's see how many bytes we produce. */
	143	#ifndef MAX_NEEDED_INPUT
	144	# define MAX_NEEDED_INPUT MIN_NEEDED_INPUT
	145	#endif
	146
	147	/* We produce at least one byte in the next round. */
	148	#ifndef MIN_NEEDED_OUTPUT
5aa8ff62	149	# error "MIN_NEEDED_OUTPUT definition missing"
c0a0f9a3 UD	150	#elif MIN_NEEDED_OUTPUT < 1
c0a0f9a3 UD	151	# error "MIN_NEEDED_OUTPUT must be >= 1"
8619129f UD	152	#endif
	153
	154	/* Let's see how many bytes we produce. */
	155	#ifndef MAX_NEEDED_OUTPUT
	156	# define MAX_NEEDED_OUTPUT MIN_NEEDED_OUTPUT
	157	#endif
	158
	159	/* Default name for the function. */
	160	#ifndef LOOPFCT
	161	# define LOOPFCT loop
	162	#endif
	163
	164	/* Make sure we have a loop body. */
	165	#ifndef BODY
	166	# error "Definition of BODY missing for function" LOOPFCT
	167	#endif
	168
8619129f	169
28f1c862 UD	170	/* If no arguments have to passed to the loop function define the macro
	171	as empty. */
	172	#ifndef EXTRA_LOOP_DECLS
	173	# define EXTRA_LOOP_DECLS
	174	#endif
	175
4b1b449d UD	176	/* Allow using UPDATE_PARAMS in macros where #ifdef UPDATE_PARAMS test
	177	isn't possible. */
	178	#ifndef UPDATE_PARAMS
	179	# define UPDATE_PARAMS do { } while (0)
	180	#endif
	181	#ifndef REINIT_PARAMS
	182	# define REINIT_PARAMS do { } while (0)
	183	#endif
	184
28f1c862	185
85830c4c UD	186	/* To make it easier for the writers of the modules, we define a macro
85830c4c UD	187	to test whether we have to ignore errors. */
b572c2da UD	188	#define ignore_errors_p() \
b572c2da UD	189	(irreversible != NULL && (flags & __GCONV_IGNORE_ERRORS))
85830c4c UD	190
85830c4c UD	191
e438a468 UD	192	/* Error handling for the FROM_LOOP direction, with ignoring of errors.
	193	Note that we cannot use the do while (0) trick since `break' and
	194	`continue' must reach certain points. */
	195	#define STANDARD_FROM_LOOP_ERR_HANDLER(Incr) \
	196	{ \
	197	result = __GCONV_ILLEGAL_INPUT; \
	198	\
	199	if (! ignore_errors_p ()) \
	200	break; \
	201	\
	202	/* We ignore the invalid input byte sequence. */ \
	203	inptr += (Incr); \
	204	++*irreversible; \
	205	/* But we keep result == __GCONV_ILLEGAL_INPUT, because of the constraint \
	206	that "iconv -c" must give the same exitcode as "iconv". */ \
	207	continue; \
	208	}
	209
	210	/* Error handling for the TO_LOOP direction, with use of transliteration/
	211	transcription functions and ignoring of errors. Note that we cannot use
	212	the do while (0) trick since `break' and `continue' must reach certain
	213	points. */
	214	#define STANDARD_TO_LOOP_ERR_HANDLER(Incr) \
d6204268	215	{ \
d6204268	216	result = __GCONV_ILLEGAL_INPUT; \
b572c2da UD	217	\
	218	if (irreversible == NULL) \
	219	/* This means we are in call from __gconv_transliterate. In this \
	220	case we are not doing any error recovery outself. */ \
	221	break; \
	222	\
4b1b449d UD	223	/* If needed, flush any conversion state, so that __gconv_transliterate \
	224	starts with current shift state. */ \
	225	UPDATE_PARAMS; \
	226	\
d6204268	227	/* First try the transliteration methods. */ \
ba7b4d29 FW	228	if ((step_data->__flags & __GCONV_TRANSLIT) != 0) \
	229	result = __gconv_transliterate \
	230	(step, step_data, *inptrp, \
	231	&inptr, inend, &outptr, irreversible); \
4b1b449d UD	232	\
	233	REINIT_PARAMS; \
	234	\
7888313d	235	/* If any of them recognized the input continue with the loop. */ \
d6204268	236	if (result != __GCONV_ILLEGAL_INPUT) \
f2a8406a	237	{ \
a1ffb40e	238	if (__glibc_unlikely (result == __GCONV_FULL_OUTPUT)) \
f2a8406a UD	239	break; \
	240	\
	241	continue; \
	242	} \
d6204268 UD	243	\
	244	/* Next see whether we have to ignore the error. If not, stop. */ \
	245	if (! ignore_errors_p ()) \
	246	break; \
b572c2da	247	\
d6204268 UD	248	/* When we come here it means we ignore the character. */ \
	249	++*irreversible; \
	250	inptr += Incr; \
e438a468 UD	251	/* But we keep result == __GCONV_ILLEGAL_INPUT, because of the constraint \
e438a468 UD	252	that "iconv -c" must give the same exitcode as "iconv". */ \
d6204268 UD	253	continue; \
	254	}
	255
	256
6900d2ca JM	257	/* With GCC 7 when compiling with -Os for 32-bit s390 the compiler
	258	warns that the variable 'ch', in the definition of BODY in
	259	sysdeps/s390/multiarch/8bit-generic.c, may be used uninitialized in
	260	the call to UNICODE_TAG_HANDLER in that macro. This variable is
	261	actually always initialized before use, in the prior loop if INDEX
	262	is nonzero and in the following 'if' if INDEX is zero. That code
	263	has a comment referencing this diagnostic disabling; updates in one
	264	place may require updates in the other. */
	265	DIAG_PUSH_NEEDS_COMMENT;
	266	DIAG_IGNORE_Os_NEEDS_COMMENT (7, "-Wmaybe-uninitialized");
9a1f6754 UD	267	/* Handling of Unicode 3.1 TAG characters. Unicode recommends
9a1f6754 UD	268	"If language codes are not relevant to the particular processing
e438a468 UD	269	operation, then they should be ignored." This macro is usually
e438a468 UD	270	called right before STANDARD_TO_LOOP_ERR_HANDLER (Incr). */
9a1f6754 UD	271	#define UNICODE_TAG_HANDLER(Character, Incr) \
	272	{ \
	273	/* TAG characters are those in the range U+E0000..U+E007F. */ \
	274	if (((Character) >> 7) == (0xe0000 >> 7)) \
	275	{ \
	276	inptr += Incr; \
	277	continue; \
	278	} \
	279	}
6900d2ca	280	DIAG_POP_NEEDS_COMMENT;
9a1f6754 UD	281
9a1f6754 UD	282
8619129f UD	283	/* The function returns the status, as defined in gconv.h. */
8619129f UD	284	static inline int
dd9423a6	285	__attribute ((always_inline))
55985355 UD	286	FCTNAME (LOOPFCT) (struct __gconv_step *step,
	287	struct __gconv_step_data *step_data,
	288	const unsigned char *inptrp, const unsigned char inend,
17427edd	289	unsigned char *outptrp, const unsigned char outend,
38677ace	290	size_t *irreversible EXTRA_LOOP_DECLS)
8619129f	291	{
55985355 UD	292	#ifdef LOOP_NEED_STATE
	293	mbstate_t *state = step_data->__statep;
	294	#endif
	295	#ifdef LOOP_NEED_FLAGS
	296	int flags = step_data->__flags;
	297	#endif
	298	#ifdef LOOP_NEED_DATA
	299	void *data = step->__data;
	300	#endif
	301	int result = __GCONV_EMPTY_INPUT;
8619129f UD	302	const unsigned char inptr = inptrp;
8619129f UD	303	unsigned char outptr = outptrp;
8619129f	304
66175fa8 UD	305	#ifdef INIT_PARAMS
	306	INIT_PARAMS;
	307	#endif
	308
55985355	309	while (inptr != inend)
8619129f	310	{
55985355	311	/* `if' cases for MIN_NEEDED_OUTPUT ==/!= 1 is made to help the
ca3c0135	312	compiler generating better code. They will be optimized away
55985355	313	since MIN_NEEDED_OUTPUT is always a constant. */
eb9dc2a2 UD	314	if (MIN_NEEDED_INPUT > 1
	315	&& __builtin_expect (inptr + MIN_NEEDED_INPUT > inend, 0))
	316	{
	317	/* We don't have enough input for another complete input
	318	character. */
	319	result = __GCONV_INCOMPLETE_INPUT;
	320	break;
	321	}
55985355 UD	322	if ((MIN_NEEDED_OUTPUT != 1
	323	&& __builtin_expect (outptr + MIN_NEEDED_OUTPUT > outend, 0))
	324	\|\| (MIN_NEEDED_OUTPUT == 1
	325	&& __builtin_expect (outptr >= outend, 0)))
	326	{
	327	/* Overflow in the output buffer. */
	328	result = __GCONV_FULL_OUTPUT;
	329	break;
	330	}
55985355 UD	331
	332	/* Here comes the body the user provides. It can stop with
	333	RESULT set to GCONV_INCOMPLETE_INPUT (if the size of the
	334	input characters vary in size), GCONV_ILLEGAL_INPUT, or
	335	GCONV_FULL_OUTPUT (if the output characters vary in size). */
	336	BODY
8619129f UD	337	}
8619129f UD	338
8619129f UD	339	/* Update the pointers pointed to by the parameters. */
	340	*inptrp = inptr;
	341	*outptrp = outptr;
66175fa8	342	UPDATE_PARAMS;
8619129f UD	343
	344	return result;
	345	}
	346
	347
b02b4774 UD	348	/* Include the file a second time to define the function to handle
b02b4774 UD	349	unaligned access. */
27822ce6	350	#if !defined DEFINE_UNALIGNED && !_STRING_ARCH_unaligned \
4a0de63b UD	351	&& MIN_NEEDED_INPUT != 1 && MAX_NEEDED_INPUT % MIN_NEEDED_INPUT == 0 \
4a0de63b UD	352	&& MIN_NEEDED_OUTPUT != 1 && MAX_NEEDED_OUTPUT % MIN_NEEDED_OUTPUT == 0
fd1b5c0f UD	353	# undef get16
	354	# undef get32
	355	# undef put16
	356	# undef put32
	357	# undef unaligned
	358
b35e58e4 UD	359	# define DEFINE_UNALIGNED
	360	# include "loop.c"
	361	# undef DEFINE_UNALIGNED
32bead5b WN	362	#else
	363	# if MAX_NEEDED_INPUT > 1
	364	# define SINGLE(fct) SINGLE2 (fct)
	365	# define SINGLE2(fct) fct##_single
fd1b5c0f	366	static inline int
dd9423a6	367	__attribute ((always_inline))
55985355 UD	368	SINGLE(LOOPFCT) (struct __gconv_step *step,
	369	struct __gconv_step_data *step_data,
	370	const unsigned char *inptrp, const unsigned char inend,
fd1b5c0f	371	unsigned char *outptrp, unsigned char outend,
55985355	372	size_t *irreversible EXTRA_LOOP_DECLS)
fd1b5c0f	373	{
55985355	374	mbstate_t *state = step_data->__statep;
32bead5b	375	# ifdef LOOP_NEED_FLAGS
55985355	376	int flags = step_data->__flags;
32bead5b WN	377	# endif
32bead5b WN	378	# ifdef LOOP_NEED_DATA
55985355	379	void *data = step->__data;
32bead5b	380	# endif
fd1b5c0f UD	381	int result = __GCONV_OK;
	382	unsigned char bytebuf[MAX_NEEDED_INPUT];
	383	const unsigned char inptr = inptrp;
	384	unsigned char outptr = outptrp;
	385	size_t inlen;
	386
32bead5b	387	# ifdef INIT_PARAMS
fd1b5c0f	388	INIT_PARAMS;
32bead5b	389	# endif
fd1b5c0f	390
32bead5b	391	# ifdef UNPACK_BYTES
fd1b5c0f	392	UNPACK_BYTES
32bead5b	393	# else
fd1b5c0f	394	/* Add the bytes from the state to the input buffer. */
5e0d0300	395	assert ((state->__count & 7) <= sizeof (state->__value));
17427edd	396	for (inlen = 0; inlen < (size_t) (state->__count & 7); ++inlen)
fd1b5c0f	397	bytebuf[inlen] = state->__value.__wchb[inlen];
32bead5b	398	# endif
fd1b5c0f UD	399
fd1b5c0f UD	400	/* Are there enough bytes in the input buffer? */
0656e90e UD	401	if (MIN_NEEDED_INPUT > 1
0656e90e UD	402	&& __builtin_expect (inptr + (MIN_NEEDED_INPUT - inlen) > inend, 0))
fd1b5c0f	403	{
fd1b5c0f	404	*inptrp = inend;
32bead5b	405	# ifdef STORE_REST
5fe8e359 AK	406
	407	/* Building with -O3 GCC emits a `array subscript is above array
	408	bounds' warning. GCC BZ #64739 has been opened for this. */
	409	DIAG_PUSH_NEEDS_COMMENT;
	410	DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Warray-bounds");
1af4e298 UD	411	while (inptr < inend)
1af4e298 UD	412	bytebuf[inlen++] = *inptr++;
5fe8e359	413	DIAG_POP_NEEDS_COMMENT;
1af4e298	414
fd1b5c0f UD	415	inptr = bytebuf;
	416	inptrp = &inptr;
	417	inend = &bytebuf[inlen];
	418
	419	STORE_REST
32bead5b	420	# else
fd1b5c0f UD	421	/* We don't have enough input for another complete input
	422	character. */
	423	while (inptr < inend)
	424	state->__value.__wchb[inlen++] = *inptr++;
32bead5b	425	# endif
fd1b5c0f UD	426
	427	return __GCONV_INCOMPLETE_INPUT;
	428	}
	429
	430	/* Enough space in output buffer. */
	431	if ((MIN_NEEDED_OUTPUT != 1 && outptr + MIN_NEEDED_OUTPUT > outend)
	432	\|\| (MIN_NEEDED_OUTPUT == 1 && outptr >= outend))
	433	/* Overflow in the output buffer. */
	434	return __GCONV_FULL_OUTPUT;
	435
	436	/* Now add characters from the normal input buffer. */
	437	do
	438	bytebuf[inlen++] = *inptr++;
316518d6	439	while (inlen < MAX_NEEDED_INPUT && inptr < inend);
fd1b5c0f UD	440
fd1b5c0f UD	441	inptr = bytebuf;
316518d6	442	inend = &bytebuf[inlen];
55985355	443
fd1b5c0f UD	444	do
	445	{
	446	BODY
	447	}
	448	while (0);
	449
316518d6 UD	450	/* Now we either have produced an output character and consumed all the
	451	bytes from the state and at least one more, or the character is still
	452	incomplete, or we have some other error (like illegal input character,
	453	no space in output buffer). */
a1ffb40e	454	if (__glibc_likely (inptr != bytebuf))
fd1b5c0f	455	{
316518d6	456	/* We found a new character. */
fd1b5c0f UD	457	assert (inptr - bytebuf > (state->__count & 7));
	458
	459	*inptrp += inptr - bytebuf - (state->__count & 7);
	460	*outptrp = outptr;
	461
316518d6 UD	462	result = __GCONV_OK;
316518d6 UD	463
fd1b5c0f	464	/* Clear the state buffer. */
32bead5b	465	# ifdef CLEAR_STATE
41f112ad	466	CLEAR_STATE;
32bead5b	467	# else
fd1b5c0f	468	state->__count &= ~7;
32bead5b	469	# endif
fd1b5c0f	470	}
316518d6 UD	471	else if (result == __GCONV_INCOMPLETE_INPUT)
	472	{
	473	/* This can only happen if we have less than MAX_NEEDED_INPUT bytes
	474	available. */
	475	assert (inend != &bytebuf[MAX_NEEDED_INPUT]);
	476
	477	*inptrp += inend - bytebuf - (state->__count & 7);
32bead5b	478	# ifdef STORE_REST
316518d6 UD	479	inptrp = &inptr;
	480
	481	STORE_REST
32bead5b	482	# else
316518d6 UD	483	/* We don't have enough input for another complete input
316518d6 UD	484	character. */
5512d89b	485	assert (inend - inptr > (state->__count & ~7));
5e0d0300	486	assert (inend - inptr <= sizeof (state->__value));
5512d89b UD	487	state->__count = (state->__count & ~7) \| (inend - inptr);
5512d89b UD	488	inlen = 0;
316518d6 UD	489	while (inptr < inend)
316518d6 UD	490	state->__value.__wchb[inlen++] = *inptr++;
32bead5b	491	# endif
316518d6	492	}
fd1b5c0f UD	493
	494	return result;
	495	}
32bead5b WN	496	# undef SINGLE
	497	# undef SINGLE2
	498	# endif
fd1b5c0f UD	499
fd1b5c0f UD	500
32bead5b	501	# ifdef ONEBYTE_BODY
f9ad060c UD	502	/* Define the shortcut function for btowc. */
	503	static wint_t
	504	gconv_btowc (struct __gconv_step *step, unsigned char c)
	505	ONEBYTE_BODY
32bead5b WN	506	# define FROM_ONEBYTE gconv_btowc
32bead5b WN	507	# endif
f9ad060c	508
32bead5b	509	#endif
f9ad060c	510
8619129f UD	511	/* We remove the macro definitions so that we can include this file again
	512	for the definition of another function. */
	513	#undef MIN_NEEDED_INPUT
	514	#undef MAX_NEEDED_INPUT
	515	#undef MIN_NEEDED_OUTPUT
	516	#undef MAX_NEEDED_OUTPUT
	517	#undef LOOPFCT
8619129f UD	518	#undef BODY
8619129f UD	519	#undef LOOPFCT
28f1c862	520	#undef EXTRA_LOOP_DECLS
66175fa8 UD	521	#undef INIT_PARAMS
66175fa8 UD	522	#undef UPDATE_PARAMS
4b1b449d	523	#undef REINIT_PARAMS
f9ad060c	524	#undef ONEBYTE_BODY
55985355	525	#undef UNPACK_BYTES
41f112ad	526	#undef CLEAR_STATE
55985355 UD	527	#undef LOOP_NEED_STATE
	528	#undef LOOP_NEED_FLAGS
	529	#undef LOOP_NEED_DATA
fd1b5c0f UD	530	#undef get16
	531	#undef get32
	532	#undef put16
	533	#undef put32
	534	#undef unaligned