2007-05-18 H.J. Lu <hongjiu.lu@intel.com>

[official-gcc.git] / libcpp / charset.c

/* CPP Library - charsets
   Copyright (C) 1998, 1999, 2000, 2001, 2002, 2003, 2004, 2006
   Free Software Foundation, Inc.

   Broken out of c-lex.c Apr 2003, adding valid C99 UCN ranges.

This program is free software; you can redistribute it and/or modify it
under the terms of the GNU General Public License as published by the
Free Software Foundation; either version 2, or (at your option) any
later version.

This program 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 General Public License for more details.

You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.  */

#include "config.h"
#include "system.h"
#include "cpplib.h"
#include "internal.h"

/* Character set handling for C-family languages.

   Terminological note: In what follows, "charset" or "character set"
   will be taken to mean both an abstract set of characters and an
   encoding for that set.

   The C99 standard discusses two character sets: source and execution.
   The source character set is used for internal processing in translation
   phases 1 through 4; the execution character set is used thereafter.
   Both are required by 5.2.1.2p1 to be multibyte encodings, not wide
   character encodings (see 3.7.2, 3.7.3 for the standardese meanings
   of these terms).  Furthermore, the "basic character set" (listed in
   5.2.1p3) is to be encoded in each with values one byte wide, and is
   to appear in the initial shift state.

   It is not explicitly mentioned, but there is also a "wide execution
   character set" used to encode wide character constants and wide
   string literals; this is supposed to be the result of applying the
   standard library function mbstowcs() to an equivalent narrow string
   (6.4.5p5).  However, the behavior of hexadecimal and octal
   \-escapes is at odds with this; they are supposed to be translated
   directly to wchar_t values (6.4.4.4p5,6).

   The source character set is not necessarily the character set used
   to encode physical source files on disk; translation phase 1 converts
   from whatever that encoding is to the source character set.

   The presence of universal character names in C99 (6.4.3 et seq.)
   forces the source character set to be isomorphic to ISO 10646,
   that is, Unicode.  There is no such constraint on the execution
   character set; note also that the conversion from source to
   execution character set does not occur for identifiers (5.1.1.2p1#5).

   For convenience of implementation, the source character set's
   encoding of the basic character set should be identical to the
   execution character set OF THE HOST SYSTEM's encoding of the basic
   character set, and it should not be a state-dependent encoding.

   cpplib uses UTF-8 or UTF-EBCDIC for the source character set,
   depending on whether the host is based on ASCII or EBCDIC (see
   respectively Unicode section 2.3/ISO10646 Amendment 2, and Unicode
   Technical Report #16).  With limited exceptions, it relies on the
   system library's iconv() primitive to do charset conversion
   (specified in SUSv2).  */

#if !HAVE_ICONV
/* Make certain that the uses of iconv(), iconv_open(), iconv_close()
   below, which are guarded only by if statements with compile-time
   constant conditions, do not cause link errors.  */
#define iconv_open(x, y) (errno = EINVAL, (iconv_t)-1)
#define iconv(a,b,c,d,e) (errno = EINVAL, (size_t)-1)
#define iconv_close(x)   (void)0
#define ICONV_CONST
#endif

#if HOST_CHARSET == HOST_CHARSET_ASCII
#define SOURCE_CHARSET "UTF-8"
#define LAST_POSSIBLY_BASIC_SOURCE_CHAR 0x7e
#elif HOST_CHARSET == HOST_CHARSET_EBCDIC
#define SOURCE_CHARSET "UTF-EBCDIC"
#define LAST_POSSIBLY_BASIC_SOURCE_CHAR 0xFF
#else
#error "Unrecognized basic host character set"
#endif

#ifndef EILSEQ
#define EILSEQ EINVAL
#endif

/* This structure is used for a resizable string buffer throughout.  */
/* Don't call it strbuf, as that conflicts with unistd.h on systems
   such as DYNIX/ptx where unistd.h includes stropts.h.  */
struct _cpp_strbuf
{
  uchar *text;
  size_t asize;
  size_t len;
};

/* This is enough to hold any string that fits on a single 80-column
   line, even if iconv quadruples its size (e.g. conversion from
   ASCII to UTF-32) rounded up to a power of two.  */
#define OUTBUF_BLOCK_SIZE 256

/* Conversions between UTF-8 and UTF-16/32 are implemented by custom
   logic.  This is because a depressing number of systems lack iconv,
   or have have iconv libraries that do not do these conversions, so
   we need a fallback implementation for them.  To ensure the fallback
   doesn't break due to neglect, it is used on all systems.

   UTF-32 encoding is nice and simple: a four-byte binary number,
   constrained to the range 00000000-7FFFFFFF to avoid questions of
   signedness.  We do have to cope with big- and little-endian
   variants.

   UTF-16 encoding uses two-byte binary numbers, again in big- and
   little-endian variants, for all values in the 00000000-0000FFFF
   range.  Values in the 00010000-0010FFFF range are encoded as pairs
   of two-byte numbers, called "surrogate pairs": given a number S in
   this range, it is mapped to a pair (H, L) as follows:

     H = (S - 0x10000) / 0x400 + 0xD800
     L = (S - 0x10000) % 0x400 + 0xDC00

   Two-byte values in the D800...DFFF range are ill-formed except as a
   component of a surrogate pair.  Even if the encoding within a
   two-byte value is little-endian, the H member of the surrogate pair
   comes first.

   There is no way to encode values in the 00110000-7FFFFFFF range,
   which is not currently a problem as there are no assigned code
   points in that range; however, the author expects that it will
   eventually become necessary to abandon UTF-16 due to this
   limitation.  Note also that, because of these pairs, UTF-16 does
   not meet the requirements of the C standard for a wide character
   encoding (see 3.7.3 and 6.4.4.4p11).

   UTF-8 encoding looks like this:

   value range         encoded as
   00000000-0000007F   0xxxxxxx
   00000080-000007FF   110xxxxx 10xxxxxx
   00000800-0000FFFF   1110xxxx 10xxxxxx 10xxxxxx
   00010000-001FFFFF   11110xxx 10xxxxxx 10xxxxxx 10xxxxxx
   00200000-03FFFFFF   111110xx 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx
   04000000-7FFFFFFF   1111110x 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx 10xxxxxx

   Values in the 0000D800 ... 0000DFFF range (surrogates) are invalid,
   which means that three-byte sequences ED xx yy, with A0 <= xx <= BF,
   never occur.  Note also that any value that can be encoded by a
   given row of the table can also be encoded by all successive rows,
   but this is not done; only the shortest possible encoding for any
   given value is valid.  For instance, the character 07C0 could be
   encoded as any of DF 80, E0 9F 80, F0 80 9F 80, F8 80 80 9F 80, or
   FC 80 80 80 9F 80.  Only the first is valid.

   An implementation note: the transformation from UTF-16 to UTF-8, or
   vice versa, is easiest done by using UTF-32 as an intermediary.  */

/* Internal primitives which go from an UTF-8 byte stream to native-endian
   UTF-32 in a cppchar_t, or vice versa; this avoids an extra marshal/unmarshal
   operation in several places below.  */
static inline int
one_utf8_to_cppchar (const uchar **inbufp, size_t *inbytesleftp,
                     cppchar_t *cp)
{
  static const uchar masks[6] = { 0x7F, 0x1F, 0x0F, 0x07, 0x02, 0x01 };
  static const uchar patns[6] = { 0x00, 0xC0, 0xE0, 0xF0, 0xF8, 0xFC };

  cppchar_t c;
  const uchar *inbuf = *inbufp;
  size_t nbytes, i;

  if (*inbytesleftp < 1)
    return EINVAL;

  c = *inbuf;
  if (c < 0x80)
    {
      *cp = c;
      *inbytesleftp -= 1;
      *inbufp += 1;
      return 0;
    }

  /* The number of leading 1-bits in the first byte indicates how many
     bytes follow.  */
  for (nbytes = 2; nbytes < 7; nbytes++)
    if ((c & ~masks[nbytes-1]) == patns[nbytes-1])
      goto found;
  return EILSEQ;
 found:

  if (*inbytesleftp < nbytes)
    return EINVAL;

  c = (c & masks[nbytes-1]);
  inbuf++;
  for (i = 1; i < nbytes; i++)
    {
      cppchar_t n = *inbuf++;
      if ((n & 0xC0) != 0x80)
        return EILSEQ;
      c = ((c << 6) + (n & 0x3F));
    }

  /* Make sure the shortest possible encoding was used.  */
  if (c <=      0x7F && nbytes > 1) return EILSEQ;
  if (c <=     0x7FF && nbytes > 2) return EILSEQ;
  if (c <=    0xFFFF && nbytes > 3) return EILSEQ;
  if (c <=  0x1FFFFF && nbytes > 4) return EILSEQ;
  if (c <= 0x3FFFFFF && nbytes > 5) return EILSEQ;

  /* Make sure the character is valid.  */
  if (c > 0x7FFFFFFF || (c >= 0xD800 && c <= 0xDFFF)) return EILSEQ;

  *cp = c;
  *inbufp = inbuf;
  *inbytesleftp -= nbytes;
  return 0;
}

static inline int
one_cppchar_to_utf8 (cppchar_t c, uchar **outbufp, size_t *outbytesleftp)
{
  static const uchar masks[6] =  { 0x00, 0xC0, 0xE0, 0xF0, 0xF8, 0xFC };
  static const uchar limits[6] = { 0x80, 0xE0, 0xF0, 0xF8, 0xFC, 0xFE };
  size_t nbytes;
  uchar buf[6], *p = &buf[6];
  uchar *outbuf = *outbufp;

  nbytes = 1;
  if (c < 0x80)
    *--p = c;
  else
    {
      do
        {
          *--p = ((c & 0x3F) | 0x80);
          c >>= 6;
          nbytes++;
        }
      while (c >= 0x3F || (c & limits[nbytes-1]));
      *--p = (c | masks[nbytes-1]);
    }

  if (*outbytesleftp < nbytes)
    return E2BIG;

  while (p < &buf[6])
    *outbuf++ = *p++;
  *outbytesleftp -= nbytes;
  *outbufp = outbuf;
  return 0;
}

/* The following four functions transform one character between the two
   encodings named in the function name.  All have the signature
   int (*)(iconv_t bigend, const uchar **inbufp, size_t *inbytesleftp,
           uchar **outbufp, size_t *outbytesleftp)

   BIGEND must have the value 0 or 1, coerced to (iconv_t); it is
   interpreted as a boolean indicating whether big-endian or
   little-endian encoding is to be used for the member of the pair
   that is not UTF-8.

   INBUFP, INBYTESLEFTP, OUTBUFP, OUTBYTESLEFTP work exactly as they
   do for iconv.

   The return value is either 0 for success, or an errno value for
   failure, which may be E2BIG (need more space), EILSEQ (ill-formed
   input sequence), ir EINVAL (incomplete input sequence).  */

static inline int
one_utf8_to_utf32 (iconv_t bigend, const uchar **inbufp, size_t *inbytesleftp,
                   uchar **outbufp, size_t *outbytesleftp)
{
  uchar *outbuf;
  cppchar_t s = 0;
  int rval;

  /* Check for space first, since we know exactly how much we need.  */
  if (*outbytesleftp < 4)
    return E2BIG;

  rval = one_utf8_to_cppchar (inbufp, inbytesleftp, &s);
  if (rval)
    return rval;

  outbuf = *outbufp;
  outbuf[bigend ? 3 : 0] = (s & 0x000000FF);
  outbuf[bigend ? 2 : 1] = (s & 0x0000FF00) >> 8;
  outbuf[bigend ? 1 : 2] = (s & 0x00FF0000) >> 16;
  outbuf[bigend ? 0 : 3] = (s & 0xFF000000) >> 24;

  *outbufp += 4;
  *outbytesleftp -= 4;
  return 0;
}

static inline int
one_utf32_to_utf8 (iconv_t bigend, const uchar **inbufp, size_t *inbytesleftp,
                   uchar **outbufp, size_t *outbytesleftp)
{
  cppchar_t s;
  int rval;
  const uchar *inbuf;

  if (*inbytesleftp < 4)
    return EINVAL;

  inbuf = *inbufp;

  s  = inbuf[bigend ? 0 : 3] << 24;
  s += inbuf[bigend ? 1 : 2] << 16;
  s += inbuf[bigend ? 2 : 1] << 8;
  s += inbuf[bigend ? 3 : 0];

  if (s >= 0x7FFFFFFF || (s >= 0xD800 && s <= 0xDFFF))
    return EILSEQ;

  rval = one_cppchar_to_utf8 (s, outbufp, outbytesleftp);
  if (rval)
    return rval;

  *inbufp += 4;
  *inbytesleftp -= 4;
  return 0;
}

static inline int
one_utf8_to_utf16 (iconv_t bigend, const uchar **inbufp, size_t *inbytesleftp,
                   uchar **outbufp, size_t *outbytesleftp)
{
  int rval;
  cppchar_t s = 0;
  const uchar *save_inbuf = *inbufp;
  size_t save_inbytesleft = *inbytesleftp;
  uchar *outbuf = *outbufp;

  rval = one_utf8_to_cppchar (inbufp, inbytesleftp, &s);
  if (rval)
    return rval;

  if (s > 0x0010FFFF)
    {
      *inbufp = save_inbuf;
      *inbytesleftp = save_inbytesleft;
      return EILSEQ;
    }

  if (s < 0xFFFF)
    {
      if (*outbytesleftp < 2)
        {
          *inbufp = save_inbuf;
          *inbytesleftp = save_inbytesleft;
          return E2BIG;
        }
      outbuf[bigend ? 1 : 0] = (s & 0x00FF);
      outbuf[bigend ? 0 : 1] = (s & 0xFF00) >> 8;

      *outbufp += 2;
      *outbytesleftp -= 2;
      return 0;
    }
  else
    {
      cppchar_t hi, lo;

      if (*outbytesleftp < 4)
        {
          *inbufp = save_inbuf;
          *inbytesleftp = save_inbytesleft;
          return E2BIG;
        }

      hi = (s - 0x10000) / 0x400 + 0xD800;
      lo = (s - 0x10000) % 0x400 + 0xDC00;

      /* Even if we are little-endian, put the high surrogate first.
         ??? Matches practice?  */
      outbuf[bigend ? 1 : 0] = (hi & 0x00FF);
      outbuf[bigend ? 0 : 1] = (hi & 0xFF00) >> 8;
      outbuf[bigend ? 3 : 2] = (lo & 0x00FF);
      outbuf[bigend ? 2 : 3] = (lo & 0xFF00) >> 8;

      *outbufp += 4;
      *outbytesleftp -= 4;
      return 0;
    }
}

static inline int
one_utf16_to_utf8 (iconv_t bigend, const uchar **inbufp, size_t *inbytesleftp,
                   uchar **outbufp, size_t *outbytesleftp)
{
  cppchar_t s;
  const uchar *inbuf = *inbufp;
  int rval;

  if (*inbytesleftp < 2)
    return EINVAL;
  s  = inbuf[bigend ? 0 : 1] << 8;
  s += inbuf[bigend ? 1 : 0];

  /* Low surrogate without immediately preceding high surrogate is invalid.  */
  if (s >= 0xDC00 && s <= 0xDFFF)
    return EILSEQ;
  /* High surrogate must have a following low surrogate.  */
  else if (s >= 0xD800 && s <= 0xDBFF)
    {
      cppchar_t hi = s, lo;
      if (*inbytesleftp < 4)
        return EINVAL;

      lo  = inbuf[bigend ? 2 : 3] << 8;
      lo += inbuf[bigend ? 3 : 2];

      if (lo < 0xDC00 || lo > 0xDFFF)
        return EILSEQ;

      s = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000;
    }

  rval = one_cppchar_to_utf8 (s, outbufp, outbytesleftp);
  if (rval)
    return rval;

  /* Success - update the input pointers (one_cppchar_to_utf8 has done
     the output pointers for us).  */
  if (s <= 0xFFFF)
    {
      *inbufp += 2;
      *inbytesleftp -= 2;
    }
  else
    {
      *inbufp += 4;
      *inbytesleftp -= 4;
    }
  return 0;
}

/* Helper routine for the next few functions.  The 'const' on
   one_conversion means that we promise not to modify what function is
   pointed to, which lets the inliner see through it.  */

static inline bool
conversion_loop (int (*const one_conversion)(iconv_t, const uchar **, size_t *,
                                             uchar **, size_t *),
                 iconv_t cd, const uchar *from, size_t flen, struct _cpp_strbuf *to)
{
  const uchar *inbuf;
  uchar *outbuf;
  size_t inbytesleft, outbytesleft;
  int rval;

  inbuf = from;
  inbytesleft = flen;
  outbuf = to->text + to->len;
  outbytesleft = to->asize - to->len;

  for (;;)
    {
      do
        rval = one_conversion (cd, &inbuf, &inbytesleft,
                               &outbuf, &outbytesleft);
      while (inbytesleft && !rval);

      if (__builtin_expect (inbytesleft == 0, 1))
        {
          to->len = to->asize - outbytesleft;
          return true;
        }
      if (rval != E2BIG)
        {
          errno = rval;
          return false;
        }

      outbytesleft += OUTBUF_BLOCK_SIZE;
      to->asize += OUTBUF_BLOCK_SIZE;
      to->text = XRESIZEVEC (uchar, to->text, to->asize);
      outbuf = to->text + to->asize - outbytesleft;
    }
}


/* These functions convert entire strings between character sets.
   They all have the signature

   bool (*)(iconv_t cd, const uchar *from, size_t flen, struct _cpp_strbuf *to);

   The input string FROM is converted as specified by the function
   name plus the iconv descriptor CD (which may be fake), and the
   result appended to TO.  On any error, false is returned, otherwise true.  */

/* These four use the custom conversion code above.  */
static bool
convert_utf8_utf16 (iconv_t cd, const uchar *from, size_t flen,
                    struct _cpp_strbuf *to)
{
  return conversion_loop (one_utf8_to_utf16, cd, from, flen, to);
}

static bool
convert_utf8_utf32 (iconv_t cd, const uchar *from, size_t flen,
                    struct _cpp_strbuf *to)
{
  return conversion_loop (one_utf8_to_utf32, cd, from, flen, to);
}

static bool
convert_utf16_utf8 (iconv_t cd, const uchar *from, size_t flen,
                    struct _cpp_strbuf *to)
{
  return conversion_loop (one_utf16_to_utf8, cd, from, flen, to);
}

static bool
convert_utf32_utf8 (iconv_t cd, const uchar *from, size_t flen,
                    struct _cpp_strbuf *to)
{
  return conversion_loop (one_utf32_to_utf8, cd, from, flen, to);
}

/* Identity conversion, used when we have no alternative.  */
static bool
convert_no_conversion (iconv_t cd ATTRIBUTE_UNUSED,
                       const uchar *from, size_t flen, struct _cpp_strbuf *to)
{
  if (to->len + flen > to->asize)
    {
      to->asize = to->len + flen;
      to->text = XRESIZEVEC (uchar, to->text, to->asize);
    }
  memcpy (to->text + to->len, from, flen);
  to->len += flen;
  return true;
}

/* And this one uses the system iconv primitive.  It's a little
   different, since iconv's interface is a little different.  */
#if HAVE_ICONV
static bool
convert_using_iconv (iconv_t cd, const uchar *from, size_t flen,
                     struct _cpp_strbuf *to)
{
  ICONV_CONST char *inbuf;
  char *outbuf;
  size_t inbytesleft, outbytesleft;

  /* Reset conversion descriptor and check that it is valid.  */
  if (iconv (cd, 0, 0, 0, 0) == (size_t)-1)
    return false;

  inbuf = (ICONV_CONST char *)from;
  inbytesleft = flen;
  outbuf = (char *)to->text + to->len;
  outbytesleft = to->asize - to->len;

  for (;;)
    {
      iconv (cd, &inbuf, &inbytesleft, &outbuf, &outbytesleft);
      if (__builtin_expect (inbytesleft == 0, 1))
        {
          to->len = to->asize - outbytesleft;
          return true;
        }
      if (errno != E2BIG)
        return false;

      outbytesleft += OUTBUF_BLOCK_SIZE;
      to->asize += OUTBUF_BLOCK_SIZE;
      to->text = XRESIZEVEC (uchar, to->text, to->asize);
      outbuf = (char *)to->text + to->asize - outbytesleft;
    }
}
#else
#define convert_using_iconv 0 /* prevent undefined symbol error below */
#endif

/* Arrange for the above custom conversion logic to be used automatically
   when conversion between a suitable pair of character sets is requested.  */

#define APPLY_CONVERSION(CONVERTER, FROM, FLEN, TO) \
   CONVERTER.func (CONVERTER.cd, FROM, FLEN, TO)

struct conversion
{
  const char *pair;
  convert_f func;
  iconv_t fake_cd;
};
static const struct conversion conversion_tab[] = {
  { "UTF-8/UTF-32LE", convert_utf8_utf32, (iconv_t)0 },
  { "UTF-8/UTF-32BE", convert_utf8_utf32, (iconv_t)1 },
  { "UTF-8/UTF-16LE", convert_utf8_utf16, (iconv_t)0 },
  { "UTF-8/UTF-16BE", convert_utf8_utf16, (iconv_t)1 },
  { "UTF-32LE/UTF-8", convert_utf32_utf8, (iconv_t)0 },
  { "UTF-32BE/UTF-8", convert_utf32_utf8, (iconv_t)1 },
  { "UTF-16LE/UTF-8", convert_utf16_utf8, (iconv_t)0 },
  { "UTF-16BE/UTF-8", convert_utf16_utf8, (iconv_t)1 },
};

/* Subroutine of cpp_init_iconv: initialize and return a
   cset_converter structure for conversion from FROM to TO.  If
   iconv_open() fails, issue an error and return an identity
   converter.  Silently return an identity converter if FROM and TO
   are identical.  */
static struct cset_converter
init_iconv_desc (cpp_reader *pfile, const char *to, const char *from)
{
  struct cset_converter ret;
  char *pair;
  size_t i;

  if (!strcasecmp (to, from))
    {
      ret.func = convert_no_conversion;
      ret.cd = (iconv_t) -1;
      return ret;
    }

  pair = (char *) alloca(strlen(to) + strlen(from) + 2);

  strcpy(pair, from);
  strcat(pair, "/");
  strcat(pair, to);
  for (i = 0; i < ARRAY_SIZE (conversion_tab); i++)
    if (!strcasecmp (pair, conversion_tab[i].pair))
      {
        ret.func = conversion_tab[i].func;
        ret.cd = conversion_tab[i].fake_cd;
        return ret;
      }

  /* No custom converter - try iconv.  */
  if (HAVE_ICONV)
    {
      ret.func = convert_using_iconv;
      ret.cd = iconv_open (to, from);

      if (ret.cd == (iconv_t) -1)
        {
          if (errno == EINVAL)
            cpp_error (pfile, CPP_DL_ERROR, /* FIXME should be DL_SORRY */
                       "conversion from %s to %s not supported by iconv",
                       from, to);
          else
            cpp_errno (pfile, CPP_DL_ERROR, "iconv_open");

          ret.func = convert_no_conversion;
        }
    }
  else
    {
      cpp_error (pfile, CPP_DL_ERROR, /* FIXME: should be DL_SORRY */
                 "no iconv implementation, cannot convert from %s to %s",
                 from, to);
      ret.func = convert_no_conversion;
      ret.cd = (iconv_t) -1;
    }
  return ret;
}

/* If charset conversion is requested, initialize iconv(3) descriptors
   for conversion from the source character set to the execution
   character sets.  If iconv is not present in the C library, and
   conversion is requested, issue an error.  */

void
cpp_init_iconv (cpp_reader *pfile)
{
  const char *ncset = CPP_OPTION (pfile, narrow_charset);
  const char *wcset = CPP_OPTION (pfile, wide_charset);
  const char *default_wcset;

  bool be = CPP_OPTION (pfile, bytes_big_endian);

  if (CPP_OPTION (pfile, wchar_precision) >= 32)
    default_wcset = be ? "UTF-32BE" : "UTF-32LE";
  else if (CPP_OPTION (pfile, wchar_precision) >= 16)
    default_wcset = be ? "UTF-16BE" : "UTF-16LE";
  else
    /* This effectively means that wide strings are not supported,
       so don't do any conversion at all.  */
   default_wcset = SOURCE_CHARSET;

  if (!ncset)
    ncset = SOURCE_CHARSET;
  if (!wcset)
    wcset = default_wcset;

  pfile->narrow_cset_desc = init_iconv_desc (pfile, ncset, SOURCE_CHARSET);
  pfile->wide_cset_desc = init_iconv_desc (pfile, wcset, SOURCE_CHARSET);
}

/* Destroy iconv(3) descriptors set up by cpp_init_iconv, if necessary.  */
void
_cpp_destroy_iconv (cpp_reader *pfile)
{
  if (HAVE_ICONV)
    {
      if (pfile->narrow_cset_desc.func == convert_using_iconv)
        iconv_close (pfile->narrow_cset_desc.cd);
      if (pfile->wide_cset_desc.func == convert_using_iconv)
        iconv_close (pfile->wide_cset_desc.cd);
    }
}

/* Utility routine for use by a full compiler.  C is a character taken
   from the *basic* source character set, encoded in the host's
   execution encoding.  Convert it to (the target's) execution
   encoding, and return that value.

   Issues an internal error if C's representation in the narrow
   execution character set fails to be a single-byte value (C99
   5.2.1p3: "The representation of each member of the source and
   execution character sets shall fit in a byte.")  May also issue an
   internal error if C fails to be a member of the basic source
   character set (testing this exactly is too hard, especially when
   the host character set is EBCDIC).  */
cppchar_t
cpp_host_to_exec_charset (cpp_reader *pfile, cppchar_t c)
{
  uchar sbuf[1];
  struct _cpp_strbuf tbuf;

  /* This test is merely an approximation, but it suffices to catch
     the most important thing, which is that we don't get handed a
     character outside the unibyte range of the host character set.  */
  if (c > LAST_POSSIBLY_BASIC_SOURCE_CHAR)
    {
      cpp_error (pfile, CPP_DL_ICE,
                 "character 0x%lx is not in the basic source character set\n",
                 (unsigned long)c);
      return 0;
    }

  /* Being a character in the unibyte range of the host character set,
     we can safely splat it into a one-byte buffer and trust that that
     is a well-formed string.  */
  sbuf[0] = c;

  /* This should never need to reallocate, but just in case... */
  tbuf.asize = 1;
  tbuf.text = XNEWVEC (uchar, tbuf.asize);
  tbuf.len = 0;

  if (!APPLY_CONVERSION (pfile->narrow_cset_desc, sbuf, 1, &tbuf))
    {
      cpp_errno (pfile, CPP_DL_ICE, "converting to execution character set");
      return 0;
    }
  if (tbuf.len != 1)
    {
      cpp_error (pfile, CPP_DL_ICE,
                 "character 0x%lx is not unibyte in execution character set",
                 (unsigned long)c);
      return 0;
    }
  c = tbuf.text[0];
  free(tbuf.text);
  return c;
}

\f

/* Utility routine that computes a mask of the form 0000...111... with
   WIDTH 1-bits.  */
static inline size_t
width_to_mask (size_t width)
{
  width = MIN (width, BITS_PER_CPPCHAR_T);
  if (width >= CHAR_BIT * sizeof (size_t))
    return ~(size_t) 0;
  else
    return ((size_t) 1 << width) - 1;
}

/* A large table of unicode character information.  */
enum {
  /* Valid in a C99 identifier?  */
  C99 = 1,
  /* Valid in a C99 identifier, but not as the first character?  */
  DIG = 2,
  /* Valid in a C++ identifier?  */
  CXX = 4,
  /* NFC representation is not valid in an identifier?  */
  CID = 8,
  /* Might be valid NFC form?  */
  NFC = 16,
  /* Might be valid NFKC form?  */
  NKC = 32,
  /* Certain preceding characters might make it not valid NFC/NKFC form?  */
  CTX = 64
};

static const struct {
  /* Bitmap of flags above.  */
  unsigned char flags;
  /* Combining class of the character.  */
  unsigned char combine;
  /* Last character in the range described by this entry.  */
  unsigned short end;
} ucnranges[] = {
#include "ucnid.h"
};

/* Returns 1 if C is valid in an identifier, 2 if C is valid except at
   the start of an identifier, and 0 if C is not valid in an
   identifier.  We assume C has already gone through the checks of
   _cpp_valid_ucn.  Also update NST for C if returning nonzero.  The
   algorithm is a simple binary search on the table defined in
   ucnid.h.  */

static int
ucn_valid_in_identifier (cpp_reader *pfile, cppchar_t c,
                         struct normalize_state *nst)
{
  int mn, mx, md;

  if (c > 0xFFFF)
    return 0;

  mn = 0;
  mx = ARRAY_SIZE (ucnranges) - 1;
  while (mx != mn)
    {
      md = (mn + mx) / 2;
      if (c <= ucnranges[md].end)
        mx = md;
      else
        mn = md + 1;
    }

  /* When -pedantic, we require the character to have been listed by
     the standard for the current language.  Otherwise, we accept the
     union of the acceptable sets for C++98 and C99.  */
  if (! (ucnranges[mn].flags & (C99 | CXX)))
      return 0;

  if (CPP_PEDANTIC (pfile)
      && ((CPP_OPTION (pfile, c99) && !(ucnranges[mn].flags & C99))
          || (CPP_OPTION (pfile, cplusplus)
              && !(ucnranges[mn].flags & CXX))))
    return 0;

  /* Update NST.  */
  if (ucnranges[mn].combine != 0 && ucnranges[mn].combine < nst->prev_class)
    nst->level = normalized_none;
  else if (ucnranges[mn].flags & CTX)
    {
      bool safe;
      cppchar_t p = nst->previous;

      /* Easy cases from Bengali, Oriya, Tamil, Jannada, and Malayalam.  */
      if (c == 0x09BE)
        safe = p != 0x09C7;  /* Use 09CB instead of 09C7 09BE.  */
      else if (c == 0x0B3E)
        safe = p != 0x0B47;  /* Use 0B4B instead of 0B47 0B3E.  */
      else if (c == 0x0BBE)
        safe = p != 0x0BC6 && p != 0x0BC7;  /* Use 0BCA/0BCB instead.  */
      else if (c == 0x0CC2)
        safe = p != 0x0CC6;  /* Use 0CCA instead of 0CC6 0CC2.  */
      else if (c == 0x0D3E)
        safe = p != 0x0D46 && p != 0x0D47;  /* Use 0D4A/0D4B instead.  */
      /* For Hangul, characters in the range AC00-D7A3 are NFC/NFKC,
         and are combined algorithmically from a sequence of the form
         1100-1112 1161-1175 11A8-11C2
         (if the third is not present, it is treated as 11A7, which is not
         really a valid character).
         Unfortunately, C99 allows (only) the NFC form, but C++ allows
         only the combining characters.  */
      else if (c >= 0x1161 && c <= 0x1175)
        safe = p < 0x1100 || p > 0x1112;
      else if (c >= 0x11A8 && c <= 0x11C2)
        safe = (p < 0xAC00 || p > 0xD7A3 || (p - 0xAC00) % 28 != 0);
      else
        {
          /* Uh-oh, someone updated ucnid.h without updating this code.  */
          cpp_error (pfile, CPP_DL_ICE, "Character %x might not be NFKC", c);
          safe = true;
        }
      if (!safe && c < 0x1161)
        nst->level = normalized_none;
      else if (!safe)
        nst->level = MAX (nst->level, normalized_identifier_C);
    }
  else if (ucnranges[mn].flags & NKC)
    ;
  else if (ucnranges[mn].flags & NFC)
    nst->level = MAX (nst->level, normalized_C);
  else if (ucnranges[mn].flags & CID)
    nst->level = MAX (nst->level, normalized_identifier_C);
  else
    nst->level = normalized_none;
  nst->previous = c;
  nst->prev_class = ucnranges[mn].combine;

  /* In C99, UCN digits may not begin identifiers.  */
  if (CPP_OPTION (pfile, c99) && (ucnranges[mn].flags & DIG))
    return 2;

  return 1;
}

/* [lex.charset]: The character designated by the universal character
   name \UNNNNNNNN is that character whose character short name in
   ISO/IEC 10646 is NNNNNNNN; the character designated by the
   universal character name \uNNNN is that character whose character
   short name in ISO/IEC 10646 is 0000NNNN.  If the hexadecimal value
   for a universal character name is less than 0x20 or in the range
   0x7F-0x9F (inclusive), or if the universal character name
   designates a character in the basic source character set, then the
   program is ill-formed.

   *PSTR must be preceded by "\u" or "\U"; it is assumed that the
   buffer end is delimited by a non-hex digit.  Returns zero if the
   UCN has not been consumed.

   Otherwise the nonzero value of the UCN, whether valid or invalid,
   is returned.  Diagnostics are emitted for invalid values.  PSTR
   is updated to point one beyond the UCN, or to the syntactically
   invalid character.

   IDENTIFIER_POS is 0 when not in an identifier, 1 for the start of
   an identifier, or 2 otherwise.  */

cppchar_t
_cpp_valid_ucn (cpp_reader *pfile, const uchar **pstr,
                const uchar *limit, int identifier_pos,
                struct normalize_state *nst)
{
  cppchar_t result, c;
  unsigned int length;
  const uchar *str = *pstr;
  const uchar *base = str - 2;

  if (!CPP_OPTION (pfile, cplusplus) && !CPP_OPTION (pfile, c99))
    cpp_error (pfile, CPP_DL_WARNING,
               "universal character names are only valid in C++ and C99");
  else if (CPP_WTRADITIONAL (pfile) && identifier_pos == 0)
    cpp_error (pfile, CPP_DL_WARNING,
               "the meaning of '\\%c' is different in traditional C",
               (int) str[-1]);

  if (str[-1] == 'u')
    length = 4;
  else if (str[-1] == 'U')
    length = 8;
  else
    {
      cpp_error (pfile, CPP_DL_ICE, "In _cpp_valid_ucn but not a UCN");
      length = 4;
    }

  result = 0;
  do
    {
      c = *str;
      if (!ISXDIGIT (c))
        break;
      str++;
      result = (result << 4) + hex_value (c);
    }
  while (--length && str < limit);

  /* Partial UCNs are not valid in strings, but decompose into
     multiple tokens in identifiers, so we can't give a helpful
     error message in that case.  */
  if (length && identifier_pos)
    return 0;
  
  *pstr = str;
  if (length)
    {
      cpp_error (pfile, CPP_DL_ERROR,
                 "incomplete universal character name %.*s",
                 (int) (str - base), base);
      result = 1;
    }
  /* The standard permits $, @ and ` to be specified as UCNs.  We use
     hex escapes so that this also works with EBCDIC hosts.  */
  else if ((result < 0xa0
            && (result != 0x24 && result != 0x40 && result != 0x60))
           || (result & 0x80000000)
           || (result >= 0xD800 && result <= 0xDFFF))
    {
      cpp_error (pfile, CPP_DL_ERROR,
                 "%.*s is not a valid universal character",
                 (int) (str - base), base);
      result = 1;
    }
  else if (identifier_pos && result == 0x24 
           && CPP_OPTION (pfile, dollars_in_ident))
    {
      if (CPP_OPTION (pfile, warn_dollars) && !pfile->state.skipping)
        {
          CPP_OPTION (pfile, warn_dollars) = 0;
          cpp_error (pfile, CPP_DL_PEDWARN, "'$' in identifier or number");
        }
      NORMALIZE_STATE_UPDATE_IDNUM (nst);
    }
  else if (identifier_pos)
    {
      int validity = ucn_valid_in_identifier (pfile, result, nst);

      if (validity == 0)
        cpp_error (pfile, CPP_DL_ERROR,
                   "universal character %.*s is not valid in an identifier",
                   (int) (str - base), base);
      else if (validity == 2 && identifier_pos == 1)
        cpp_error (pfile, CPP_DL_ERROR,
   "universal character %.*s is not valid at the start of an identifier",
                   (int) (str - base), base);
    }

  if (result == 0)
    result = 1;

  return result;
}

/* Convert an UCN, pointed to by FROM, to UTF-8 encoding, then translate
   it to the execution character set and write the result into TBUF.
   An advanced pointer is returned.  Issues all relevant diagnostics.  */
static const uchar *
convert_ucn (cpp_reader *pfile, const uchar *from, const uchar *limit,
             struct _cpp_strbuf *tbuf, bool wide)
{
  cppchar_t ucn;
  uchar buf[6];
  uchar *bufp = buf;
  size_t bytesleft = 6;
  int rval;
  struct cset_converter cvt
    = wide ? pfile->wide_cset_desc : pfile->narrow_cset_desc;
  struct normalize_state nst = INITIAL_NORMALIZE_STATE;

  from++;  /* Skip u/U.  */
  ucn = _cpp_valid_ucn (pfile, &from, limit, 0, &nst);

  rval = one_cppchar_to_utf8 (ucn, &bufp, &bytesleft);
  if (rval)
    {
      errno = rval;
      cpp_errno (pfile, CPP_DL_ERROR,
                 "converting UCN to source character set");
    }
  else if (!APPLY_CONVERSION (cvt, buf, 6 - bytesleft, tbuf))
    cpp_errno (pfile, CPP_DL_ERROR,
               "converting UCN to execution character set");

  return from;
}

/* Subroutine of convert_hex and convert_oct.  N is the representation
   in the execution character set of a numeric escape; write it into the
   string buffer TBUF and update the end-of-string pointer therein.  WIDE
   is true if it's a wide string that's being assembled in TBUF.  This
   function issues no diagnostics and never fails.  */
static void
emit_numeric_escape (cpp_reader *pfile, cppchar_t n,
                     struct _cpp_strbuf *tbuf, bool wide)
{
  if (wide)
    {
      /* We have to render this into the target byte order, which may not
         be our byte order.  */
      bool bigend = CPP_OPTION (pfile, bytes_big_endian);
      size_t width = CPP_OPTION (pfile, wchar_precision);
      size_t cwidth = CPP_OPTION (pfile, char_precision);
      size_t cmask = width_to_mask (cwidth);
      size_t nbwc = width / cwidth;
      size_t i;
      size_t off = tbuf->len;
      cppchar_t c;

      if (tbuf->len + nbwc > tbuf->asize)
        {
          tbuf->asize += OUTBUF_BLOCK_SIZE;
          tbuf->text = XRESIZEVEC (uchar, tbuf->text, tbuf->asize);
        }

      for (i = 0; i < nbwc; i++)
        {
          c = n & cmask;
          n >>= cwidth;
          tbuf->text[off + (bigend ? nbwc - i - 1 : i)] = c;
        }
      tbuf->len += nbwc;
    }
  else
    {
      /* Note: this code does not handle the case where the target
         and host have a different number of bits in a byte.  */
      if (tbuf->len + 1 > tbuf->asize)
        {
          tbuf->asize += OUTBUF_BLOCK_SIZE;
          tbuf->text = XRESIZEVEC (uchar, tbuf->text, tbuf->asize);
        }
      tbuf->text[tbuf->len++] = n;
    }
}

/* Convert a hexadecimal escape, pointed to by FROM, to the execution
   character set and write it into the string buffer TBUF.  Returns an
   advanced pointer, and issues diagnostics as necessary.
   No character set translation occurs; this routine always produces the
   execution-set character with numeric value equal to the given hex
   number.  You can, e.g. generate surrogate pairs this way.  */
static const uchar *
convert_hex (cpp_reader *pfile, const uchar *from, const uchar *limit,
             struct _cpp_strbuf *tbuf, bool wide)
{
  cppchar_t c, n = 0, overflow = 0;
  int digits_found = 0;
  size_t width = (wide ? CPP_OPTION (pfile, wchar_precision)
                  : CPP_OPTION (pfile, char_precision));
  size_t mask = width_to_mask (width);

  if (CPP_WTRADITIONAL (pfile))
    cpp_error (pfile, CPP_DL_WARNING,
               "the meaning of '\\x' is different in traditional C");

  from++;  /* Skip 'x'.  */
  while (from < limit)
    {
      c = *from;
      if (! hex_p (c))
        break;
      from++;
      overflow |= n ^ (n << 4 >> 4);
      n = (n << 4) + hex_value (c);
      digits_found = 1;
    }

  if (!digits_found)
    {
      cpp_error (pfile, CPP_DL_ERROR,
                 "\\x used with no following hex digits");
      return from;
    }

  if (overflow | (n != (n & mask)))
    {
      cpp_error (pfile, CPP_DL_PEDWARN,
                 "hex escape sequence out of range");
      n &= mask;
    }

  emit_numeric_escape (pfile, n, tbuf, wide);

  return from;
}

/* Convert an octal escape, pointed to by FROM, to the execution
   character set and write it into the string buffer TBUF.  Returns an
   advanced pointer, and issues diagnostics as necessary.
   No character set translation occurs; this routine always produces the
   execution-set character with numeric value equal to the given octal
   number.  */
static const uchar *
convert_oct (cpp_reader *pfile, const uchar *from, const uchar *limit,
             struct _cpp_strbuf *tbuf, bool wide)
{
  size_t count = 0;
  cppchar_t c, n = 0;
  size_t width = (wide ? CPP_OPTION (pfile, wchar_precision)
                  : CPP_OPTION (pfile, char_precision));
  size_t mask = width_to_mask (width);
  bool overflow = false;

  while (from < limit && count++ < 3)
    {
      c = *from;
      if (c < '0' || c > '7')
        break;
      from++;
      overflow |= n ^ (n << 3 >> 3);
      n = (n << 3) + c - '0';
    }

  if (n != (n & mask))
    {
      cpp_error (pfile, CPP_DL_PEDWARN,
                 "octal escape sequence out of range");
      n &= mask;
    }

  emit_numeric_escape (pfile, n, tbuf, wide);

  return from;
}

/* Convert an escape sequence (pointed to by FROM) to its value on
   the target, and to the execution character set.  Do not scan past
   LIMIT.  Write the converted value into TBUF.  Returns an advanced
   pointer.  Handles all relevant diagnostics.  */
static const uchar *
convert_escape (cpp_reader *pfile, const uchar *from, const uchar *limit,
                struct _cpp_strbuf *tbuf, bool wide)
{
  /* Values of \a \b \e \f \n \r \t \v respectively.  */
#if HOST_CHARSET == HOST_CHARSET_ASCII
  static const uchar charconsts[] = {  7,  8, 27, 12, 10, 13,  9, 11 };
#elif HOST_CHARSET == HOST_CHARSET_EBCDIC
  static const uchar charconsts[] = { 47, 22, 39, 12, 21, 13,  5, 11 };
#else
#error "unknown host character set"
#endif

  uchar c;
  struct cset_converter cvt
    = wide ? pfile->wide_cset_desc : pfile->narrow_cset_desc;

  c = *from;
  switch (c)
    {
      /* UCNs, hex escapes, and octal escapes are processed separately.  */
    case 'u': case 'U':
      return convert_ucn (pfile, from, limit, tbuf, wide);

    case 'x':
      return convert_hex (pfile, from, limit, tbuf, wide);
      break;

    case '0':  case '1':  case '2':  case '3':
    case '4':  case '5':  case '6':  case '7':
      return convert_oct (pfile, from, limit, tbuf, wide);

      /* Various letter escapes.  Get the appropriate host-charset
         value into C.  */
    case '\\': case '\'': case '"': case '?': break;

    case '(': case '{': case '[': case '%':
      /* '\(', etc, can be used at the beginning of a line in a long
         string split onto multiple lines with \-newline, to prevent
         Emacs or other text editors from getting confused.  '\%' can
         be used to prevent SCCS from mangling printf format strings.  */
      if (CPP_PEDANTIC (pfile))
        goto unknown;
      break;

    case 'b': c = charconsts[1];  break;
    case 'f': c = charconsts[3];  break;
    case 'n': c = charconsts[4];  break;
    case 'r': c = charconsts[5];  break;
    case 't': c = charconsts[6];  break;
    case 'v': c = charconsts[7];  break;

    case 'a':
      if (CPP_WTRADITIONAL (pfile))
        cpp_error (pfile, CPP_DL_WARNING,
                   "the meaning of '\\a' is different in traditional C");
      c = charconsts[0];
      break;

    case 'e': case 'E':
      if (CPP_PEDANTIC (pfile))
        cpp_error (pfile, CPP_DL_PEDWARN,
                   "non-ISO-standard escape sequence, '\\%c'", (int) c);
      c = charconsts[2];
      break;

    default:
    unknown:
      if (ISGRAPH (c))
        cpp_error (pfile, CPP_DL_PEDWARN,
                   "unknown escape sequence '\\%c'", (int) c);
      else
        {
          /* diagnostic.c does not support "%03o".  When it does, this
             code can use %03o directly in the diagnostic again.  */
          char buf[32];
          sprintf(buf, "%03o", (int) c);
          cpp_error (pfile, CPP_DL_PEDWARN,
                     "unknown escape sequence: '\\%s'", buf);
        }
    }

  /* Now convert what we have to the execution character set.  */
  if (!APPLY_CONVERSION (cvt, &c, 1, tbuf))
    cpp_errno (pfile, CPP_DL_ERROR,
               "converting escape sequence to execution character set");

  return from + 1;
}
\f
/* FROM is an array of cpp_string structures of length COUNT.  These
   are to be converted from the source to the execution character set,
   escape sequences translated, and finally all are to be
   concatenated.  WIDE indicates whether or not to produce a wide
   string.  The result is written into TO.  Returns true for success,
   false for failure.  */
bool
cpp_interpret_string (cpp_reader *pfile, const cpp_string *from, size_t count,
                      cpp_string *to, bool wide)
{
  struct _cpp_strbuf tbuf;
  const uchar *p, *base, *limit;
  size_t i;
  struct cset_converter cvt
    = wide ? pfile->wide_cset_desc : pfile->narrow_cset_desc;

  tbuf.asize = MAX (OUTBUF_BLOCK_SIZE, from->len);
  tbuf.text = XNEWVEC (uchar, tbuf.asize);
  tbuf.len = 0;

  for (i = 0; i < count; i++)
    {
      p = from[i].text;
      if (*p == 'L') p++;
      p++; /* Skip leading quote.  */
      limit = from[i].text + from[i].len - 1; /* Skip trailing quote.  */

      for (;;)
        {
          base = p;
          while (p < limit && *p != '\\')
            p++;
          if (p > base)
            {
              /* We have a run of normal characters; these can be fed
                 directly to convert_cset.  */
              if (!APPLY_CONVERSION (cvt, base, p - base, &tbuf))
                goto fail;
            }
          if (p == limit)
            break;

          p = convert_escape (pfile, p + 1, limit, &tbuf, wide);
        }
    }
  /* NUL-terminate the 'to' buffer and translate it to a cpp_string
     structure.  */
  emit_numeric_escape (pfile, 0, &tbuf, wide);
  tbuf.text = XRESIZEVEC (uchar, tbuf.text, tbuf.len);
  to->text = tbuf.text;
  to->len = tbuf.len;
  return true;

 fail:
  cpp_errno (pfile, CPP_DL_ERROR, "converting to execution character set");
  free (tbuf.text);
  return false;
}

/* Subroutine of do_line and do_linemarker.  Convert escape sequences
   in a string, but do not perform character set conversion.  */
bool
cpp_interpret_string_notranslate (cpp_reader *pfile, const cpp_string *from,
                                  size_t count, cpp_string *to, bool wide)
{
  struct cset_converter save_narrow_cset_desc = pfile->narrow_cset_desc;
  bool retval;

  pfile->narrow_cset_desc.func = convert_no_conversion;
  pfile->narrow_cset_desc.cd = (iconv_t) -1;

  retval = cpp_interpret_string (pfile, from, count, to, wide);

  pfile->narrow_cset_desc = save_narrow_cset_desc;
  return retval;
}

\f
/* Subroutine of cpp_interpret_charconst which performs the conversion
   to a number, for narrow strings.  STR is the string structure returned
   by cpp_interpret_string.  PCHARS_SEEN and UNSIGNEDP are as for
   cpp_interpret_charconst.  */
static cppchar_t
narrow_str_to_charconst (cpp_reader *pfile, cpp_string str,
                         unsigned int *pchars_seen, int *unsignedp)
{
  size_t width = CPP_OPTION (pfile, char_precision);
  size_t max_chars = CPP_OPTION (pfile, int_precision) / width;
  size_t mask = width_to_mask (width);
  size_t i;
  cppchar_t result, c;
  bool unsigned_p;

  /* The value of a multi-character character constant, or a
     single-character character constant whose representation in the
     execution character set is more than one byte long, is
     implementation defined.  This implementation defines it to be the
     number formed by interpreting the byte sequence in memory as a
     big-endian binary number.  If overflow occurs, the high bytes are
     lost, and a warning is issued.

     We don't want to process the NUL terminator handed back by
     cpp_interpret_string.  */
  result = 0;
  for (i = 0; i < str.len - 1; i++)
    {
      c = str.text[i] & mask;
      if (width < BITS_PER_CPPCHAR_T)
        result = (result << width) | c;
      else
        result = c;
    }

  if (i > max_chars)
    {
      i = max_chars;
      cpp_error (pfile, CPP_DL_WARNING,
                 "character constant too long for its type");
    }
  else if (i > 1 && CPP_OPTION (pfile, warn_multichar))
    cpp_error (pfile, CPP_DL_WARNING, "multi-character character constant");

  /* Multichar constants are of type int and therefore signed.  */
  if (i > 1)
    unsigned_p = 0;
  else
    unsigned_p = CPP_OPTION (pfile, unsigned_char);

  /* Truncate the constant to its natural width, and simultaneously
     sign- or zero-extend to the full width of cppchar_t.
     For single-character constants, the value is WIDTH bits wide.
     For multi-character constants, the value is INT_PRECISION bits wide.  */
  if (i > 1)
    width = CPP_OPTION (pfile, int_precision);
  if (width < BITS_PER_CPPCHAR_T)
    {
      mask = ((cppchar_t) 1 << width) - 1;
      if (unsigned_p || !(result & (1 << (width - 1))))
        result &= mask;
      else
        result |= ~mask;
    }
  *pchars_seen = i;
  *unsignedp = unsigned_p;
  return result;
}

/* Subroutine of cpp_interpret_charconst which performs the conversion
   to a number, for wide strings.  STR is the string structure returned
   by cpp_interpret_string.  PCHARS_SEEN and UNSIGNEDP are as for
   cpp_interpret_charconst.  */
static cppchar_t
wide_str_to_charconst (cpp_reader *pfile, cpp_string str,
                       unsigned int *pchars_seen, int *unsignedp)
{
  bool bigend = CPP_OPTION (pfile, bytes_big_endian);
  size_t width = CPP_OPTION (pfile, wchar_precision);
  size_t cwidth = CPP_OPTION (pfile, char_precision);
  size_t mask = width_to_mask (width);
  size_t cmask = width_to_mask (cwidth);
  size_t nbwc = width / cwidth;
  size_t off, i;
  cppchar_t result = 0, c;

  /* This is finicky because the string is in the target's byte order,
     which may not be our byte order.  Only the last character, ignoring
     the NUL terminator, is relevant.  */
  off = str.len - (nbwc * 2);
  result = 0;
  for (i = 0; i < nbwc; i++)
    {
      c = bigend ? str.text[off + i] : str.text[off + nbwc - i - 1];
      result = (result << cwidth) | (c & cmask);
    }

  /* Wide character constants have type wchar_t, and a single
     character exactly fills a wchar_t, so a multi-character wide
     character constant is guaranteed to overflow.  */
  if (off > 0)
    cpp_error (pfile, CPP_DL_WARNING,
               "character constant too long for its type");

  /* Truncate the constant to its natural width, and simultaneously
     sign- or zero-extend to the full width of cppchar_t.  */
  if (width < BITS_PER_CPPCHAR_T)
    {
      if (CPP_OPTION (pfile, unsigned_wchar) || !(result & (1 << (width - 1))))
        result &= mask;
      else
        result |= ~mask;
    }

  *unsignedp = CPP_OPTION (pfile, unsigned_wchar);
  *pchars_seen = 1;
  return result;
}

/* Interpret a (possibly wide) character constant in TOKEN.
   PCHARS_SEEN points to a variable that is filled in with the number
   of characters seen, and UNSIGNEDP to a variable that indicates
   whether the result has signed type.  */
cppchar_t
cpp_interpret_charconst (cpp_reader *pfile, const cpp_token *token,
                         unsigned int *pchars_seen, int *unsignedp)
{
  cpp_string str = { 0, 0 };
  bool wide = (token->type == CPP_WCHAR);
  cppchar_t result;

  /* an empty constant will appear as L'' or '' */
  if (token->val.str.len == (size_t) (2 + wide))
    {
      cpp_error (pfile, CPP_DL_ERROR, "empty character constant");
      return 0;
    }
  else if (!cpp_interpret_string (pfile, &token->val.str, 1, &str, wide))
    return 0;

  if (wide)
    result = wide_str_to_charconst (pfile, str, pchars_seen, unsignedp);
  else
    result = narrow_str_to_charconst (pfile, str, pchars_seen, unsignedp);

  if (str.text != token->val.str.text)
    free ((void *)str.text);

  return result;
}
\f
/* Convert an identifier denoted by ID and LEN, which might contain
   UCN escapes, to the source character set, either UTF-8 or
   UTF-EBCDIC.  Assumes that the identifier is actually a valid identifier.  */
cpp_hashnode *
_cpp_interpret_identifier (cpp_reader *pfile, const uchar *id, size_t len)
{
  /* It turns out that a UCN escape always turns into fewer characters
     than the escape itself, so we can allocate a temporary in advance.  */
  uchar * buf = (uchar *) alloca (len + 1);
  uchar * bufp = buf;
  size_t idp;
  
  for (idp = 0; idp < len; idp++)
    if (id[idp] != '\\')
      *bufp++ = id[idp];
    else
      {
        unsigned length = id[idp+1] == 'u' ? 4 : 8;
        cppchar_t value = 0;
        size_t bufleft = len - (bufp - buf);
        int rval;

        idp += 2;
        while (length && idp < len && ISXDIGIT (id[idp]))
          {
            value = (value << 4) + hex_value (id[idp]);
            idp++;
            length--;
          }
        idp--;

        /* Special case for EBCDIC: if the identifier contains
           a '$' specified using a UCN, translate it to EBCDIC.  */
        if (value == 0x24)
          {
            *bufp++ = '$';
            continue;
          }

        rval = one_cppchar_to_utf8 (value, &bufp, &bufleft);
        if (rval)
          {
            errno = rval;
            cpp_errno (pfile, CPP_DL_ERROR,
                       "converting UCN to source character set");
            break;
          }
      }

  return CPP_HASHNODE (ht_lookup (pfile->hash_table, 
                                  buf, bufp - buf, HT_ALLOC));
}
\f
/* Convert an input buffer (containing the complete contents of one
   source file) from INPUT_CHARSET to the source character set.  INPUT
   points to the input buffer, SIZE is its allocated size, and LEN is
   the length of the meaningful data within the buffer.  The
   translated buffer is returned, and *ST_SIZE is set to the length of
   the meaningful data within the translated buffer.

   INPUT is expected to have been allocated with xmalloc.  This function
   will either return INPUT, or free it and return a pointer to another
   xmalloc-allocated block of memory.  */
uchar * 
_cpp_convert_input (cpp_reader *pfile, const char *input_charset,
                    uchar *input, size_t size, size_t len, off_t *st_size)
{
  struct cset_converter input_cset;
  struct _cpp_strbuf to;

  input_cset = init_iconv_desc (pfile, SOURCE_CHARSET, input_charset);
  if (input_cset.func == convert_no_conversion)
    {
      to.text = input;
      to.asize = size;
      to.len = len;
    }
  else
    {
      to.asize = MAX (65536, len);
      to.text = XNEWVEC (uchar, to.asize);
      to.len = 0;

      if (!APPLY_CONVERSION (input_cset, input, len, &to))
        cpp_error (pfile, CPP_DL_ERROR,
                   "failure to convert %s to %s",
                   CPP_OPTION (pfile, input_charset), SOURCE_CHARSET);

      free (input);
    }

  /* Clean up the mess.  */
  if (input_cset.func == convert_using_iconv)
    iconv_close (input_cset.cd);

  /* Resize buffer if we allocated substantially too much, or if we
     haven't enough space for the \n-terminator.  */
  if (to.len + 4096 < to.asize || to.len >= to.asize)
    to.text = XRESIZEVEC (uchar, to.text, to.len + 1);

  /* If the file is using old-school Mac line endings (\r only),
     terminate with another \r, not an \n, so that we do not mistake
     the \r\n sequence for a single DOS line ending and erroneously
     issue the "No newline at end of file" diagnostic.  */
  if (to.len && to.text[to.len - 1] == '\r')
    to.text[to.len] = '\r';
  else
    to.text[to.len] = '\n';

  *st_size = to.len;
  return to.text;
}

/* Decide on the default encoding to assume for input files.  */
const char *
_cpp_default_encoding (void)
{
  const char *current_encoding = NULL;

  /* We disable this because the default codeset is 7-bit ASCII on
     most platforms, and this causes conversion failures on every
     file in GCC that happens to have one of the upper 128 characters
     in it -- most likely, as part of the name of a contributor.
     We should definitely recognize in-band markers of file encoding,
     like:
     - the appropriate Unicode byte-order mark (FE FF) to recognize
       UTF16 and UCS4 (in both big-endian and little-endian flavors)
       and UTF8
     - a "#i", "#d", "/ *", "//", " #p" or "#p" (for #pragma) to
       distinguish ASCII and EBCDIC.
     - now we can parse something like "#pragma GCC encoding <xyz>
       on the first line, or even Emacs/VIM's mode line tags (there's
       a problem here in that VIM uses the last line, and Emacs has
       its more elaborate "local variables" convention).
     - investigate whether Java has another common convention, which
       would be friendly to support.
     (Zack Weinberg and Paolo Bonzini, May 20th 2004)  */
#if defined (HAVE_LOCALE_H) && defined (HAVE_LANGINFO_CODESET) && 0
  setlocale (LC_CTYPE, "");
  current_encoding = nl_langinfo (CODESET);
#endif
  if (current_encoding == NULL || *current_encoding == '\0')
    current_encoding = SOURCE_CHARSET;

  return current_encoding;
}