/**********************************************************************

  pack.c -

  $Author$
  created at: Thu Feb 10 15:17:05 JST 1994

  Copyright (C) 1993-2007 Yukihiro Matsumoto

**********************************************************************/

#include "ruby/encoding.h"
#include "internal.h"
#include <sys/types.h>
#include <ctype.h>
#include <errno.h>
#include <float.h>

/*
 * It is intentional that the condition for natstr is HAVE_TRUE_LONG_LONG
 * instead of HAVE_LONG_LONG or LONG_LONG.
 * This means q! and Q! means always the standard long long type and
 * causes ArgumentError for platforms which has no long long type,
 * even if the platform has an implementation specific 64bit type.
 * This behavior is consistent with the document of pack/unpack.
 */
#ifdef HAVE_TRUE_LONG_LONG
static const char natstr[] = "sSiIlLqQjJ";
#else
static const char natstr[] = "sSiIlLjJ";
#endif
static const char endstr[] = "sSiIlLqQjJ";

#ifdef HAVE_TRUE_LONG_LONG
/* It is intentional to use long long instead of LONG_LONG. */
# define NATINT_LEN_Q NATINT_LEN(long long, 8)
#else
# define NATINT_LEN_Q 8
#endif

#if SIZEOF_SHORT != 2 || SIZEOF_LONG != 4 || (defined(HAVE_TRUE_LONG_LONG) && SIZEOF_LONG_LONG != 8)
# define NATINT_PACK
#endif

#ifdef DYNAMIC_ENDIAN
/* for universal binary of NEXTSTEP and MacOS X */
/* useless since autoconf 2.63? */
static int
is_bigendian(void)
{
    static int init = 0;
    static int endian_value;
    char *p;

    if (init) return endian_value;
    init = 1;
    p = (char*)&init;
    return endian_value = p[0]?0:1;
}
# define BIGENDIAN_P() (is_bigendian())
#elif defined(WORDS_BIGENDIAN)
# define BIGENDIAN_P() 1
#else
# define BIGENDIAN_P() 0
#endif

#ifdef NATINT_PACK
# define NATINT_LEN(type,len) (natint?(int)sizeof(type):(int)(len))
#else
# define NATINT_LEN(type,len) ((int)sizeof(type))
#endif

typedef union {
    float f;
    uint32_t u;
    char buf[4];
} FLOAT_SWAPPER;
typedef union {
    double d;
    uint64_t u;
    char buf[8];
} DOUBLE_SWAPPER;
#define swapf(x) swap32(x)
#define swapd(x) swap64(x)

#define rb_ntohf(x) (BIGENDIAN_P()?(x):swapf(x))
#define rb_ntohd(x) (BIGENDIAN_P()?(x):swapd(x))
#define rb_htonf(x) (BIGENDIAN_P()?(x):swapf(x))
#define rb_htond(x) (BIGENDIAN_P()?(x):swapd(x))
#define rb_htovf(x) (BIGENDIAN_P()?swapf(x):(x))
#define rb_htovd(x) (BIGENDIAN_P()?swapd(x):(x))
#define rb_vtohf(x) (BIGENDIAN_P()?swapf(x):(x))
#define rb_vtohd(x) (BIGENDIAN_P()?swapd(x):(x))

#define FLOAT_CONVWITH(x)	FLOAT_SWAPPER x;
#define HTONF(x)	((x).u = rb_htonf((x).u))
#define HTOVF(x)	((x).u = rb_htovf((x).u))
#define NTOHF(x)	((x).u = rb_ntohf((x).u))
#define VTOHF(x)	((x).u = rb_vtohf((x).u))

#define DOUBLE_CONVWITH(x)	DOUBLE_SWAPPER x;
#define HTOND(x)	((x).u = rb_htond((x).u))
#define HTOVD(x)	((x).u = rb_htovd((x).u))
#define NTOHD(x)	((x).u = rb_ntohd((x).u))
#define VTOHD(x)	((x).u = rb_vtohd((x).u))

#define MAX_INTEGER_PACK_SIZE 8

static const char toofew[] = "too few arguments";

static void encodes(VALUE,const char*,long,int,int);
static void qpencode(VALUE,VALUE,long);

static unsigned long utf8_to_uv(const char*,long*);

static ID id_associated;

static void
str_associate(VALUE str, VALUE add)
{
    /* assert(NIL_P(rb_attr_get(str, id_associated))); */
    rb_ivar_set(str, id_associated, add);
}

static VALUE
str_associated(VALUE str)
{
    return rb_ivar_lookup(str, id_associated, Qfalse);
}

static void
unknown_directive(const char *mode, char type, VALUE fmt)
{
    VALUE f;
    char unknown[5];

    if (ISPRINT(type)) {
        unknown[0] = type;
        unknown[1] = '\0';
    }
    else {
        snprintf(unknown, sizeof(unknown), "\\x%.2x", type & 0xff);
    }
    f = rb_str_quote_unprintable(fmt);
    if (f != fmt) {
        fmt = rb_str_subseq(f, 1, RSTRING_LEN(f) - 2);
    }
    rb_warning("unknown %s directive '%s' in '%"PRIsVALUE"'",
               mode, unknown, fmt);
}

static float
VALUE_to_float(VALUE obj)
{
    VALUE v = rb_to_float(obj);
    double d = RFLOAT_VALUE(v);

    if (isnan(d)) {
        return NAN;
    }
    else if (d < -FLT_MAX) {
        return -INFINITY;
    }
    else if (d <= FLT_MAX) {
        return d;
    }
    else {
        return INFINITY;
    }
}

/*
 *  call-seq:
 *     arr.pack( aTemplateString ) -> aBinaryString
 *     arr.pack( aTemplateString, buffer: aBufferString ) -> aBufferString
 *
 *  Packs the contents of <i>arr</i> into a binary sequence according to
 *  the directives in <i>aTemplateString</i> (see the table below)
 *  Directives ``A,'' ``a,'' and ``Z'' may be followed by a count,
 *  which gives the width of the resulting field. The remaining
 *  directives also may take a count, indicating the number of array
 *  elements to convert. If the count is an asterisk
 *  (``<code>*</code>''), all remaining array elements will be
 *  converted. Any of the directives ``<code>sSiIlL</code>'' may be
 *  followed by an underscore (``<code>_</code>'') or
 *  exclamation mark (``<code>!</code>'') to use the underlying
 *  platform's native size for the specified type; otherwise, they use a
 *  platform-independent size. Spaces are ignored in the template
 *  string. See also String#unpack.
 *
 *     a = [ "a", "b", "c" ]
 *     n = [ 65, 66, 67 ]
 *     a.pack("A3A3A3")   #=> "a  b  c  "
 *     a.pack("a3a3a3")   #=> "a\000\000b\000\000c\000\000"
 *     n.pack("ccc")      #=> "ABC"
 *
 *  If <i>aBufferString</i> is specified and its capacity is enough,
 *  +pack+ uses it as the buffer and returns it.
 *  When the offset is specified by the beginning of <i>aTemplateString</i>,
 *  the result is filled after the offset.
 *  If original contents of <i>aBufferString</i> exists and it's longer than
 *  the offset, the rest of <i>offsetOfBuffer</i> are overwritten by the result.
 *  If it's shorter, the gap is filled with ``<code>\0</code>''.
 *
 *  Note that ``buffer:'' option does not guarantee not to allocate memory
 *  in +pack+.  If the capacity of <i>aBufferString</i> is not enough,
 *  +pack+ allocates memory.
 *
 *  Directives for +pack+.
 *
 *   Integer       | Array   |
 *   Directive     | Element | Meaning
 *   ----------------------------------------------------------------------------
 *   C             | Integer | 8-bit unsigned (unsigned char)
 *   S             | Integer | 16-bit unsigned, native endian (uint16_t)
 *   L             | Integer | 32-bit unsigned, native endian (uint32_t)
 *   Q             | Integer | 64-bit unsigned, native endian (uint64_t)
 *   J             | Integer | pointer width unsigned, native endian (uintptr_t)
 *                 |         | (J is available since Ruby 2.3.)
 *                 |         |
 *   c             | Integer | 8-bit signed (signed char)
 *   s             | Integer | 16-bit signed, native endian (int16_t)
 *   l             | Integer | 32-bit signed, native endian (int32_t)
 *   q             | Integer | 64-bit signed, native endian (int64_t)
 *   j             | Integer | pointer width signed, native endian (intptr_t)
 *                 |         | (j is available since Ruby 2.3.)
 *                 |         |
 *   S_ S!         | Integer | unsigned short, native endian
 *   I I_ I!       | Integer | unsigned int, native endian
 *   L_ L!         | Integer | unsigned long, native endian
 *   Q_ Q!         | Integer | unsigned long long, native endian (ArgumentError
 *                 |         | if the platform has no long long type.)
 *                 |         | (Q_ and Q! is available since Ruby 2.1.)
 *   J!            | Integer | uintptr_t, native endian (same with J)
 *                 |         | (J! is available since Ruby 2.3.)
 *                 |         |
 *   s_ s!         | Integer | signed short, native endian
 *   i i_ i!       | Integer | signed int, native endian
 *   l_ l!         | Integer | signed long, native endian
 *   q_ q!         | Integer | signed long long, native endian (ArgumentError
 *                 |         | if the platform has no long long type.)
 *                 |         | (q_ and q! is available since Ruby 2.1.)
 *   j!            | Integer | intptr_t, native endian (same with j)
 *                 |         | (j! is available since Ruby 2.3.)
 *                 |         |
 *   S> s> S!> s!> | Integer | same as the directives without ">" except
 *   L> l> L!> l!> |         | big endian
 *   I!> i!>       |         | (available since Ruby 1.9.3)
 *   Q> q> Q!> q!> |         | "S>" is same as "n"
 *   J> j> J!> j!> |         | "L>" is same as "N"
 *                 |         |
 *   S< s< S!< s!< | Integer | same as the directives without "<" except
 *   L< l< L!< l!< |         | little endian
 *   I!< i!<       |         | (available since Ruby 1.9.3)
 *   Q< q< Q!< q!< |         | "S<" is same as "v"
 *   J< j< J!< j!< |         | "L<" is same as "V"
 *                 |         |
 *   n             | Integer | 16-bit unsigned, network (big-endian) byte order
 *   N             | Integer | 32-bit unsigned, network (big-endian) byte order
 *   v             | Integer | 16-bit unsigned, VAX (little-endian) byte order
 *   V             | Integer | 32-bit unsigned, VAX (little-endian) byte order
 *                 |         |
 *   U             | Integer | UTF-8 character
 *   w             | Integer | BER-compressed integer
 *
 *   Float        | Array   |
 *   Directive    | Element | Meaning
 *   ---------------------------------------------------------------------------
 *   D d          | Float   | double-precision, native format
 *   F f          | Float   | single-precision, native format
 *   E            | Float   | double-precision, little-endian byte order
 *   e            | Float   | single-precision, little-endian byte order
 *   G            | Float   | double-precision, network (big-endian) byte order
 *   g            | Float   | single-precision, network (big-endian) byte order
 *
 *   String       | Array   |
 *   Directive    | Element | Meaning
 *   ---------------------------------------------------------------------------
 *   A            | String  | arbitrary binary string (space padded, count is width)
 *   a            | String  | arbitrary binary string (null padded, count is width)
 *   Z            | String  | same as ``a'', except that null is added with *
 *   B            | String  | bit string (MSB first)
 *   b            | String  | bit string (LSB first)
 *   H            | String  | hex string (high nibble first)
 *   h            | String  | hex string (low nibble first)
 *   u            | String  | UU-encoded string
 *   M            | String  | quoted printable, MIME encoding (see also RFC2045)
 *                |         | (text mode but input must use LF and output LF)
 *   m            | String  | base64 encoded string (see RFC 2045)
 *                |         | (if count is 0, no line feed are added, see RFC 4648)
 *                |         | (count specifies input bytes between each LF,
 *                |         | rounded down to nearest multiple of 3)
 *   P            | String  | pointer to a structure (fixed-length string)
 *   p            | String  | pointer to a null-terminated string
 *
 *   Misc.        | Array   |
 *   Directive    | Element | Meaning
 *   ---------------------------------------------------------------------------
 *   @            | ---     | moves to absolute position
 *   X            | ---     | back up a byte
 *   x            | ---     | null byte
 */

static VALUE
pack_pack(int argc, VALUE *argv, VALUE ary)
{
    static const char nul10[] = "\0\0\0\0\0\0\0\0\0\0";
    static const char spc10[] = "          ";
    const char *p, *pend;
    VALUE fmt, opt = Qnil, res, from, associates = 0, buffer = 0;
    char type;
    long len, idx, plen;
    const char *ptr;
    int enc_info = 1;		/* 0 - BINARY, 1 - US-ASCII, 2 - UTF-8 */
#ifdef NATINT_PACK
    int natint;		/* native integer */
#endif
    int integer_size, bigendian_p;

    rb_scan_args(argc, argv, "10:", &fmt, &opt);

    StringValue(fmt);
    p = RSTRING_PTR(fmt);
    pend = p + RSTRING_LEN(fmt);
    if (!NIL_P(opt)) {
	static ID keyword_ids[1];
	if (!keyword_ids[0])
	    CONST_ID(keyword_ids[0], "buffer");

	rb_get_kwargs(opt, keyword_ids, 0, 1, &buffer);

	if (buffer != Qundef && !RB_TYPE_P(buffer, T_STRING))
	    rb_raise(rb_eTypeError, "buffer must be String, not %s", rb_obj_classname(buffer));
    }
    if (buffer)
	res = buffer;
    else
	res = rb_str_buf_new(0);

    idx = 0;

#define TOO_FEW (rb_raise(rb_eArgError, toofew), 0)
#define MORE_ITEM (idx < RARRAY_LEN(ary))
#define THISFROM (MORE_ITEM ? RARRAY_AREF(ary, idx) : TOO_FEW)
#define NEXTFROM (MORE_ITEM ? RARRAY_AREF(ary, idx++) : TOO_FEW)

    while (p < pend) {
	int explicit_endian = 0;
	if (RSTRING_PTR(fmt) + RSTRING_LEN(fmt) != pend) {
	    rb_raise(rb_eRuntimeError, "format string modified");
	}
	type = *p++;		/* get data type */
#ifdef NATINT_PACK
	natint = 0;
#endif

	if (ISSPACE(type)) continue;
	if (type == '#') {
	    while ((p < pend) && (*p != '\n')) {
		p++;
	    }
	    continue;
	}

	{
          modifiers:
	    switch (*p) {
	      case '_':
	      case '!':
		if (strchr(natstr, type)) {
#ifdef NATINT_PACK
		    natint = 1;
#endif
		    p++;
		}
		else {
		    rb_raise(rb_eArgError, "'%c' allowed only after types %s", *p, natstr);
		}
		goto modifiers;

	      case '<':
	      case '>':
		if (!strchr(endstr, type)) {
		    rb_raise(rb_eArgError, "'%c' allowed only after types %s", *p, endstr);
		}
		if (explicit_endian) {
		    rb_raise(rb_eRangeError, "Can't use both '<' and '>'");
		}
		explicit_endian = *p++;
		goto modifiers;
	    }
	}

	if (*p == '*') {	/* set data length */
	    len = strchr("@Xxu", type) ? 0
                : strchr("PMm", type) ? 1
                : RARRAY_LEN(ary) - idx;
	    p++;
	}
	else if (ISDIGIT(*p)) {
	    errno = 0;
	    len = STRTOUL(p, (char**)&p, 10);
	    if (errno) {
		rb_raise(rb_eRangeError, "pack length too big");
	    }
	}
	else {
	    len = 1;
	}

	switch (type) {
	  case 'U':
	    /* if encoding is US-ASCII, upgrade to UTF-8 */
	    if (enc_info == 1) enc_info = 2;
	    break;
	  case 'm': case 'M': case 'u':
	    /* keep US-ASCII (do nothing) */
	    break;
	  default:
	    /* fall back to BINARY */
	    enc_info = 0;
	    break;
	}
	switch (type) {
	  case 'A': case 'a': case 'Z':
	  case 'B': case 'b':
	  case 'H': case 'h':
	    from = NEXTFROM;
	    if (NIL_P(from)) {
		ptr = "";
		plen = 0;
	    }
	    else {
		StringValue(from);
		ptr = RSTRING_PTR(from);
		plen = RSTRING_LEN(from);
		OBJ_INFECT(res, from);
	    }

	    if (p[-1] == '*')
		len = plen;

	    switch (type) {
	      case 'a':		/* arbitrary binary string (null padded)  */
	      case 'A':         /* arbitrary binary string (ASCII space padded) */
	      case 'Z':         /* null terminated string  */
		if (plen >= len) {
		    rb_str_buf_cat(res, ptr, len);
		    if (p[-1] == '*' && type == 'Z')
			rb_str_buf_cat(res, nul10, 1);
		}
		else {
		    rb_str_buf_cat(res, ptr, plen);
		    len -= plen;
		    while (len >= 10) {
			rb_str_buf_cat(res, (type == 'A')?spc10:nul10, 10);
			len -= 10;
		    }
		    rb_str_buf_cat(res, (type == 'A')?spc10:nul10, len);
		}
		break;

#define castchar(from) (char)((from) & 0xff)

	      case 'b':		/* bit string (ascending) */
		{
		    int byte = 0;
		    long i, j = 0;

		    if (len > plen) {
			j = (len - plen + 1)/2;
			len = plen;
		    }
		    for (i=0; i++ < len; ptr++) {
			if (*ptr & 1)
			    byte |= 128;
			if (i & 7)
			    byte >>= 1;
			else {
			    char c = castchar(byte);
			    rb_str_buf_cat(res, &c, 1);
			    byte = 0;
			}
		    }
		    if (len & 7) {
			char c;
			byte >>= 7 - (len & 7);
			c = castchar(byte);
			rb_str_buf_cat(res, &c, 1);
		    }
		    len = j;
		    goto grow;
		}
		break;

	      case 'B':		/* bit string (descending) */
		{
		    int byte = 0;
		    long i, j = 0;

		    if (len > plen) {
			j = (len - plen + 1)/2;
			len = plen;
		    }
		    for (i=0; i++ < len; ptr++) {
			byte |= *ptr & 1;
			if (i & 7)
			    byte <<= 1;
			else {
			    char c = castchar(byte);
			    rb_str_buf_cat(res, &c, 1);
			    byte = 0;
			}
		    }
		    if (len & 7) {
			char c;
			byte <<= 7 - (len & 7);
			c = castchar(byte);
			rb_str_buf_cat(res, &c, 1);
		    }
		    len = j;
		    goto grow;
		}
		break;

	      case 'h':		/* hex string (low nibble first) */
		{
		    int byte = 0;
		    long i, j = 0;

		    if (len > plen) {
			j = (len + 1) / 2 - (plen + 1) / 2;
			len = plen;
		    }
		    for (i=0; i++ < len; ptr++) {
			if (ISALPHA(*ptr))
			    byte |= (((*ptr & 15) + 9) & 15) << 4;
			else
			    byte |= (*ptr & 15) << 4;
			if (i & 1)
			    byte >>= 4;
			else {
			    char c = castchar(byte);
			    rb_str_buf_cat(res, &c, 1);
			    byte = 0;
			}
		    }
		    if (len & 1) {
			char c = castchar(byte);
			rb_str_buf_cat(res, &c, 1);
		    }
		    len = j;
		    goto grow;
		}
		break;

	      case 'H':		/* hex string (high nibble first) */
		{
		    int byte = 0;
		    long i, j = 0;

		    if (len > plen) {
			j = (len + 1) / 2 - (plen + 1) / 2;
			len = plen;
		    }
		    for (i=0; i++ < len; ptr++) {
			if (ISALPHA(*ptr))
			    byte |= ((*ptr & 15) + 9) & 15;
			else
			    byte |= *ptr & 15;
			if (i & 1)
			    byte <<= 4;
			else {
			    char c = castchar(byte);
			    rb_str_buf_cat(res, &c, 1);
			    byte = 0;
			}
		    }
		    if (len & 1) {
			char c = castchar(byte);
			rb_str_buf_cat(res, &c, 1);
		    }
		    len = j;
		    goto grow;
		}
		break;
	    }
	    break;

	  case 'c':		/* signed char */
	  case 'C':		/* unsigned char */
            integer_size = 1;
            bigendian_p = BIGENDIAN_P(); /* not effective */
            goto pack_integer;

	  case 's':		/* s for int16_t, s! for signed short */
            integer_size = NATINT_LEN(short, 2);
            bigendian_p = BIGENDIAN_P();
            goto pack_integer;

	  case 'S':		/* S for uint16_t, S! for unsigned short */
            integer_size = NATINT_LEN(short, 2);
            bigendian_p = BIGENDIAN_P();
            goto pack_integer;

	  case 'i':		/* i and i! for signed int */
            integer_size = (int)sizeof(int);
            bigendian_p = BIGENDIAN_P();
            goto pack_integer;

	  case 'I':		/* I and I! for unsigned int */
            integer_size = (int)sizeof(int);
            bigendian_p = BIGENDIAN_P();
            goto pack_integer;

	  case 'l':		/* l for int32_t, l! for signed long */
            integer_size = NATINT_LEN(long, 4);
            bigendian_p = BIGENDIAN_P();
            goto pack_integer;

	  case 'L':		/* L for uint32_t, L! for unsigned long */
            integer_size = NATINT_LEN(long, 4);
            bigendian_p = BIGENDIAN_P();
            goto pack_integer;

	  case 'q':		/* q for int64_t, q! for signed long long */
	    integer_size = NATINT_LEN_Q;
            bigendian_p = BIGENDIAN_P();
            goto pack_integer;

	  case 'Q':		/* Q for uint64_t, Q! for unsigned long long */
	    integer_size = NATINT_LEN_Q;
            bigendian_p = BIGENDIAN_P();
            goto pack_integer;

	  case 'j':		/* j for intptr_t */
	    integer_size = sizeof(intptr_t);
	    bigendian_p = BIGENDIAN_P();
	    goto pack_integer;

	  case 'J':		/* J for uintptr_t */
	    integer_size = sizeof(uintptr_t);
	    bigendian_p = BIGENDIAN_P();
	    goto pack_integer;

	  case 'n':		/* 16 bit (2 bytes) integer (network byte-order)  */
            integer_size = 2;
            bigendian_p = 1;
            goto pack_integer;

	  case 'N':		/* 32 bit (4 bytes) integer (network byte-order) */
            integer_size = 4;
            bigendian_p = 1;
            goto pack_integer;

	  case 'v':		/* 16 bit (2 bytes) integer (VAX byte-order) */
            integer_size = 2;
            bigendian_p = 0;
            goto pack_integer;

	  case 'V':		/* 32 bit (4 bytes) integer (VAX byte-order) */
            integer_size = 4;
            bigendian_p = 0;
            goto pack_integer;

          pack_integer:
	    if (explicit_endian) {
		bigendian_p = explicit_endian == '>';
	    }
            if (integer_size > MAX_INTEGER_PACK_SIZE)
                rb_bug("unexpected intger size for pack: %d", integer_size);
            while (len-- > 0) {
                char intbuf[MAX_INTEGER_PACK_SIZE];

                from = NEXTFROM;
                rb_integer_pack(from, intbuf, integer_size, 1, 0,
                    INTEGER_PACK_2COMP |
                    (bigendian_p ? INTEGER_PACK_BIG_ENDIAN : INTEGER_PACK_LITTLE_ENDIAN));
                rb_str_buf_cat(res, intbuf, integer_size);
            }
	    break;

	  case 'f':		/* single precision float in native format */
	  case 'F':		/* ditto */
	    while (len-- > 0) {
		float f;

		from = NEXTFROM;
                f = VALUE_to_float(from);
		rb_str_buf_cat(res, (char*)&f, sizeof(float));
	    }
	    break;

	  case 'e':		/* single precision float in VAX byte-order */
	    while (len-- > 0) {
		FLOAT_CONVWITH(tmp);

		from = NEXTFROM;
                tmp.f = VALUE_to_float(from);
		HTOVF(tmp);
		rb_str_buf_cat(res, tmp.buf, sizeof(float));
	    }
	    break;

	  case 'E':		/* double precision float in VAX byte-order */
	    while (len-- > 0) {
		DOUBLE_CONVWITH(tmp);
		from = NEXTFROM;
		tmp.d = RFLOAT_VALUE(rb_to_float(from));
		HTOVD(tmp);
		rb_str_buf_cat(res, tmp.buf, sizeof(double));
	    }
	    break;

	  case 'd':		/* double precision float in native format */
	  case 'D':		/* ditto */
	    while (len-- > 0) {
		double d;

		from = NEXTFROM;
		d = RFLOAT_VALUE(rb_to_float(from));
		rb_str_buf_cat(res, (char*)&d, sizeof(double));
	    }
	    break;

	  case 'g':		/* single precision float in network byte-order */
	    while (len-- > 0) {
		FLOAT_CONVWITH(tmp);
		from = NEXTFROM;
                tmp.f = VALUE_to_float(from);
		HTONF(tmp);
		rb_str_buf_cat(res, tmp.buf, sizeof(float));
	    }
	    break;

	  case 'G':		/* double precision float in network byte-order */
	    while (len-- > 0) {
		DOUBLE_CONVWITH(tmp);

		from = NEXTFROM;
		tmp.d = RFLOAT_VALUE(rb_to_float(from));
		HTOND(tmp);
		rb_str_buf_cat(res, tmp.buf, sizeof(double));
	    }
	    break;

	  case 'x':		/* null byte */
	  grow:
	    while (len >= 10) {
		rb_str_buf_cat(res, nul10, 10);
		len -= 10;
	    }
	    rb_str_buf_cat(res, nul10, len);
	    break;

	  case 'X':		/* back up byte */
	  shrink:
	    plen = RSTRING_LEN(res);
	    if (plen < len)
		rb_raise(rb_eArgError, "X outside of string");
	    rb_str_set_len(res, plen - len);
	    break;

	  case '@':		/* null fill to absolute position */
	    len -= RSTRING_LEN(res);
	    if (len > 0) goto grow;
	    len = -len;
	    if (len > 0) goto shrink;
	    break;

	  case '%':
	    rb_raise(rb_eArgError, "%% is not supported");
	    break;

	  case 'U':		/* Unicode character */
	    while (len-- > 0) {
		SIGNED_VALUE l;
		char buf[8];
		int le;

		from = NEXTFROM;
		from = rb_to_int(from);
		l = NUM2LONG(from);
		if (l < 0) {
		    rb_raise(rb_eRangeError, "pack(U): value out of range");
		}
		le = rb_uv_to_utf8(buf, l);
		rb_str_buf_cat(res, (char*)buf, le);
	    }
	    break;

	  case 'u':		/* uuencoded string */
	  case 'm':		/* base64 encoded string */
	    from = NEXTFROM;
	    StringValue(from);
	    ptr = RSTRING_PTR(from);
	    plen = RSTRING_LEN(from);
            OBJ_INFECT(res, from);

	    if (len == 0 && type == 'm') {
		encodes(res, ptr, plen, type, 0);
		ptr += plen;
		break;
	    }
	    if (len <= 2)
		len = 45;
	    else if (len > 63 && type == 'u')
		len = 63;
	    else
		len = len / 3 * 3;
	    while (plen > 0) {
		long todo;

		if (plen > len)
		    todo = len;
		else
		    todo = plen;
		encodes(res, ptr, todo, type, 1);
		plen -= todo;
		ptr += todo;
	    }
	    break;

	  case 'M':		/* quoted-printable encoded string */
	    from = rb_obj_as_string(NEXTFROM);
            OBJ_INFECT(res, from);
	    if (len <= 1)
		len = 72;
	    qpencode(res, from, len);
	    break;

	  case 'P':		/* pointer to packed byte string */
	    from = THISFROM;
	    if (!NIL_P(from)) {
		StringValue(from);
		if (RSTRING_LEN(from) < len) {
		    rb_raise(rb_eArgError, "too short buffer for P(%ld for %ld)",
			     RSTRING_LEN(from), len);
		}
	    }
	    len = 1;
	    /* FALL THROUGH */
	  case 'p':		/* pointer to string */
	    while (len-- > 0) {
		char *t;
		from = NEXTFROM;
		if (NIL_P(from)) {
		    t = 0;
		}
		else {
		    t = StringValuePtr(from);
                    OBJ_INFECT(res, from);
		    rb_obj_taint(from);
		}
		if (!associates) {
		    associates = rb_ary_new();
		}
		rb_ary_push(associates, from);
		rb_str_buf_cat(res, (char*)&t, sizeof(char*));
	    }
	    break;

	  case 'w':		/* BER compressed integer  */
	    while (len-- > 0) {
		VALUE buf = rb_str_new(0, 0);
                size_t numbytes;
                int sign;
                char *cp;

		from = NEXTFROM;
                from = rb_to_int(from);
                numbytes = rb_absint_numwords(from, 7, NULL);
                if (numbytes == 0)
                    numbytes = 1;
                buf = rb_str_new(NULL, numbytes);

                sign = rb_integer_pack(from, RSTRING_PTR(buf), RSTRING_LEN(buf), 1, 1, INTEGER_PACK_BIG_ENDIAN);

                if (sign < 0)
                    rb_raise(rb_eArgError, "can't compress negative numbers");
                if (sign == 2)
                    rb_bug("buffer size problem?");

                cp = RSTRING_PTR(buf);
                while (1 < numbytes) {
                    *cp |= 0x80;
                    cp++;
                    numbytes--;
                }

                rb_str_buf_cat(res, RSTRING_PTR(buf), RSTRING_LEN(buf));
	    }
	    break;

	  default: {
            unknown_directive("pack", type, fmt);
	    break;
	  }
	}
    }

    if (associates) {
	str_associate(res, associates);
    }
    OBJ_INFECT(res, fmt);
    switch (enc_info) {
      case 1:
	ENCODING_CODERANGE_SET(res, rb_usascii_encindex(), ENC_CODERANGE_7BIT);
	break;
      case 2:
	rb_enc_set_index(res, rb_utf8_encindex());
	break;
      default:
	/* do nothing, keep ASCII-8BIT */
	break;
    }
    return res;
}

static const char uu_table[] =
"`!\"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\\]^_";
static const char b64_table[] =
"ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/";

static void
encodes(VALUE str, const char *s0, long len, int type, int tail_lf)
{
    enum {buff_size = 4096, encoded_unit = 4, input_unit = 3};
    char buff[buff_size + 1];	/* +1 for tail_lf */
    long i = 0;
    const char *const trans = type == 'u' ? uu_table : b64_table;
    char padding;
    const unsigned char *s = (const unsigned char *)s0;

    if (type == 'u') {
	buff[i++] = (char)len + ' ';
	padding = '`';
    }
    else {
	padding = '=';
    }
    while (len >= input_unit) {
        while (len >= input_unit && buff_size-i >= encoded_unit) {
            buff[i++] = trans[077 & (*s >> 2)];
            buff[i++] = trans[077 & (((*s << 4) & 060) | ((s[1] >> 4) & 017))];
            buff[i++] = trans[077 & (((s[1] << 2) & 074) | ((s[2] >> 6) & 03))];
            buff[i++] = trans[077 & s[2]];
            s += input_unit;
            len -= input_unit;
        }
        if (buff_size-i < encoded_unit) {
            rb_str_buf_cat(str, buff, i);
            i = 0;
        }
    }

    if (len == 2) {
	buff[i++] = trans[077 & (*s >> 2)];
	buff[i++] = trans[077 & (((*s << 4) & 060) | ((s[1] >> 4) & 017))];
	buff[i++] = trans[077 & (((s[1] << 2) & 074) | (('\0' >> 6) & 03))];
	buff[i++] = padding;
    }
    else if (len == 1) {
	buff[i++] = trans[077 & (*s >> 2)];
	buff[i++] = trans[077 & (((*s << 4) & 060) | (('\0' >> 4) & 017))];
	buff[i++] = padding;
	buff[i++] = padding;
    }
    if (tail_lf) buff[i++] = '\n';
    rb_str_buf_cat(str, buff, i);
    if ((size_t)i > sizeof(buff)) rb_bug("encodes() buffer overrun");
}

static const char hex_table[] = "0123456789ABCDEF";

static void
qpencode(VALUE str, VALUE from, long len)
{
    char buff[1024];
    long i = 0, n = 0, prev = EOF;
    unsigned char *s = (unsigned char*)RSTRING_PTR(from);
    unsigned char *send = s + RSTRING_LEN(from);

    while (s < send) {
        if ((*s > 126) ||
	    (*s < 32 && *s != '\n' && *s != '\t') ||
	    (*s == '=')) {
	    buff[i++] = '=';
	    buff[i++] = hex_table[*s >> 4];
	    buff[i++] = hex_table[*s & 0x0f];
            n += 3;
            prev = EOF;
        }
	else if (*s == '\n') {
            if (prev == ' ' || prev == '\t') {
		buff[i++] = '=';
		buff[i++] = *s;
            }
	    buff[i++] = *s;
            n = 0;
            prev = *s;
        }
	else {
	    buff[i++] = *s;
            n++;
            prev = *s;
        }
        if (n > len) {
	    buff[i++] = '=';
	    buff[i++] = '\n';
            n = 0;
            prev = '\n';
        }
	if (i > 1024 - 5) {
	    rb_str_buf_cat(str, buff, i);
	    i = 0;
	}
	s++;
    }
    if (n > 0) {
	buff[i++] = '=';
	buff[i++] = '\n';
    }
    if (i > 0) {
	rb_str_buf_cat(str, buff, i);
    }
}

static inline int
hex2num(char c)
{
    int n;
    n = ruby_digit36_to_number_table[(unsigned char)c];
    if (16 <= n)
        n = -1;
    return n;
}

#define PACK_LENGTH_ADJUST_SIZE(sz) do {	\
    tmp_len = 0;				\
    if (len > (long)((send-s)/(sz))) {		\
        if (!star) {				\
	    tmp_len = len-(send-s)/(sz);	\
        }					\
	len = (send-s)/(sz);			\
    }						\
} while (0)

#define PACK_ITEM_ADJUST() do { \
    if (tmp_len > 0 && mode == UNPACK_ARRAY) \
	rb_ary_store(ary, RARRAY_LEN(ary)+tmp_len-1, Qnil); \
} while (0)

/* Workaround for Oracle Developer Studio (Oracle Solaris Studio)
 * 12.4/12.5/12.6 C compiler optimization bug
 * with "-xO4" optimization option.
 */
#if defined(__SUNPRO_C) && 0x5130 <= __SUNPRO_C && __SUNPRO_C <= 0x5150
# define AVOID_CC_BUG volatile
#else
# define AVOID_CC_BUG
#endif

static VALUE
infected_str_new(const char *ptr, long len, VALUE str)
{
    VALUE s = rb_str_new(ptr, len);

    OBJ_INFECT(s, str);
    return s;
}

/* unpack mode */
#define UNPACK_ARRAY 0
#define UNPACK_BLOCK 1
#define UNPACK_1 2

static VALUE
pack_unpack_internal(VALUE str, VALUE fmt, int mode)
{
#define hexdigits ruby_hexdigits
    char *s, *send;
    char *p, *pend;
    VALUE ary;
    char type;
    long len;
    AVOID_CC_BUG long tmp_len;
    int star;
#ifdef NATINT_PACK
    int natint;			/* native integer */
#endif
    int signed_p, integer_size, bigendian_p;
#define UNPACK_PUSH(item) do {\
	VALUE item_val = (item);\
	if ((mode) == UNPACK_BLOCK) {\
	    rb_yield(item_val);\
	}\
	else if ((mode) == UNPACK_ARRAY) {\
	    rb_ary_push(ary, item_val);\
	}\
	else /* if ((mode) == UNPACK_1) { */ {\
	    return item_val; \
	}\
    } while (0)

    StringValue(str);
    StringValue(fmt);
    s = RSTRING_PTR(str);
    send = s + RSTRING_LEN(str);
    p = RSTRING_PTR(fmt);
    pend = p + RSTRING_LEN(fmt);

    ary = mode == UNPACK_ARRAY ? rb_ary_new() : Qnil;
    while (p < pend) {
	int explicit_endian = 0;
	type = *p++;
#ifdef NATINT_PACK
	natint = 0;
#endif

	if (ISSPACE(type)) continue;
	if (type == '#') {
	    while ((p < pend) && (*p != '\n')) {
		p++;
	    }
	    continue;
	}

	star = 0;
	{
          modifiers:
	    switch (*p) {
	      case '_':
	      case '!':

		if (strchr(natstr, type)) {
#ifdef NATINT_PACK
		    natint = 1;
#endif
		    p++;
		}
		else {
		    rb_raise(rb_eArgError, "'%c' allowed only after types %s", *p, natstr);
		}
		goto modifiers;

	      case '<':
	      case '>':
		if (!strchr(endstr, type)) {
		    rb_raise(rb_eArgError, "'%c' allowed only after types %s", *p, endstr);
		}
		if (explicit_endian) {
		    rb_raise(rb_eRangeError, "Can't use both '<' and '>'");
		}
		explicit_endian = *p++;
		goto modifiers;
	    }
	}

	if (p >= pend)
	    len = 1;
	else if (*p == '*') {
	    star = 1;
	    len = send - s;
	    p++;
	}
	else if (ISDIGIT(*p)) {
	    errno = 0;
	    len = STRTOUL(p, (char**)&p, 10);
	    if (len < 0 || errno) {
		rb_raise(rb_eRangeError, "pack length too big");
	    }
	}
	else {
	    len = (type != '@');
	}

	switch (type) {
	  case '%':
	    rb_raise(rb_eArgError, "%% is not supported");
	    break;

	  case 'A':
	    if (len > send - s) len = send - s;
	    {
		long end = len;
		char *t = s + len - 1;

		while (t >= s) {
		    if (*t != ' ' && *t != '\0') break;
		    t--; len--;
		}
		UNPACK_PUSH(infected_str_new(s, len, str));
		s += end;
	    }
	    break;

	  case 'Z':
	    {
		char *t = s;

		if (len > send-s) len = send-s;
		while (t < s+len && *t) t++;
		UNPACK_PUSH(infected_str_new(s, t-s, str));
		if (t < send) t++;
		s = star ? t : s+len;
	    }
	    break;

	  case 'a':
	    if (len > send - s) len = send - s;
	    UNPACK_PUSH(infected_str_new(s, len, str));
	    s += len;
	    break;

	  case 'b':
	    {
		VALUE bitstr;
		char *t;
		int bits;
		long i;

		if (p[-1] == '*' || len > (send - s) * 8)
		    len = (send - s) * 8;
		bits = 0;
		bitstr = rb_usascii_str_new(0, len);
                OBJ_INFECT(bitstr, str);
		t = RSTRING_PTR(bitstr);
		for (i=0; i<len; i++) {
		    if (i & 7) bits >>= 1;
		    else bits = (unsigned char)*s++;
		    *t++ = (bits & 1) ? '1' : '0';
		}
		UNPACK_PUSH(bitstr);
	    }
	    break;

	  case 'B':
	    {
		VALUE bitstr;
		char *t;
		int bits;
		long i;

		if (p[-1] == '*' || len > (send - s) * 8)
		    len = (send - s) * 8;
		bits = 0;
		bitstr = rb_usascii_str_new(0, len);
                OBJ_INFECT(bitstr, str);
		t = RSTRING_PTR(bitstr);
		for (i=0; i<len; i++) {
		    if (i & 7) bits <<= 1;
		    else bits = (unsigned char)*s++;
		    *t++ = (bits & 128) ? '1' : '0';
		}
		UNPACK_PUSH(bitstr);
	    }
	    break;

	  case 'h':
	    {
		VALUE bitstr;
		char *t;
		int bits;
		long i;

		if (p[-1] == '*' || len > (send - s) * 2)
		    len = (send - s) * 2;
		bits = 0;
		bitstr = rb_usascii_str_new(0, len);
                OBJ_INFECT(bitstr, str);
		t = RSTRING_PTR(bitstr);
		for (i=0; i<len; i++) {
		    if (i & 1)
			bits >>= 4;
		    else
			bits = (unsigned char)*s++;
		    *t++ = hexdigits[bits & 15];
		}
		UNPACK_PUSH(bitstr);
	    }
	    break;

	  case 'H':
	    {
		VALUE bitstr;
		char *t;
		int bits;
		long i;

		if (p[-1] == '*' || len > (send - s) * 2)
		    len = (send - s) * 2;
		bits = 0;
		bitstr = rb_usascii_str_new(0, len);
                OBJ_INFECT(bitstr, str);
		t = RSTRING_PTR(bitstr);
		for (i=0; i<len; i++) {
		    if (i & 1)
			bits <<= 4;
		    else
			bits = (unsigned char)*s++;
		    *t++ = hexdigits[(bits >> 4) & 15];
		}
		UNPACK_PUSH(bitstr);
	    }
	    break;

	  case 'c':
	    signed_p = 1;
	    integer_size = 1;
	    bigendian_p = BIGENDIAN_P(); /* not effective */
	    goto unpack_integer;

	  case 'C':
	    signed_p = 0;
	    integer_size = 1;
	    bigendian_p = BIGENDIAN_P(); /* not effective */
	    goto unpack_integer;

	  case 's':
	    signed_p = 1;
	    integer_size = NATINT_LEN(short, 2);
	    bigendian_p = BIGENDIAN_P();
	    goto unpack_integer;

	  case 'S':
	    signed_p = 0;
	    integer_size = NATINT_LEN(short, 2);
	    bigendian_p = BIGENDIAN_P();
	    goto unpack_integer;

	  case 'i':
	    signed_p = 1;
	    integer_size = (int)sizeof(int);
	    bigendian_p = BIGENDIAN_P();
	    goto unpack_integer;

	  case 'I':
	    signed_p = 0;
	    integer_size = (int)sizeof(int);
	    bigendian_p = BIGENDIAN_P();
	    goto unpack_integer;

	  case 'l':
	    signed_p = 1;
	    integer_size = NATINT_LEN(long, 4);
	    bigendian_p = BIGENDIAN_P();
	    goto unpack_integer;

	  case 'L':
	    signed_p = 0;
	    integer_size = NATINT_LEN(long, 4);
	    bigendian_p = BIGENDIAN_P();
	    goto unpack_integer;

	  case 'q':
	    signed_p = 1;
	    integer_size = NATINT_LEN_Q;
	    bigendian_p = BIGENDIAN_P();
	    goto unpack_integer;

	  case 'Q':
	    signed_p = 0;
	    integer_size = NATINT_LEN_Q;
	    bigendian_p = BIGENDIAN_P();
	    goto unpack_integer;

	  case 'j':
	    signed_p = 1;
	    integer_size = sizeof(intptr_t);
	    bigendian_p = BIGENDIAN_P();
	    goto unpack_integer;

	  case 'J':
	    signed_p = 0;
	    integer_size = sizeof(uintptr_t);
	    bigendian_p = BIGENDIAN_P();
	    goto unpack_integer;

	  case 'n':
	    signed_p = 0;
	    integer_size = 2;
	    bigendian_p = 1;
	    goto unpack_integer;

	  case 'N':
	    signed_p = 0;
	    integer_size = 4;
	    bigendian_p = 1;
	    goto unpack_integer;

	  case 'v':
	    signed_p = 0;
	    integer_size = 2;
	    bigendian_p = 0;
	    goto unpack_integer;

	  case 'V':
	    signed_p = 0;
	    integer_size = 4;
	    bigendian_p = 0;
	    goto unpack_integer;

	  unpack_integer:
	    if (explicit_endian) {
		bigendian_p = explicit_endian == '>';
	    }
            PACK_LENGTH_ADJUST_SIZE(integer_size);
            while (len-- > 0) {
                int flags = bigendian_p ? INTEGER_PACK_BIG_ENDIAN : INTEGER_PACK_LITTLE_ENDIAN;
                VALUE val;
                if (signed_p)
                    flags |= INTEGER_PACK_2COMP;
                val = rb_integer_unpack(s, integer_size, 1, 0, flags);
                UNPACK_PUSH(val);
                s += integer_size;
            }
            PACK_ITEM_ADJUST();
            break;

	  case 'f':
	  case 'F':
	    PACK_LENGTH_ADJUST_SIZE(sizeof(float));
	    while (len-- > 0) {
		float tmp;
		memcpy(&tmp, s, sizeof(float));
		s += sizeof(float);
		UNPACK_PUSH(DBL2NUM((double)tmp));
	    }
	    PACK_ITEM_ADJUST();
	    break;

	  case 'e':
	    PACK_LENGTH_ADJUST_SIZE(sizeof(float));
	    while (len-- > 0) {
		FLOAT_CONVWITH(tmp);
		memcpy(tmp.buf, s, sizeof(float));
		s += sizeof(float);
		VTOHF(tmp);
		UNPACK_PUSH(DBL2NUM(tmp.f));
	    }
	    PACK_ITEM_ADJUST();
	    break;

	  case 'E':
	    PACK_LENGTH_ADJUST_SIZE(sizeof(double));
	    while (len-- > 0) {
		DOUBLE_CONVWITH(tmp);
		memcpy(tmp.buf, s, sizeof(double));
		s += sizeof(double);
		VTOHD(tmp);
		UNPACK_PUSH(DBL2NUM(tmp.d));
	    }
	    PACK_ITEM_ADJUST();
	    break;

	  case 'D':
	  case 'd':
	    PACK_LENGTH_ADJUST_SIZE(sizeof(double));
	    while (len-- > 0) {
		double tmp;
		memcpy(&tmp, s, sizeof(double));
		s += sizeof(double);
		UNPACK_PUSH(DBL2NUM(tmp));
	    }
	    PACK_ITEM_ADJUST();
	    break;

	  case 'g':
	    PACK_LENGTH_ADJUST_SIZE(sizeof(float));
	    while (len-- > 0) {
		FLOAT_CONVWITH(tmp);
		memcpy(tmp.buf, s, sizeof(float));
		s += sizeof(float);
		NTOHF(tmp);
		UNPACK_PUSH(DBL2NUM(tmp.f));
	    }
	    PACK_ITEM_ADJUST();
	    break;

	  case 'G':
	    PACK_LENGTH_ADJUST_SIZE(sizeof(double));
	    while (len-- > 0) {
		DOUBLE_CONVWITH(tmp);
		memcpy(tmp.buf, s, sizeof(double));
		s += sizeof(double);
		NTOHD(tmp);
		UNPACK_PUSH(DBL2NUM(tmp.d));
	    }
	    PACK_ITEM_ADJUST();
	    break;

	  case 'U':
	    if (len > send - s) len = send - s;
	    while (len > 0 && s < send) {
		long alen = send - s;
		unsigned long l;

		l = utf8_to_uv(s, &alen);
		s += alen; len--;
		UNPACK_PUSH(ULONG2NUM(l));
	    }
	    break;

	  case 'u':
	    {
		VALUE buf = infected_str_new(0, (send - s)*3/4, str);
		char *ptr = RSTRING_PTR(buf);
		long total = 0;

		while (s < send && (unsigned char)*s > ' ' && (unsigned char)*s < 'a') {
		    long a,b,c,d;
		    char hunk[3];

		    len = ((unsigned char)*s++ - ' ') & 077;

		    total += len;
		    if (total > RSTRING_LEN(buf)) {
			len -= total - RSTRING_LEN(buf);
			total = RSTRING_LEN(buf);
		    }

		    while (len > 0) {
			long mlen = len > 3 ? 3 : len;

			if (s < send && (unsigned char)*s >= ' ' && (unsigned char)*s < 'a')
			    a = ((unsigned char)*s++ - ' ') & 077;
			else
			    a = 0;
			if (s < send && (unsigned char)*s >= ' ' && (unsigned char)*s < 'a')
			    b = ((unsigned char)*s++ - ' ') & 077;
			else
			    b = 0;
			if (s < send && (unsigned char)*s >= ' ' && (unsigned char)*s < 'a')
			    c = ((unsigned char)*s++ - ' ') & 077;
			else
			    c = 0;
			if (s < send && (unsigned char)*s >= ' ' && (unsigned char)*s < 'a')
			    d = ((unsigned char)*s++ - ' ') & 077;
			else
			    d = 0;
			hunk[0] = (char)(a << 2 | b >> 4);
			hunk[1] = (char)(b << 4 | c >> 2);
			hunk[2] = (char)(c << 6 | d);
			memcpy(ptr, hunk, mlen);
			ptr += mlen;
			len -= mlen;
		    }
		    if (s < send && (unsigned char)*s != '\r' && *s != '\n')
			s++;	/* possible checksum byte */
		    if (s < send && *s == '\r') s++;
		    if (s < send && *s == '\n') s++;
		}

		rb_str_set_len(buf, total);
		UNPACK_PUSH(buf);
	    }
	    break;

	  case 'm':
	    {
		VALUE buf = infected_str_new(0, (send - s + 3)*3/4, str); /* +3 is for skipping paddings */
		char *ptr = RSTRING_PTR(buf);
		int a = -1,b = -1,c = 0,d = 0;
		static signed char b64_xtable[256];

		if (b64_xtable['/'] <= 0) {
		    int i;

		    for (i = 0; i < 256; i++) {
			b64_xtable[i] = -1;
		    }
		    for (i = 0; i < 64; i++) {
			b64_xtable[(unsigned char)b64_table[i]] = (char)i;
		    }
		}
		if (len == 0) {
		    while (s < send) {
			a = b = c = d = -1;
			a = b64_xtable[(unsigned char)*s++];
			if (s >= send || a == -1) rb_raise(rb_eArgError, "invalid base64");
			b = b64_xtable[(unsigned char)*s++];
			if (s >= send || b == -1) rb_raise(rb_eArgError, "invalid base64");
			if (*s == '=') {
			    if (s + 2 == send && *(s + 1) == '=') break;
			    rb_raise(rb_eArgError, "invalid base64");
			}
			c = b64_xtable[(unsigned char)*s++];
			if (s >= send || c == -1) rb_raise(rb_eArgError, "invalid base64");
			if (s + 1 == send && *s == '=') break;
			d = b64_xtable[(unsigned char)*s++];
			if (d == -1) rb_raise(rb_eArgError, "invalid base64");
			*ptr++ = castchar(a << 2 | b >> 4);
			*ptr++ = castchar(b << 4 | c >> 2);
			*ptr++ = castchar(c << 6 | d);
		    }
		    if (c == -1) {
			*ptr++ = castchar(a << 2 | b >> 4);
			if (b & 0xf) rb_raise(rb_eArgError, "invalid base64");
		    }
		    else if (d == -1) {
			*ptr++ = castchar(a << 2 | b >> 4);
			*ptr++ = castchar(b << 4 | c >> 2);
			if (c & 0x3) rb_raise(rb_eArgError, "invalid base64");
		    }
		}
		else {
		    while (s < send) {
			a = b = c = d = -1;
			while ((a = b64_xtable[(unsigned char)*s]) == -1 && s < send) {s++;}
			if (s >= send) break;
			s++;
			while ((b = b64_xtable[(unsigned char)*s]) == -1 && s < send) {s++;}
			if (s >= send) break;
			s++;
			while ((c = b64_xtable[(unsigned char)*s]) == -1 && s < send) {if (*s == '=') break; s++;}
			if (*s == '=' || s >= send) break;
			s++;
			while ((d = b64_xtable[(unsigned char)*s]) == -1 && s < send) {if (*s == '=') break; s++;}
			if (*s == '=' || s >= send) break;
			s++;
			*ptr++ = castchar(a << 2 | b >> 4);
			*ptr++ = castchar(b << 4 | c >> 2);
			*ptr++ = castchar(c << 6 | d);
			a = -1;
		    }
		    if (a != -1 && b != -1) {
			if (c == -1)
			    *ptr++ = castchar(a << 2 | b >> 4);
			else {
			    *ptr++ = castchar(a << 2 | b >> 4);
			    *ptr++ = castchar(b << 4 | c >> 2);
			}
		    }
		}
		rb_str_set_len(buf, ptr - RSTRING_PTR(buf));
		UNPACK_PUSH(buf);
	    }
	    break;

	  case 'M':
	    {
		VALUE buf = infected_str_new(0, send - s, str);
		char *ptr = RSTRING_PTR(buf), *ss = s;
		int csum = 0;
		int c1, c2;

		while (s < send) {
		    if (*s == '=') {
			if (++s == send) break;
			if (s+1 < send && *s == '\r' && *(s+1) == '\n')
			    s++;
			if (*s != '\n') {
			    if ((c1 = hex2num(*s)) == -1) break;
			    if (++s == send) break;
			    if ((c2 = hex2num(*s)) == -1) break;
			    csum |= *ptr++ = castchar(c1 << 4 | c2);
			}
		    }
		    else {
			csum |= *ptr++ = *s;
		    }
		    s++;
		    ss = s;
		}
		rb_str_set_len(buf, ptr - RSTRING_PTR(buf));
		rb_str_buf_cat(buf, ss, send-ss);
		csum = ISASCII(csum) ? ENC_CODERANGE_7BIT : ENC_CODERANGE_VALID;
		ENCODING_CODERANGE_SET(buf, rb_ascii8bit_encindex(), csum);
		UNPACK_PUSH(buf);
	    }
	    break;

	  case '@':
	    if (len > RSTRING_LEN(str))
		rb_raise(rb_eArgError, "@ outside of string");
	    s = RSTRING_PTR(str) + len;
	    break;

	  case 'X':
	    if (len > s - RSTRING_PTR(str))
		rb_raise(rb_eArgError, "X outside of string");
	    s -= len;
	    break;

	  case 'x':
	    if (len > send - s)
		rb_raise(rb_eArgError, "x outside of string");
	    s += len;
	    break;

	  case 'P':
	    if (sizeof(char *) <= (size_t)(send - s)) {
		VALUE tmp = Qnil;
		char *t;

		memcpy(&t, s, sizeof(char *));
		s += sizeof(char *);

		if (t) {
		    VALUE a;
		    const VALUE *p, *pend;

		    if (!(a = str_associated(str))) {
			rb_raise(rb_eArgError, "no associated pointer");
		    }
		    p = RARRAY_CONST_PTR(a);
		    pend = p + RARRAY_LEN(a);
		    while (p < pend) {
			if (RB_TYPE_P(*p, T_STRING) && RSTRING_PTR(*p) == t) {
			    if (len < RSTRING_LEN(*p)) {
				tmp = rb_tainted_str_new(t, len);
				str_associate(tmp, a);
			    }
			    else {
				tmp = *p;
			    }
			    break;
			}
			p++;
		    }
		    if (p == pend) {
			rb_raise(rb_eArgError, "non associated pointer");
		    }
		}
		UNPACK_PUSH(tmp);
	    }
	    break;

	  case 'p':
	    if (len > (long)((send - s) / sizeof(char *)))
		len = (send - s) / sizeof(char *);
	    while (len-- > 0) {
		if ((size_t)(send - s) < sizeof(char *))
		    break;
		else {
		    VALUE tmp = Qnil;
		    char *t;

		    memcpy(&t, s, sizeof(char *));
		    s += sizeof(char *);

		    if (t) {
			VALUE a;
			const VALUE *p, *pend;

			if (!(a = str_associated(str))) {
			    rb_raise(rb_eArgError, "no associated pointer");
			}
			p = RARRAY_CONST_PTR(a);
			pend = p + RARRAY_LEN(a);
			while (p < pend) {
			    if (RB_TYPE_P(*p, T_STRING) && RSTRING_PTR(*p) == t) {
				tmp = *p;
				break;
			    }
			    p++;
			}
			if (p == pend) {
			    rb_raise(rb_eArgError, "non associated pointer");
			}
		    }
		    UNPACK_PUSH(tmp);
		}
	    }
	    break;

	  case 'w':
	    {
                char *s0 = s;
                while (len > 0 && s < send) {
                    if (*s & 0x80) {
                        s++;
                    }
                    else {
                        s++;
                        UNPACK_PUSH(rb_integer_unpack(s0, s-s0, 1, 1, INTEGER_PACK_BIG_ENDIAN));
                        len--;
                        s0 = s;
                    }
                }
	    }
	    break;

	  default:
            unknown_directive("unpack", type, fmt);
	    break;
	}
    }

    return ary;
}

/*
 *  call-seq:
 *     str.unpack(format)    ->  anArray
 *
 *  Decodes <i>str</i> (which may contain binary data) according to the
 *  format string, returning an array of each value extracted. The
 *  format string consists of a sequence of single-character directives,
 *  summarized in the table at the end of this entry.
 *  Each directive may be followed
 *  by a number, indicating the number of times to repeat with this
 *  directive. An asterisk (``<code>*</code>'') will use up all
 *  remaining elements. The directives <code>sSiIlL</code> may each be
 *  followed by an underscore (``<code>_</code>'') or
 *  exclamation mark (``<code>!</code>'') to use the underlying
 *  platform's native size for the specified type; otherwise, it uses a
 *  platform-independent consistent size. Spaces are ignored in the
 *  format string. See also String#unpack1,  Array#pack.
 *
 *     "abc \0\0abc \0\0".unpack('A6Z6')   #=> ["abc", "abc "]
 *     "abc \0\0".unpack('a3a3')           #=> ["abc", " \000\000"]
 *     "abc \0abc \0".unpack('Z*Z*')       #=> ["abc ", "abc "]
 *     "aa".unpack('b8B8')                 #=> ["10000110", "01100001"]
 *     "aaa".unpack('h2H2c')               #=> ["16", "61", 97]
 *     "\xfe\xff\xfe\xff".unpack('sS')     #=> [-2, 65534]
 *     "now=20is".unpack('M*')             #=> ["now is"]
 *     "whole".unpack('xax2aX2aX1aX2a')    #=> ["h", "e", "l", "l", "o"]
 *
 *  This table summarizes the various formats and the Ruby classes
 *  returned by each.
 *
 *   Integer       |         |
 *   Directive     | Returns | Meaning
 *   ------------------------------------------------------------------
 *   C             | Integer | 8-bit unsigned (unsigned char)
 *   S             | Integer | 16-bit unsigned, native endian (uint16_t)
 *   L             | Integer | 32-bit unsigned, native endian (uint32_t)
 *   Q             | Integer | 64-bit unsigned, native endian (uint64_t)
 *   J             | Integer | pointer width unsigned, native endian (uintptr_t)
 *                 |         |
 *   c             | Integer | 8-bit signed (signed char)
 *   s             | Integer | 16-bit signed, native endian (int16_t)
 *   l             | Integer | 32-bit signed, native endian (int32_t)
 *   q             | Integer | 64-bit signed, native endian (int64_t)
 *   j             | Integer | pointer width signed, native endian (intptr_t)
 *                 |         |
 *   S_ S!         | Integer | unsigned short, native endian
 *   I I_ I!       | Integer | unsigned int, native endian
 *   L_ L!         | Integer | unsigned long, native endian
 *   Q_ Q!         | Integer | unsigned long long, native endian (ArgumentError
 *                 |         | if the platform has no long long type.)
 *   J!            | Integer | uintptr_t, native endian (same with J)
 *                 |         |
 *   s_ s!         | Integer | signed short, native endian
 *   i i_ i!       | Integer | signed int, native endian
 *   l_ l!         | Integer | signed long, native endian
 *   q_ q!         | Integer | signed long long, native endian (ArgumentError
 *                 |         | if the platform has no long long type.)
 *   j!            | Integer | intptr_t, native endian (same with j)
 *                 |         |
 *   S> s> S!> s!> | Integer | same as the directives without ">" except
 *   L> l> L!> l!> |         | big endian
 *   I!> i!>       |         |
 *   Q> q> Q!> q!> |         | "S>" is same as "n"
 *   J> j> J!> j!> |         | "L>" is same as "N"
 *                 |         |
 *   S< s< S!< s!< | Integer | same as the directives without "<" except
 *   L< l< L!< l!< |         | little endian
 *   I!< i!<       |         |
 *   Q< q< Q!< q!< |         | "S<" is same as "v"
 *   J< j< J!< j!< |         | "L<" is same as "V"
 *                 |         |
 *   n             | Integer | 16-bit unsigned, network (big-endian) byte order
 *   N             | Integer | 32-bit unsigned, network (big-endian) byte order
 *   v             | Integer | 16-bit unsigned, VAX (little-endian) byte order
 *   V             | Integer | 32-bit unsigned, VAX (little-endian) byte order
 *                 |         |
 *   U             | Integer | UTF-8 character
 *   w             | Integer | BER-compressed integer (see Array.pack)
 *
 *   Float        |         |
 *   Directive    | Returns | Meaning
 *   -----------------------------------------------------------------
 *   D d          | Float   | double-precision, native format
 *   F f          | Float   | single-precision, native format
 *   E            | Float   | double-precision, little-endian byte order
 *   e            | Float   | single-precision, little-endian byte order
 *   G            | Float   | double-precision, network (big-endian) byte order
 *   g            | Float   | single-precision, network (big-endian) byte order
 *
 *   String       |         |
 *   Directive    | Returns | Meaning
 *   -----------------------------------------------------------------
 *   A            | String  | arbitrary binary string (remove trailing nulls and ASCII spaces)
 *   a            | String  | arbitrary binary string
 *   Z            | String  | null-terminated string
 *   B            | String  | bit string (MSB first)
 *   b            | String  | bit string (LSB first)
 *   H            | String  | hex string (high nibble first)
 *   h            | String  | hex string (low nibble first)
 *   u            | String  | UU-encoded string
 *   M            | String  | quoted-printable, MIME encoding (see RFC2045)
 *   m            | String  | base64 encoded string (RFC 2045) (default)
 *                |         | base64 encoded string (RFC 4648) if followed by 0
 *   P            | String  | pointer to a structure (fixed-length string)
 *   p            | String  | pointer to a null-terminated string
 *
 *   Misc.        |         |
 *   Directive    | Returns | Meaning
 *   -----------------------------------------------------------------
 *   @            | ---     | skip to the offset given by the length argument
 *   X            | ---     | skip backward one byte
 *   x            | ---     | skip forward one byte
 *
 *  HISTORY
 *
 *  * J, J! j, and j! are available since Ruby 2.3.
 *  * Q_, Q!, q_, and q! are available since Ruby 2.1.
 *  * I!<, i!<, I!>, and i!> are available since Ruby 1.9.3.
 */

static VALUE
pack_unpack(VALUE str, VALUE fmt)
{
    int mode = rb_block_given_p() ? UNPACK_BLOCK : UNPACK_ARRAY;
    return pack_unpack_internal(str, fmt, mode);
}

/*
 *  call-seq:
 *     str.unpack1(format)    ->  obj
 *
 *  Decodes <i>str</i> (which may contain binary data) according to the
 *  format string, returning the first value extracted.
 *  See also String#unpack, Array#pack.
 *
 *  Contrast with String#unpack:
 *
 *     "abc \0\0abc \0\0".unpack('A6Z6')   #=> ["abc", "abc "]
 *     "abc \0\0abc \0\0".unpack1('A6Z6')  #=> "abc"
 *
 *  In that case data would be lost but often it's the case that the array
 *  only holds one value, especially when unpacking binary data. For instance:
 *
 *  "\xff\x00\x00\x00".unpack("l")         #=>  [255]
 *  "\xff\x00\x00\x00".unpack1("l")        #=>  255
 *
 *  Thus unpack1 is convenient, makes clear the intention and signals
 *  the expected return value to those reading the code.
 */

static VALUE
pack_unpack1(VALUE str, VALUE fmt)
{
    return pack_unpack_internal(str, fmt, UNPACK_1);
}

int
rb_uv_to_utf8(char buf[6], unsigned long uv)
{
    if (uv <= 0x7f) {
	buf[0] = (char)uv;
	return 1;
    }
    if (uv <= 0x7ff) {
	buf[0] = castchar(((uv>>6)&0xff)|0xc0);
	buf[1] = castchar((uv&0x3f)|0x80);
	return 2;
    }
    if (uv <= 0xffff) {
	buf[0] = castchar(((uv>>12)&0xff)|0xe0);
	buf[1] = castchar(((uv>>6)&0x3f)|0x80);
	buf[2] = castchar((uv&0x3f)|0x80);
	return 3;
    }
    if (uv <= 0x1fffff) {
	buf[0] = castchar(((uv>>18)&0xff)|0xf0);
	buf[1] = castchar(((uv>>12)&0x3f)|0x80);
	buf[2] = castchar(((uv>>6)&0x3f)|0x80);
	buf[3] = castchar((uv&0x3f)|0x80);
	return 4;
    }
    if (uv <= 0x3ffffff) {
	buf[0] = castchar(((uv>>24)&0xff)|0xf8);
	buf[1] = castchar(((uv>>18)&0x3f)|0x80);
	buf[2] = castchar(((uv>>12)&0x3f)|0x80);
	buf[3] = castchar(((uv>>6)&0x3f)|0x80);
	buf[4] = castchar((uv&0x3f)|0x80);
	return 5;
    }
    if (uv <= 0x7fffffff) {
	buf[0] = castchar(((uv>>30)&0xff)|0xfc);
	buf[1] = castchar(((uv>>24)&0x3f)|0x80);
	buf[2] = castchar(((uv>>18)&0x3f)|0x80);
	buf[3] = castchar(((uv>>12)&0x3f)|0x80);
	buf[4] = castchar(((uv>>6)&0x3f)|0x80);
	buf[5] = castchar((uv&0x3f)|0x80);
	return 6;
    }
    rb_raise(rb_eRangeError, "pack(U): value out of range");

    UNREACHABLE_RETURN(Qnil);
}

static const unsigned long utf8_limits[] = {
    0x0,			/* 1 */
    0x80,			/* 2 */
    0x800,			/* 3 */
    0x10000,			/* 4 */
    0x200000,			/* 5 */
    0x4000000,			/* 6 */
    0x80000000,			/* 7 */
};

static unsigned long
utf8_to_uv(const char *p, long *lenp)
{
    int c = *p++ & 0xff;
    unsigned long uv = c;
    long n;

    if (!(uv & 0x80)) {
	*lenp = 1;
        return uv;
    }
    if (!(uv & 0x40)) {
	*lenp = 1;
	rb_raise(rb_eArgError, "malformed UTF-8 character");
    }

    if      (!(uv & 0x20)) { n = 2; uv &= 0x1f; }
    else if (!(uv & 0x10)) { n = 3; uv &= 0x0f; }
    else if (!(uv & 0x08)) { n = 4; uv &= 0x07; }
    else if (!(uv & 0x04)) { n = 5; uv &= 0x03; }
    else if (!(uv & 0x02)) { n = 6; uv &= 0x01; }
    else {
	*lenp = 1;
	rb_raise(rb_eArgError, "malformed UTF-8 character");
    }
    if (n > *lenp) {
	rb_raise(rb_eArgError, "malformed UTF-8 character (expected %ld bytes, given %ld bytes)",
		 n, *lenp);
    }
    *lenp = n--;
    if (n != 0) {
	while (n--) {
	    c = *p++ & 0xff;
	    if ((c & 0xc0) != 0x80) {
		*lenp -= n + 1;
		rb_raise(rb_eArgError, "malformed UTF-8 character");
	    }
	    else {
		c &= 0x3f;
		uv = uv << 6 | c;
	    }
	}
    }
    n = *lenp - 1;
    if (uv < utf8_limits[n]) {
	rb_raise(rb_eArgError, "redundant UTF-8 sequence");
    }
    return uv;
}

void
Init_pack(void)
{
    rb_define_method(rb_cArray, "pack", pack_pack, -1);
    rb_define_method(rb_cString, "unpack", pack_unpack, 1);
    rb_define_method(rb_cString, "unpack1", pack_unpack1, 1);

    id_associated = rb_make_internal_id();
}