/* 
  Copyright 2008-2010 LibRaw LLC (info@libraw.org)

LibRaw is free software; you can redistribute it and/or modify
it under the terms of the one of three licenses as you choose:

1. GNU LESSER GENERAL PUBLIC LICENSE version 2.1
   (See file LICENSE.LGPL provided in LibRaw distribution archive for details).

2. COMMON DEVELOPMENT AND DISTRIBUTION LICENSE (CDDL) Version 1.0
   (See file LICENSE.CDDL provided in LibRaw distribution archive for details).

3. LibRaw Software License 27032010
   (See file LICENSE.LibRaw.pdf provided in LibRaw distribution archive for details).

   This file is generated from Dave Coffin's dcraw.c
   dcraw.c -- Dave Coffin's raw photo decoder
   Copyright 1997-2010 by Dave Coffin, dcoffin a cybercom o net

   Look into dcraw homepage (probably http://cybercom.net/~dcoffin/dcraw/)
   for more information
*/

#include <math.h>
#define CLASS LibRaw::
#include "libraw/libraw_types.h"
#define LIBRAW_LIBRARY_BUILD
#define LIBRAW_IO_REDEFINED
#include "libraw/libraw.h"
#include "internal/defines.h"
#include "internal/var_defines.h"
#include "internal/libraw_bytebuffer.h"

#ifdef ANDROID
#include <asm/byteorder.h> 
void swab(const void *from, void*to, ssize_t n) 
{ 
	if (n < 0) 
		return; 
 		 
	for (ssize_t i = 0; i < (n/2)*2; i += 2) 
	{
		*((uint16_t*)to+i) = __arch__swab16(*((uint16_t*)from+i)); 
	}
} 
#endif

#ifndef __GLIBC__
char *my_memmem (char *haystack, size_t haystacklen,
	      char *needle, size_t needlelen)
{
  char *c;
  for (c = haystack; c <= haystack + haystacklen - needlelen; c++)
    if (!memcmp (c, needle, needlelen))
      return c;
  return 0;
}
#define memmem my_memmem
#endif


ushort CLASS sget2 (uchar *s)
{
  if (order == 0x4949)		/* "II" means little-endian */
    return s[0] | s[1] << 8;
  else				/* "MM" means big-endian */
    return s[0] << 8 | s[1];
}

ushort CLASS get2()
{
  uchar str[2] = { 0xff,0xff };
  fread (str, 1, 2, ifp);
  return sget2(str);
}

unsigned CLASS sget4 (uchar *s)
{
  if (order == 0x4949)
    return s[0] | s[1] << 8 | s[2] << 16 | s[3] << 24;
  else
    return s[0] << 24 | s[1] << 16 | s[2] << 8 | s[3];
}
#define sget4(s) sget4((uchar *)s)

unsigned CLASS get4()
{
  uchar str[4] = { 0xff,0xff,0xff,0xff };
  fread (str, 1, 4, ifp);
  return sget4(str);
}

unsigned CLASS getint (int type)
{
  return type == 3 ? get2() : get4();
}

float CLASS int_to_float (int i)
{
  union { int i; float f; } u;
  u.i = i;
  return u.f;
}

double CLASS getreal (int type)
{
  union { char c[8]; double d; } u;
  int i, rev;

  switch (type) {
    case 3: return (unsigned short) get2();
    case 4: return (unsigned int) get4();
    case 5:  u.d = (unsigned int) get4();
      return u.d / (unsigned int) get4();
    case 8: return (signed short) get2();
    case 9: return (signed int) get4();
    case 10: u.d = (signed int) get4();
      return u.d / (signed int) get4();
    case 11: return int_to_float (get4());
    case 12:
      rev = 7 * ((order == 0x4949) == (ntohs(0x1234) == 0x1234));
      for (i=0; i < 8; i++)
	u.c[i ^ rev] = fgetc(ifp);
      return u.d;
    default: return fgetc(ifp);
  }
}

void CLASS read_shorts (ushort *pixel, int count)
{
  if (fread (pixel, 2, count, ifp) < count) derror();
  if ((order == 0x4949) == (ntohs(0x1234) == 0x1234))
      swab ((char*)pixel, (char*)pixel, count*2);
}

#ifdef LIBRAW_LIBRARY_BUILD
#define RBAYER(x,y) raw_image[(x)*raw_width+(y)]
#define RRBAYER(x,y) raw_image[((x)+top_margin)*raw_width+(y)+left_margin]
#define CBAYER(x,y) color_image[((x)+top_margin)*raw_width+(y)+left_margin][FC((x),(y))]
#else
#define CBAYER(x,y) BAYER((x),(y))
#define RBAYER(x,y) BAYER((x),(y))
#define RRBAYER(x,y) BAYER((x),(y))
#endif

void CLASS canon_600_fixed_wb (int temp)
{
  static const short mul[4][5] = {
    {  667, 358,397,565,452 },
    {  731, 390,367,499,517 },
    { 1119, 396,348,448,537 },
    { 1399, 485,431,508,688 } };
  int lo, hi, i;
  float frac=0;

  for (lo=4; --lo; )
    if (*mul[lo] <= temp) break;
  for (hi=0; hi < 3; hi++)
    if (*mul[hi] >= temp) break;
  if (lo != hi)
    frac = (float) (temp - *mul[lo]) / (*mul[hi] - *mul[lo]);
  for (i=1; i < 5; i++)
    pre_mul[i-1] = 1 / (frac * mul[hi][i] + (1-frac) * mul[lo][i]);
#ifdef LIBRAW_LIBRARY_BUILD
  color_flags.pre_mul_state = LIBRAW_COLORSTATE_CONST;
#endif
}

/* Return values:  0 = white  1 = near white  2 = not white */
int CLASS canon_600_color (int ratio[2], int mar)
{
  int clipped=0, target, miss;

  if (flash_used) {
    if (ratio[1] < -104)
      { ratio[1] = -104; clipped = 1; }
    if (ratio[1] >   12)
      { ratio[1] =   12; clipped = 1; }
  } else {
    if (ratio[1] < -264 || ratio[1] > 461) return 2;
    if (ratio[1] < -50)
      { ratio[1] = -50; clipped = 1; }
    if (ratio[1] > 307)
      { ratio[1] = 307; clipped = 1; }
  }
  target = flash_used || ratio[1] < 197
	? -38 - (398 * ratio[1] >> 10)
	: -123 + (48 * ratio[1] >> 10);
  if (target - mar <= ratio[0] &&
      target + 20  >= ratio[0] && !clipped) return 0;
  miss = target - ratio[0];
  if (abs(miss) >= mar*4) return 2;
  if (miss < -20) miss = -20;
  if (miss > mar) miss = mar;
  ratio[0] = target - miss;
  return 1;
}

void CLASS canon_600_auto_wb()
{
  int mar, row, col, i, j, st, count[] = { 0,0 };
  int test[8], total[2][8], ratio[2][2], stat[2];

  memset (&total, 0, sizeof total);
  i = canon_ev + 0.5;
  if      (i < 10) mar = 150;
  else if (i > 12) mar = 20;
  else mar = 280 - 20 * i;
  if (flash_used) mar = 80;
  for (row=14; row < height-14; row+=4)
    for (col=10; col < width; col+=2) {
      for (i=0; i < 8; i++)
	test[(i & 4) + FC(row+(i >> 1),col+(i & 1))] =
		    BAYER(row+(i >> 1),col+(i & 1));
      for (i=0; i < 8; i++)
	if (test[i] < 150 || test[i] > 1500) goto next;
      for (i=0; i < 4; i++)
	if (abs(test[i] - test[i+4]) > 50) goto next;
      for (i=0; i < 2; i++) {
	for (j=0; j < 4; j+=2)
	  ratio[i][j >> 1] = ((test[i*4+j+1]-test[i*4+j]) << 10) / test[i*4+j];
	stat[i] = canon_600_color (ratio[i], mar);
      }
      if ((st = stat[0] | stat[1]) > 1) goto next;
      for (i=0; i < 2; i++)
	if (stat[i])
	  for (j=0; j < 2; j++)
	    test[i*4+j*2+1] = test[i*4+j*2] * (0x400 + ratio[i][j]) >> 10;
      for (i=0; i < 8; i++)
	total[st][i] += test[i];
      count[st]++;
next: ;
    }
  if (count[0] | count[1]) {
    st = count[0]*200 < count[1];
    for (i=0; i < 4; i++)
      pre_mul[i] = 1.0 / (total[st][i] + total[st][i+4]);
#ifdef LIBRAW_LIBRARY_BUILD
    color_flags.pre_mul_state = LIBRAW_COLORSTATE_CALCULATED;
#endif
  }
}

void CLASS canon_600_coeff()
{
  static const short table[6][12] = {
    { -190,702,-1878,2390,   1861,-1349,905,-393, -432,944,2617,-2105  },
    { -1203,1715,-1136,1648, 1388,-876,267,245,  -1641,2153,3921,-3409 },
    { -615,1127,-1563,2075,  1437,-925,509,3,     -756,1268,2519,-2007 },
    { -190,702,-1886,2398,   2153,-1641,763,-251, -452,964,3040,-2528  },
    { -190,702,-1878,2390,   1861,-1349,905,-393, -432,944,2617,-2105  },
    { -807,1319,-1785,2297,  1388,-876,769,-257,  -230,742,2067,-1555  } };
  int t=0, i, c;
  float mc, yc;

  mc = pre_mul[1] / pre_mul[2];
  yc = pre_mul[3] / pre_mul[2];
  if (mc > 1 && mc <= 1.28 && yc < 0.8789) t=1;
  if (mc > 1.28 && mc <= 2) {
    if  (yc < 0.8789) t=3;
    else if (yc <= 2) t=4;
  }
  if (flash_used) t=5;
  for (raw_color = i=0; i < 3; i++)
    FORCC rgb_cam[i][c] = table[t][i*4 + c] / 1024.0;
#ifdef LIBRAW_LIBRARY_BUILD
  color_flags.rgb_cam_state = LIBRAW_COLORSTATE_CALCULATED;
#endif
}

void CLASS canon_600_load_raw()
{
  uchar  data[1120], *dp;
  ushort pixel[896], *pix;
  int irow, row, col, val;
  static const short mul[4][2] =
  { { 1141,1145 }, { 1128,1109 }, { 1178,1149 }, { 1128,1109 } };

  for (irow=row=0; irow < height; irow++) {
    if (fread (data, 1, raw_width*5/4, ifp) < raw_width*5/4) derror();
    for (dp=data, pix=pixel; dp < data+1120; dp+=10, pix+=8) {
      pix[0] = (dp[0] << 2) + (dp[1] >> 6    );
      pix[1] = (dp[2] << 2) + (dp[1] >> 4 & 3);
      pix[2] = (dp[3] << 2) + (dp[1] >> 2 & 3);
      pix[3] = (dp[4] << 2) + (dp[1]      & 3);
      pix[4] = (dp[5] << 2) + (dp[9]      & 3);
      pix[5] = (dp[6] << 2) + (dp[9] >> 2 & 3);
      pix[6] = (dp[7] << 2) + (dp[9] >> 4 & 3);
      pix[7] = (dp[8] << 2) + (dp[9] >> 6    );
    }
#ifndef LIBRAW_LIBRARY_BUILD
    for (col=0; col < width; col++)
        {
            BAYER(row,col) = pixel[col];
        }
#else
    for (col=0; col < raw_width; col++)
            RBAYER(row,col) = pixel[col];
#endif

    for (col=width; col < raw_width; col++)
        {
            black += pixel[col];
        }
    if ((row+=2) > height) row = 1;
  }
  if (raw_width > width)
    black = black / ((raw_width - width) * height) - 4;
#ifndef LIBRAW_LIBRARY_BUILD
  for (row=0; row < height; row++)
    for (col=0; col < width; col++) {
      if ((val = BAYER(row,col) - black) < 0) val = 0;
      val = val * mul[row & 3][col & 1] >> 9;
      BAYER(row,col) = val;
    }
#else
  for (row=0; row < height; row++)
    for (col=0; col < raw_width; col++) {
      if ((val = RBAYER(row,col) - black) < 0) val = 0;
      val = val * mul[row & 3][col & 1] >> 9;
      RBAYER(row,col) = val;
    }

#endif
  canon_600_fixed_wb(1311);
  canon_600_auto_wb();
  canon_600_coeff();
  maximum = (0x3ff - black) * 1109 >> 9;
  black = 0;
}

void CLASS remove_zeroes()
{
  unsigned row, col, tot, n, r, c;

#ifdef LIBRAW_LIBRARY_BUILD
  RUN_CALLBACK(LIBRAW_PROGRESS_REMOVE_ZEROES,0,2);
#endif
  for (row=0; row < height; row++)
    for (col=0; col < width; col++)
      if (BAYER(row,col) == 0) {
	tot = n = 0;
	for (r = row-2; r <= row+2; r++)
	  for (c = col-2; c <= col+2; c++)
	    if (r < height && c < width &&
		FC(r,c) == FC(row,col) && BAYER(r,c))
	      tot += (n++,BAYER(r,c));
	if (n) BAYER(row,col) = tot/n;
      }
#ifdef LIBRAW_LIBRARY_BUILD
  RUN_CALLBACK(LIBRAW_PROGRESS_REMOVE_ZEROES,1,2);
#endif
}

int CLASS canon_s2is()
{
  unsigned row;

  for (row=0; row < 100; row++) {
    fseek (ifp, row*3340 + 3284, SEEK_SET);
    if (getc(ifp) > 15) return 1;
  }
  return 0;
}

/*
   getbits(-1) initializes the buffer
   getbits(n) where 0 <= n <= 25 returns an n-bit integer
 */

unsigned CLASS getbithuff (int nbits, ushort *huff)
{
#ifdef LIBRAW_NOTHREADS
  static unsigned bitbuf=0;
  static int vbits=0, reset=0;
#else
#define bitbuf tls->getbits.bitbuf
#define vbits  tls->getbits.vbits
#define reset  tls->getbits.reset
#endif
  unsigned c;

  if (nbits == -1)
    return bitbuf = vbits = reset = 0;
  if (nbits == 0 || vbits < 0) return 0;
  while (!reset && vbits < nbits && (c = fgetc(ifp)) != EOF &&
    !(reset = zero_after_ff && c == 0xff && fgetc(ifp))) {
    bitbuf = (bitbuf << 8) + (uchar) c;
    vbits += 8;
  }
  c = bitbuf << (32-vbits) >> (32-nbits);
  if (huff) {
    vbits -= huff[c] >> 8;
    c = (uchar) huff[c];
  } else
    vbits -= nbits;
  if (vbits < 0) derror();
  return c;
#ifndef LIBRAW_NOTHREADS
#undef bitbuf
#undef vbits
#undef reset
#endif
}

#define getbits(n) getbithuff(n,0)
#define gethuff(h) getbithuff(*h,h+1)

/*
   Construct a decode tree according to the specification in *source.
   The first 16 bytes specify how many codes should be 1-bit, 2-bit
   3-bit, etc.  Bytes after that are the leaf values.

   For example, if the source is

    { 0,1,4,2,3,1,2,0,0,0,0,0,0,0,0,0,
      0x04,0x03,0x05,0x06,0x02,0x07,0x01,0x08,0x09,0x00,0x0a,0x0b,0xff  },

   then the code is

	00		0x04
	010		0x03
	011		0x05
	100		0x06
	101		0x02
	1100		0x07
	1101		0x01
	11100		0x08
	11101		0x09
	11110		0x00
	111110		0x0a
	1111110		0x0b
	1111111		0xff
 */
ushort * CLASS make_decoder_ref (const uchar **source)
{
  int max, len, h, i, j;
  const uchar *count;
  ushort *huff;

  count = (*source += 16) - 17;
  for (max=16; max && !count[max]; max--);
  huff = (ushort *) calloc (1 + (1 << max), sizeof *huff);
  merror (huff, "make_decoder()");
  huff[0] = max;
  for (h=len=1; len <= max; len++)
    for (i=0; i < count[len]; i++, ++*source)
      for (j=0; j < 1 << (max-len); j++)
	if (h <= 1 << max)
	  huff[h++] = len << 8 | **source;
  return huff;
}

ushort * CLASS make_decoder (const uchar *source)
{
  return make_decoder_ref (&source);
}

void CLASS crw_init_tables (unsigned table, ushort *huff[2])
{
  static const uchar first_tree[3][29] = {
    { 0,1,4,2,3,1,2,0,0,0,0,0,0,0,0,0,
      0x04,0x03,0x05,0x06,0x02,0x07,0x01,0x08,0x09,0x00,0x0a,0x0b,0xff  },
    { 0,2,2,3,1,1,1,1,2,0,0,0,0,0,0,0,
      0x03,0x02,0x04,0x01,0x05,0x00,0x06,0x07,0x09,0x08,0x0a,0x0b,0xff  },
    { 0,0,6,3,1,1,2,0,0,0,0,0,0,0,0,0,

      0x06,0x05,0x07,0x04,0x08,0x03,0x09,0x02,0x00,0x0a,0x01,0x0b,0xff  },
  };
  static const uchar second_tree[3][180] = {
    { 0,2,2,2,1,4,2,1,2,5,1,1,0,0,0,139,
      0x03,0x04,0x02,0x05,0x01,0x06,0x07,0x08,
      0x12,0x13,0x11,0x14,0x09,0x15,0x22,0x00,0x21,0x16,0x0a,0xf0,
      0x23,0x17,0x24,0x31,0x32,0x18,0x19,0x33,0x25,0x41,0x34,0x42,
      0x35,0x51,0x36,0x37,0x38,0x29,0x79,0x26,0x1a,0x39,0x56,0x57,
      0x28,0x27,0x52,0x55,0x58,0x43,0x76,0x59,0x77,0x54,0x61,0xf9,
      0x71,0x78,0x75,0x96,0x97,0x49,0xb7,0x53,0xd7,0x74,0xb6,0x98,
      0x47,0x48,0x95,0x69,0x99,0x91,0xfa,0xb8,0x68,0xb5,0xb9,0xd6,
      0xf7,0xd8,0x67,0x46,0x45,0x94,0x89,0xf8,0x81,0xd5,0xf6,0xb4,
      0x88,0xb1,0x2a,0x44,0x72,0xd9,0x87,0x66,0xd4,0xf5,0x3a,0xa7,
      0x73,0xa9,0xa8,0x86,0x62,0xc7,0x65,0xc8,0xc9,0xa1,0xf4,0xd1,
      0xe9,0x5a,0x92,0x85,0xa6,0xe7,0x93,0xe8,0xc1,0xc6,0x7a,0x64,
      0xe1,0x4a,0x6a,0xe6,0xb3,0xf1,0xd3,0xa5,0x8a,0xb2,0x9a,0xba,
      0x84,0xa4,0x63,0xe5,0xc5,0xf3,0xd2,0xc4,0x82,0xaa,0xda,0xe4,
      0xf2,0xca,0x83,0xa3,0xa2,0xc3,0xea,0xc2,0xe2,0xe3,0xff,0xff  },
    { 0,2,2,1,4,1,4,1,3,3,1,0,0,0,0,140,
      0x02,0x03,0x01,0x04,0x05,0x12,0x11,0x06,
      0x13,0x07,0x08,0x14,0x22,0x09,0x21,0x00,0x23,0x15,0x31,0x32,
      0x0a,0x16,0xf0,0x24,0x33,0x41,0x42,0x19,0x17,0x25,0x18,0x51,
      0x34,0x43,0x52,0x29,0x35,0x61,0x39,0x71,0x62,0x36,0x53,0x26,
      0x38,0x1a,0x37,0x81,0x27,0x91,0x79,0x55,0x45,0x28,0x72,0x59,
      0xa1,0xb1,0x44,0x69,0x54,0x58,0xd1,0xfa,0x57,0xe1,0xf1,0xb9,
      0x49,0x47,0x63,0x6a,0xf9,0x56,0x46,0xa8,0x2a,0x4a,0x78,0x99,
      0x3a,0x75,0x74,0x86,0x65,0xc1,0x76,0xb6,0x96,0xd6,0x89,0x85,
      0xc9,0xf5,0x95,0xb4,0xc7,0xf7,0x8a,0x97,0xb8,0x73,0xb7,0xd8,
      0xd9,0x87,0xa7,0x7a,0x48,0x82,0x84,0xea,0xf4,0xa6,0xc5,0x5a,
      0x94,0xa4,0xc6,0x92,0xc3,0x68,0xb5,0xc8,0xe4,0xe5,0xe6,0xe9,
      0xa2,0xa3,0xe3,0xc2,0x66,0x67,0x93,0xaa,0xd4,0xd5,0xe7,0xf8,
      0x88,0x9a,0xd7,0x77,0xc4,0x64,0xe2,0x98,0xa5,0xca,0xda,0xe8,
      0xf3,0xf6,0xa9,0xb2,0xb3,0xf2,0xd2,0x83,0xba,0xd3,0xff,0xff  },
    { 0,0,6,2,1,3,3,2,5,1,2,2,8,10,0,117,
      0x04,0x05,0x03,0x06,0x02,0x07,0x01,0x08,
      0x09,0x12,0x13,0x14,0x11,0x15,0x0a,0x16,0x17,0xf0,0x00,0x22,
      0x21,0x18,0x23,0x19,0x24,0x32,0x31,0x25,0x33,0x38,0x37,0x34,
      0x35,0x36,0x39,0x79,0x57,0x58,0x59,0x28,0x56,0x78,0x27,0x41,
      0x29,0x77,0x26,0x42,0x76,0x99,0x1a,0x55,0x98,0x97,0xf9,0x48,
      0x54,0x96,0x89,0x47,0xb7,0x49,0xfa,0x75,0x68,0xb6,0x67,0x69,
      0xb9,0xb8,0xd8,0x52,0xd7,0x88,0xb5,0x74,0x51,0x46,0xd9,0xf8,
      0x3a,0xd6,0x87,0x45,0x7a,0x95,0xd5,0xf6,0x86,0xb4,0xa9,0x94,
      0x53,0x2a,0xa8,0x43,0xf5,0xf7,0xd4,0x66,0xa7,0x5a,0x44,0x8a,
      0xc9,0xe8,0xc8,0xe7,0x9a,0x6a,0x73,0x4a,0x61,0xc7,0xf4,0xc6,
      0x65,0xe9,0x72,0xe6,0x71,0x91,0x93,0xa6,0xda,0x92,0x85,0x62,
      0xf3,0xc5,0xb2,0xa4,0x84,0xba,0x64,0xa5,0xb3,0xd2,0x81,0xe5,
      0xd3,0xaa,0xc4,0xca,0xf2,0xb1,0xe4,0xd1,0x83,0x63,0xea,0xc3,
      0xe2,0x82,0xf1,0xa3,0xc2,0xa1,0xc1,0xe3,0xa2,0xe1,0xff,0xff  }
  };
  if (table > 2) table = 2;
  huff[0] = make_decoder ( first_tree[table]);
  huff[1] = make_decoder (second_tree[table]);
}

/*
   Return 0 if the image starts with compressed data,
   1 if it starts with uncompressed low-order bits.

   In Canon compressed data, 0xff is always followed by 0x00.
 */
int CLASS canon_has_lowbits()
{
  uchar test[0x4000];
  int ret=1;
  unsigned i;

  fseek (ifp, 0, SEEK_SET);
  fread (test, 1, sizeof test, ifp);
  for (i=540; i < sizeof test - 1; i++)
    if (test[i] == 0xff) {
      if (test[i+1]) return 1;
      ret=0;
    }
  return ret;
}

void CLASS canon_compressed_load_raw()
{
  ushort *pixel, *prow, *huff[2];
  int nblocks, lowbits, i, c, row, r, col, save, val;
  unsigned irow, icol;
  int block, diffbuf[64], leaf, len, diff, carry=0, pnum=0, base[2];

  crw_init_tables (tiff_compress, huff);
  pixel = (ushort *) calloc (raw_width*8, sizeof *pixel);
  merror (pixel, "canon_compressed_load_raw()");
  lowbits = canon_has_lowbits();
  if (!lowbits) maximum = 0x3ff;
  fseek (ifp, 540 + lowbits*raw_height*raw_width/4, SEEK_SET);
  zero_after_ff = 1;
  getbits(-1);
  for (row=0; row < raw_height; row+=8) {
    nblocks = MIN (8, raw_height-row) * raw_width >> 6;
    for (block=0; block < nblocks; block++) {
      memset (diffbuf, 0, sizeof diffbuf);
      for (i=0; i < 64; i++ ) {
	leaf = gethuff(huff[i > 0]);
	if (leaf == 0 && i) break;
	if (leaf == 0xff) continue;
	i  += leaf >> 4;
	len = leaf & 15;
	if (len == 0) continue;
	diff = getbits(len);
	if ((diff & (1 << (len-1))) == 0)
	  diff -= (1 << len) - 1;
	if (i < 64) diffbuf[i] = diff;
      }
      diffbuf[0] += carry;
      carry = diffbuf[0];
      for (i=0; i < 64; i++ ) {
	if (pnum++ % raw_width == 0)
	  base[0] = base[1] = 512;
	if ((pixel[(block << 6) + i] = base[i & 1] += diffbuf[i]) >> 10)
	  derror();
      }
    }
    if (lowbits) {
      save = ftell(ifp);
      fseek (ifp, 26 + row*raw_width/4, SEEK_SET);
      for (prow=pixel, i=0; i < raw_width*2; i++) {
	c = fgetc(ifp);
	for (r=0; r < 8; r+=2, prow++) {
	  val = (*prow << 2) + ((c >> r) & 3);
	  if (raw_width == 2672 && val < 512) val += 2;
	  *prow = val;
	}
      }
      fseek (ifp, save, SEEK_SET);
    }

#ifdef LIBRAW_LIBRARY_BUILD
    for (r=0; r < 8; r++) {
        if(row+r>=raw_height) break; // Not sure that raw_height is always N*8
        // MOVE entire row into place
        memmove(&raw_image[(row+r)*raw_width],&pixel[r*raw_width],raw_width*sizeof(pixel[0]));

        irow = row - top_margin + r;
        if (irow >= height) continue; // if row above image area than irow is VERY positive :)

        // only margins!
        for (col=0; col < left_margin; col++) 
            {
                icol = col - left_margin;
                c = FC(irow,icol);
                if (icol >= width && col > 1 && (unsigned) (col-left_margin+2) > width+3)
                    cblack[c] += (cblack[4+c]++,pixel[r*raw_width+col]);
            }
        for (col=width+left_margin; col < raw_width; col++) 
            {
                icol = col - left_margin;
                c = FC(irow,icol);
                if (icol >= width && col > 1 && (unsigned) (col-left_margin+2) > width+3)
                    cblack[c] += (cblack[4+c]++,pixel[r*raw_width+col]);
            }

    }
#else
    // dcraw original code
    for (r=0; r < 8; r++) {
      irow = row - top_margin + r;
      if (irow >= height) continue;
      for (col=0; col < raw_width; col++) {
	icol = col - left_margin;
	c = FC(irow,icol);
	if (icol < width)
	  BAYER(irow,icol) = pixel[r*raw_width+col];
	else if (col > 1 && (unsigned) (col-left_margin+2) > width+3)
	  cblack[c] += (cblack[4+c]++,pixel[r*raw_width+col]);
      }
    }
#endif
  }
  free (pixel);
  FORC(2) free (huff[c]);
  FORC4 if (cblack[4+c]) cblack[c] /= cblack[4+c];
}

int CLASS ljpeg_start (struct jhead *jh, int info_only)
{
  int c, tag, len;
  uchar data[0x10000];
  const uchar *dp;

  memset (jh, 0, sizeof *jh);
  jh->restart = INT_MAX;
  fread (data, 2, 1, ifp);
  if (data[1] != 0xd8) return 0;
  do {
    fread (data, 2, 2, ifp);
    tag =  data[0] << 8 | data[1];
    len = (data[2] << 8 | data[3]) - 2;
    if (tag <= 0xff00) return 0;
    fread (data, 1, len, ifp);
    switch (tag) {
      case 0xffc3:
	jh->sraw = ((data[7] >> 4) * (data[7] & 15) - 1) & 3;
      case 0xffc0:
	jh->bits = data[0];
	jh->high = data[1] << 8 | data[2];
	jh->wide = data[3] << 8 | data[4];
	jh->clrs = data[5] + jh->sraw;
	if (len == 9 && !dng_version) getc(ifp);
	break;
      case 0xffc4:
	if (info_only) break;
	for (dp = data; dp < data+len && (c = *dp++) < 4; )
	  jh->free[c] = jh->huff[c] = make_decoder_ref (&dp);
	break;
      case 0xffda:
	jh->psv = data[1+data[0]*2];
	jh->bits -= data[3+data[0]*2] & 15;
	break;
      case 0xffdd:
	jh->restart = data[0] << 8 | data[1];
    }
  } while (tag != 0xffda);
  if (info_only) return 1;
  FORC(5) if (!jh->huff[c+1]) jh->huff[c+1] = jh->huff[c];
  if (jh->sraw) {
    FORC(4)        jh->huff[2+c] = jh->huff[1];
    FORC(jh->sraw) jh->huff[1+c] = jh->huff[0];
  }
  jh->row = (ushort *) calloc (jh->wide*jh->clrs, 4);
  merror (jh->row, "ljpeg_start()");
  return zero_after_ff = 1;
}

void CLASS ljpeg_end (struct jhead *jh)
{
  int c;
  FORC4 if (jh->free[c]) free (jh->free[c]);
  free (jh->row);
}

// used for kodak-262 decoder
int CLASS ljpeg_diff (ushort *huff)
{
  int len, diff;

  len = gethuff(huff);
  if (len == 16 && (!dng_version || dng_version >= 0x1010000))
    return -32768;
  diff = getbits(len);
  if ((diff & (1 << (len-1))) == 0)
    diff -= (1 << len) - 1;
  return diff;
}

#ifdef LIBRAW_LIBRARY_BUILD
int CLASS ljpeg_diff_new (LibRaw_bit_buffer& bits, LibRaw_byte_buffer* buf,ushort *huff)
{
  int len, diff;

  len = bits._gethuff_lj(buf,*huff,huff+1);
  if (len == 16 && (!dng_version || dng_version >= 0x1010000))
    return -32768;
  diff = bits._getbits_lj(buf,len);
  if ((diff & (1 << (len-1))) == 0)
    diff -= (1 << len) - 1;
  return diff;
}

int CLASS ljpeg_diff_pef (LibRaw_bit_buffer& bits, LibRaw_byte_buffer* buf,ushort *huff)
{
  int len, diff;

  len = bits._gethuff(buf,*huff,huff+1,zero_after_ff);
  if (len == 16 && (!dng_version || dng_version >= 0x1010000))
    return -32768;
  diff = bits._getbits(buf,len,zero_after_ff);
  if ((diff & (1 << (len-1))) == 0)
    diff -= (1 << len) - 1;
  return diff;
}


#endif


ushort * CLASS ljpeg_row (int jrow, struct jhead *jh)
{
  int col, c, diff, pred, spred=0;
  ushort mark=0, *row[3];

  if (jrow * jh->wide % jh->restart == 0) {
    FORC(6) jh->vpred[c] = 1 << (jh->bits-1);
    if (jrow) {
      fseek (ifp, -2, SEEK_CUR);
      do mark = (mark << 8) + (c = fgetc(ifp));
      while (c != EOF && mark >> 4 != 0xffd);
    }
    getbits(-1);
  }
  FORC3 row[c] = jh->row + jh->wide*jh->clrs*((jrow+c) & 1);
  for (col=0; col < jh->wide; col++)
    FORC(jh->clrs) {
      diff = ljpeg_diff (jh->huff[c]);
      if (jh->sraw && c <= jh->sraw && (col | c))
		    pred = spred;
      else if (col) pred = row[0][-jh->clrs];
      else	    pred = (jh->vpred[c] += diff) - diff;
      if (jrow && col) switch (jh->psv) {
	case 1:	break;
	case 2: pred = row[1][0];					break;
	case 3: pred = row[1][-jh->clrs];				break;
	case 4: pred = pred +   row[1][0] - row[1][-jh->clrs];		break;
	case 5: pred = pred + ((row[1][0] - row[1][-jh->clrs]) >> 1);	break;
	case 6: pred = row[1][0] + ((pred - row[1][-jh->clrs]) >> 1);	break;
	case 7: pred = (pred + row[1][0]) >> 1;				break;
	default: pred = 0;
      }
      if ((**row = pred + diff) >> jh->bits) derror();
      if (c <= jh->sraw) spred = **row;
      row[0]++; row[1]++;
    }
  return row[2];
}

#ifdef LIBRAW_LIBRARY_BUILD
ushort * CLASS ljpeg_row_new (int jrow, struct jhead *jh, LibRaw_bit_buffer& bits,LibRaw_byte_buffer* bytes)
{
  int col, c, diff, pred, spred=0;
  ushort mark=0, *row[3];

  if (jrow * jh->wide % jh->restart == 0) {
    FORC(6) jh->vpred[c] = 1 << (jh->bits-1);
    if (jrow) {
        bytes->unseek2();
        do mark = (mark << 8) + (c = bytes->get_byte());
        while (c != EOF && mark >> 4 != 0xffd);
    }
    bits.reset();
  }
  FORC3 row[c] = jh->row + jh->wide*jh->clrs*((jrow+c) & 1);
  for (col=0; col < jh->wide; col++)
    FORC(jh->clrs) {
        diff = ljpeg_diff_new (bits,bytes,jh->huff[c]);
      if (jh->sraw && c <= jh->sraw && (col | c))
		    pred = spred;
      else if (col) pred = row[0][-jh->clrs];
      else	    pred = (jh->vpred[c] += diff) - diff;
      if (jrow && col) switch (jh->psv) {
	case 1:	break;
	case 2: pred = row[1][0];					break;
	case 3: pred = row[1][-jh->clrs];				break;
	case 4: pred = pred +   row[1][0] - row[1][-jh->clrs];		break;
	case 5: pred = pred + ((row[1][0] - row[1][-jh->clrs]) >> 1);	break;
	case 6: pred = row[1][0] + ((pred - row[1][-jh->clrs]) >> 1);	break;
	case 7: pred = (pred + row[1][0]) >> 1;				break;
	default: pred = 0;
      }
      if ((**row = pred + diff) >> jh->bits) derror();
      if (c <= jh->sraw) spred = **row;
      row[0]++; row[1]++;
    }
  return row[2];
}

#endif

void CLASS lossless_jpeg_load_raw()
{
  int jwide, jrow, jcol, val, c, i, row=0, col=0;
#ifndef LIBRAW_LIBRARY_BUILD
  int jidx,j;
#endif
  struct jhead jh;
  int min=INT_MAX;
  ushort *rp;
#ifdef LIBRAW_LIBRARY_BUILD
  int save_min = 0;
  unsigned slicesW[16],slicesWcnt=0,slices;
  unsigned *offset;
  unsigned t_y=0,t_x=0,t_s=0,slice=0,pixelsInSlice,pixno;
  if (!strcasecmp(make,"KODAK"))
      save_min = 1;
#endif

#ifdef LIBRAW_LIBRARY_BUILD
  if (cr2_slice[0]>15)
      throw LIBRAW_EXCEPTION_IO_EOF; // change many slices
#else
  fprintf(stderr,"Too many CR2 slices: %d\n",cr2_slice[0]+1);
  return;
#endif


  if (!ljpeg_start (&jh, 0)) return;
  jwide = jh.wide * jh.clrs;

#ifdef LIBRAW_LIBRARY_BUILD
  if(cr2_slice[0])
      {
          for(i=0;i<cr2_slice[0];i++)
              slicesW[slicesWcnt++] = cr2_slice[1];
          slicesW[slicesWcnt++] = cr2_slice[2];
      }
  else
      {
          // not sliced
          slicesW[slicesWcnt++] = raw_width; // safe fallback
      }
       
  slices = slicesWcnt * jh.high;
  offset = (unsigned*)calloc(slices+1,sizeof(offset[0]));
  
  for(slice=0;slice<slices;slice++)
      {
          offset[slice] = (t_x + t_y * raw_width)| (t_s<<28);
          if(offset[slice] & 0x0fffffff >= raw_width * raw_height)
              throw LIBRAW_EXCEPTION_IO_BADFILE; 
          t_y++;
          if(t_y == jh.high)
              {
                  t_y = 0;
                  t_x += slicesW[t_s++];
              }
      }
  offset[slices] = offset[slices-1];
  slice = 1; // next slice
  pixno = offset[0];
  pixelsInSlice = slicesW[0];
#endif

#ifdef LIBRAW_LIBRARY_BUILD
  LibRaw_byte_buffer *buf=NULL;
  if(data_size)
      buf = ifp->make_byte_buffer(data_size);
  LibRaw_bit_buffer bits;
#endif
  for (jrow=0; jrow < jh.high; jrow++) {
#ifdef LIBRAW_LIBRARY_BUILD
      if (data_size)
          rp = ljpeg_row_new (jrow, &jh,bits,buf);
      else
#endif
    rp = ljpeg_row (jrow, &jh);

    if (load_flags & 1)
      row = jrow & 1 ? height-1-jrow/2 : jrow/2;
    for (jcol=0; jcol < jwide; jcol++) {
      val = *rp++;
      if (jh.bits <= 12)
	val = curve[val & 0xfff];
#ifndef LIBRAW_LIBRARY_BUILD
      // slow dcraw way to calculate row/col
      if (cr2_slice[0]) {
	jidx = jrow*jwide + jcol;
	i = jidx / (cr2_slice[1]*jh.high);
	if ((j = i >= cr2_slice[0]))
		 i  = cr2_slice[0];
	jidx -= i * (cr2_slice[1]*jh.high);
	row = jidx / cr2_slice[1+j];
	col = jidx % cr2_slice[1+j] + i*cr2_slice[1];
      }
#else
      // new fast one, but for data_size defined only (i.e. new CR2 format, not 1D/1Ds)
      if(data_size) 
          {
              row = pixno/raw_width;
              col = pixno % raw_width;
              pixno++;
              if (0 == --pixelsInSlice)
                  {
                      unsigned o = offset[slice++];
                      pixno = o & 0x0fffffff;
                      pixelsInSlice = slicesW[o>>28];
                  }
          }
#endif
#ifndef LIBRAW_LIBRARY_BUILD

      if (raw_width == 3984 && (col -= 2) < 0)
              col += (row--,raw_width);

      if ((unsigned) (row-top_margin) < height) {
	c = FC(row-top_margin,col-left_margin);
	if ((unsigned) (col-left_margin) < width) {
	  BAYER(row-top_margin,col-left_margin) = val;
	  if (min > val) min = val;
	} else if (col > 1 && (unsigned) (col-left_margin+2) > width+3)
	  cblack[c] += (cblack[4+c]++,val);
      }
#else
      if (raw_width == 3984)
          {
              if ( (col -= 2) < 0)
                  col += (row--,raw_width);
              if(row >= 0 && row < raw_height && col >= 0 && col < raw_width)
                  RBAYER(row,col) = val;
          }
      else 
          RBAYER(row,col) = val;

      if ((unsigned) (row-top_margin) < height) 
          {
              // within image height
              if ((unsigned) (col-left_margin) < width) 
                  {
                      // within image area, save min
                      if(save_min)
                          if (min > val) min = val;
                  } 
              else if (col > 1 && (unsigned) (col-left_margin+2) > width+3) 
                  {
                      c = FC(row-top_margin,col-left_margin);
                      cblack[c] += (cblack[4+c]++,val);
                  }
          }
#endif

#ifndef LIBRAW_LIBRARY_BUILD
      if (++col >= raw_width)
	col = (row++,0);
#else
      if(!data_size) // 1D or 1Ds case
          if (++col >= raw_width)
              col = (row++,0);
#endif
    }
  }
  ljpeg_end (&jh);
  FORC4 if (cblack[4+c]) cblack[c] /= cblack[4+c];
  if (!strcasecmp(make,"KODAK"))
    black = min;
#ifdef LIBRAW_LIBRARY_BUILD
  if(buf)
      delete buf;
  free(offset);
#endif
}

void CLASS canon_sraw_load_raw()
{
  struct jhead jh;
  short *rp=0, (*ip)[4];
  int jwide, slice, scol, ecol, row, col, jrow=0, jcol=0, pix[3], c;
  int v[3]={0,0,0}, ver, hue;
  char *cp;

  if (!ljpeg_start (&jh, 0)) return;
  jwide = (jh.wide >>= 1) * jh.clrs;

#ifdef LIBRAW_LIBRARY_BUILD
  if(!data_size)
      throw LIBRAW_EXCEPTION_IO_BADFILE;
  LibRaw_byte_buffer *buf = ifp->make_byte_buffer(data_size);
  LibRaw_bit_buffer bits;
#endif


  for (ecol=slice=0; slice <= cr2_slice[0]; slice++) {
    scol = ecol;
    ecol += cr2_slice[1] * 2 / jh.clrs;
    if (!cr2_slice[0] || ecol > raw_width-1) ecol = raw_width & -2;
    for (row=0; row < height; row += (jh.clrs >> 1) - 1) {
      ip = (short (*)[4]) image + row*width;
      for (col=scol; col < ecol; col+=2, jcol+=jh.clrs) {
	if ((jcol %= jwide) == 0)
#ifdef LIBRAW_LIBRARY_BUILD
            rp = (short*) ljpeg_row_new (jrow++, &jh,bits,buf);
#else
            rp = (short *) ljpeg_row (jrow++, &jh);
#endif
	if (col >= width) continue;
	FORC (jh.clrs-2)
	  ip[col + (c >> 1)*width + (c & 1)][0] = rp[jcol+c];
	ip[col][1] = rp[jcol+jh.clrs-2] - 16384;
	ip[col][2] = rp[jcol+jh.clrs-1] - 16384;
      }
    }
  }
  for (cp=model2; *cp && !isdigit(*cp); cp++);
  sscanf (cp, "%d.%d.%d", v, v+1, v+2);
  ver = (v[0]*1000 + v[1])*1000 + v[2];
  hue = (jh.sraw+1) << 2;
  if (unique_id >= 0x80000281 || (unique_id == 0x80000218 && ver > 1000006))
    hue = jh.sraw << 1;
  ip = (short (*)[4]) image;
  rp = ip[0];
  for (row=0; row < height; row++, ip+=width) {
    if (row & (jh.sraw >> 1))
      for (col=0; col < width; col+=2)
	for (c=1; c < 3; c++)
	  if (row == height-1)
	       ip[col][c] =  ip[col-width][c];
	  else ip[col][c] = (ip[col-width][c] + ip[col+width][c] + 1) >> 1;
    for (col=1; col < width; col+=2)
      for (c=1; c < 3; c++)
	if (col == width-1)
	     ip[col][c] =  ip[col-1][c];
	else ip[col][c] = (ip[col-1][c] + ip[col+1][c] + 1) >> 1;
  }
  for ( ; rp < ip[0]; rp+=4) {
    if (unique_id < 0x80000218) {
      pix[0] = rp[0] + rp[2] - 512;
      pix[2] = rp[0] + rp[1] - 512;
      pix[1] = rp[0] + ((-778*rp[1] - (rp[2] << 11)) >> 12) - 512;
    } else {
      rp[1] = (rp[1] << 2) + hue;
      rp[2] = (rp[2] << 2) + hue;
      pix[0] = rp[0] + ((   50*rp[1] + 22929*rp[2]) >> 14);
      pix[1] = rp[0] + ((-5640*rp[1] - 11751*rp[2]) >> 14);
      pix[2] = rp[0] + ((29040*rp[1] -   101*rp[2]) >> 14);
    }
    FORC3 
        rp[c] = CLIP(pix[c] * sraw_mul[c] >> 10);
  }
#ifdef LIBRAW_LIBRARY_BUILD
  delete buf;
#endif
  ljpeg_end (&jh);
  maximum = 0x3fff;
}

void CLASS adobe_copy_pixel (int row, int col, ushort **rp)
{
  unsigned r, c;

#ifndef LIBRAW_LIBRARY_BUILD
  r = row -= top_margin;
  c = col -= left_margin;
  if (is_raw == 2 && shot_select) (*rp)++;
  if (filters) {
    if (fuji_width) {
      r = row + fuji_width - 1 - (col >> 1);
      c = row + ((col+1) >> 1);
    }
    if (r < height && c < width)
      BAYER(r,c) = **rp < 0x1000 ? curve[**rp] : **rp;
    *rp += is_raw;
  } else {
    if (r < height && c < width)
      FORC(tiff_samples)
	image[row*width+col][c] = (*rp)[c] < 0x1000 ? curve[(*rp)[c]]:(*rp)[c];
    *rp += tiff_samples;
  }
#else
  if (is_raw == 2 && shot_select) (*rp)++;
  if (filters) {
      if(row < raw_height && col < raw_width)
          RBAYER(row,col) = **rp < 0x1000 ? curve[**rp] : **rp;
    *rp += is_raw;
  } else {
      if (row < raw_height && col < raw_width)
          FORC(tiff_samples)
              color_image[row*raw_width+col][c] = (*rp)[c]<0x1000 ? curve[(*rp)[c]]:(*rp)[c];
      *rp += tiff_samples;
  }

#endif
  if (is_raw == 2 && shot_select) (*rp)--;

}

void CLASS adobe_dng_load_raw_lj()
{
  unsigned save, trow=0, tcol=0, jwide, jrow, jcol, row, col;
  struct jhead jh;
  ushort *rp;
  while (trow < raw_height) {
    save = ftell(ifp);
    if (tile_length < INT_MAX)
      fseek (ifp, get4(), SEEK_SET);
    if (!ljpeg_start (&jh, 0)) break;
    jwide = jh.wide;
    if (filters) jwide *= jh.clrs;
    jwide /= is_raw;
#ifdef LIBRAW_LIBRARY_BUILD
    if(!data_size)
        throw LIBRAW_EXCEPTION_IO_BADFILE;
    LibRaw_byte_buffer *buf = ifp->make_byte_buffer(data_size);
    LibRaw_bit_buffer bits;
#endif
    for (row=col=jrow=0; jrow < jh.high; jrow++) {
#ifdef LIBRAW_LIBRARY_BUILD
        rp = ljpeg_row_new (jrow, &jh,bits,buf);
#else
        rp = ljpeg_row (jrow, &jh);
#endif
      for (jcol=0; jcol < jwide; jcol++) {
	adobe_copy_pixel (trow+row, tcol+col, &rp);
	if (++col >= tile_width || col >= raw_width)
	  row += 1 + (col = 0);
      }
    }
    fseek (ifp, save+4, SEEK_SET);
    if ((tcol += tile_width) >= raw_width)
      trow += tile_length + (tcol = 0);
    ljpeg_end (&jh);
#ifdef LIBRAW_LIBRARY_BUILD
    delete buf;
#endif
  }
}

void CLASS adobe_dng_load_raw_nc()
{
  ushort *pixel, *rp;
  int row, col;

  pixel = (ushort *) calloc (raw_width * tiff_samples, sizeof *pixel);
  merror (pixel, "adobe_dng_load_raw_nc()");


#ifdef LIBRAW_LIBRARY_BUILD
  int dsz= raw_height*raw_width * tiff_samples * tiff_bps/8;
  LibRaw_byte_buffer *buf = NULL;
  if (tiff_bps != 16)
      {
          buf = ifp->make_byte_buffer(dsz);
      }
  LibRaw_bit_buffer bits;
#endif

  for (row=0; row < raw_height; row++) {
    if (tiff_bps == 16)
      read_shorts (pixel, raw_width * tiff_samples);
    else {
#ifdef LIBRAW_LIBRARY_BUILD
        bits.reset();
        for (col=0; col < raw_width * tiff_samples; col++)
            pixel[col] = bits._getbits(buf,tiff_bps,zero_after_ff);

#else
      getbits(-1);
      for (col=0; col < raw_width * tiff_samples; col++)
	pixel[col] = getbits(tiff_bps);
#endif
    }
    for (rp=pixel, col=0; col < raw_width; col++)
      adobe_copy_pixel (row, col, &rp);
  }
  free (pixel);
#ifdef LIBRAW_LIBRARY_BUILD
    if(buf)
        delete buf;
#endif
}

void CLASS pentax_load_raw()
{
  ushort bit[2][15], huff[4097];
  int dep, row, col, diff, c, i;
  ushort vpred[2][2] = {{0,0},{0,0}}, hpred[2];

  fseek (ifp, meta_offset, SEEK_SET);
  dep = (get2() + 12) & 15;
  fseek (ifp, 12, SEEK_CUR);
  FORC(dep) bit[0][c] = get2();
  FORC(dep) bit[1][c] = fgetc(ifp);
  FORC(dep)
    for (i=bit[0][c]; i <= ((bit[0][c]+(4096 >> bit[1][c])-1) & 4095); )
      huff[++i] = bit[1][c] << 8 | c;
  huff[0] = 12;
  fseek (ifp, data_offset, SEEK_SET);
#ifdef LIBRAW_LIBRARY_BUILD
  if(!data_size)
      throw LIBRAW_EXCEPTION_IO_BADFILE;
  LibRaw_byte_buffer *buf = ifp->make_byte_buffer(data_size);
  LibRaw_bit_buffer bits;
  bits.reset();
#else
  getbits(-1);
#endif
  for (row=0; row < raw_height; row++)
      {
          for (col=0; col < raw_width; col++) {
#ifdef LIBRAW_LIBRARY_BUILD
              diff = ljpeg_diff_pef(bits,buf,huff);
#else
              diff = ljpeg_diff (huff);
#endif
              if (col < 2) hpred[col] = vpred[row & 1][col] += diff;
              else	   hpred[col & 1] += diff;

              unsigned val = hpred[col & 1];

#ifndef LIBRAW_LIBRARY_BUILD
              if ((unsigned) (row-top_margin) < height && 
                  (unsigned)(col-left_margin) < width)
                  BAYER(row-top_margin,col-left_margin) = val;
#else
              RBAYER(row,col) = val;
#endif
              if (val >> tiff_bps) derror();
          }
      }
#ifdef LIBRAW_LIBRARY_BUILD
  delete buf;
#endif
}

void CLASS nikon_compressed_load_raw()
{
  static const uchar nikon_tree[][32] = {
    { 0,1,5,1,1,1,1,1,1,2,0,0,0,0,0,0,	/* 12-bit lossy */
      5,4,3,6,2,7,1,0,8,9,11,10,12 },
    { 0,1,5,1,1,1,1,1,1,2,0,0,0,0,0,0,	/* 12-bit lossy after split */
      0x39,0x5a,0x38,0x27,0x16,5,4,3,2,1,0,11,12,12 },
    { 0,1,4,2,3,1,2,0,0,0,0,0,0,0,0,0,  /* 12-bit lossless */
      5,4,6,3,7,2,8,1,9,0,10,11,12 },
    { 0,1,4,3,1,1,1,1,1,2,0,0,0,0,0,0,	/* 14-bit lossy */
      5,6,4,7,8,3,9,2,1,0,10,11,12,13,14 },
    { 0,1,5,1,1,1,1,1,1,1,2,0,0,0,0,0,	/* 14-bit lossy after split */
      8,0x5c,0x4b,0x3a,0x29,7,6,5,4,3,2,1,0,13,14 },
    { 0,1,4,2,2,3,1,2,0,0,0,0,0,0,0,0,	/* 14-bit lossless */
      7,6,8,5,9,4,10,3,11,12,2,0,1,13,14 } };
  ushort *huff, ver0, ver1, vpred[2][2], hpred[2], csize;
  int i, min, max, step=0, tree=0, split=0, row, col, len, shl, diff;

  fseek (ifp, meta_offset, SEEK_SET);
  ver0 = fgetc(ifp);
  ver1 = fgetc(ifp);
  if (ver0 == 0x49 || ver1 == 0x58)
    fseek (ifp, 2110, SEEK_CUR);
  if (ver0 == 0x46) tree = 2;
  if (tiff_bps == 14) tree += 3;
  read_shorts (vpred[0], 4);
  max = 1 << tiff_bps & 0x7fff;
  if ((csize = get2()) > 1)
    step = max / (csize-1);
  if (ver0 == 0x44 && ver1 == 0x20 && step > 0) {
    for (i=0; i < csize; i++)
      curve[i*step] = get2();
    for (i=0; i < max; i++)
      curve[i] = ( curve[i-i%step]*(step-i%step) +
		   curve[i-i%step+step]*(i%step) ) / step;
#ifdef LIBRAW_LIBRARY_BUILD
    color_flags.curve_state = LIBRAW_COLORSTATE_LOADED;
#endif
    fseek (ifp, meta_offset+562, SEEK_SET);
    split = get2();
  } else if (ver0 != 0x46 && csize <= 0x4001)
      {
    read_shorts (curve, max=csize);
#ifdef LIBRAW_LIBRARY_BUILD
    color_flags.curve_state = LIBRAW_COLORSTATE_LOADED;
#endif
      }
  while (curve[max-2] == curve[max-1]) max--;
  huff = make_decoder (nikon_tree[tree]);
  fseek (ifp, data_offset, SEEK_SET);
#ifdef LIBRAW_LIBRARY_BUILD
  if(!data_size)
      throw LIBRAW_EXCEPTION_IO_BADFILE;
  LibRaw_byte_buffer *buf = ifp->make_byte_buffer(data_size);
  LibRaw_bit_buffer bits;
  bits.reset();
#else
  getbits(-1);
#endif
  for (min=row=0; row < height; row++) {
    if (split && row == split) {
      free (huff);
      huff = make_decoder (nikon_tree[tree+1]);
      max += (min = 16) << 1;
    }
    for (col=0; col < raw_width; col++) {
#ifdef LIBRAW_LIBRARY_BUILD
        i = bits._gethuff(buf,*huff,huff+1,zero_after_ff);
#else
      i = gethuff(huff);
#endif
      len = i & 15;
      shl = i >> 4;
#ifdef LIBRAW_LIBRARY_BUILD
      diff = ((bits._getbits(buf,len-shl,zero_after_ff) << 1) + 1) << shl >> 1;
#else
      diff = ((getbits(len-shl) << 1) + 1) << shl >> 1;
#endif
      if ((diff & (1 << (len-1))) == 0)
	diff -= (1 << len) - !shl;
      if (col < 2) hpred[col] = vpred[row & 1][col] += diff;
      else	   hpred[col & 1] += diff;
      if ((ushort)(hpred[col & 1] + min) >= max) derror();
#ifndef LIBRAW_LIBRARY_BUILD
      if ((unsigned) (col-left_margin) < width)
	BAYER(row,col-left_margin) =  curve[LIM((short)hpred[col & 1],0,0x3fff)];
#else
      ushort xval = hpred[col & 1];
      xval = curve[LIM((short)xval,0,0x3fff)];
      RBAYER(row,col) = xval;
#endif

    }
  }
#ifdef LIBRAW_LIBRARY_BUILD
  delete buf;
#endif
  free (huff);
}

/*
   Figure out if a NEF file is compressed.  These fancy heuristics
   are only needed for the D100, thanks to a bug in some cameras
   that tags all images as "compressed".
 */
int CLASS nikon_is_compressed()
{
  uchar test[256];
  int i;

  fseek (ifp, data_offset, SEEK_SET);
  fread (test, 1, 256, ifp);
  for (i=15; i < 256; i+=16)
    if (test[i]) return 1;
  return 0;
}

/*
   Returns 1 for a Coolpix 995, 0 for anything else.
 */
int CLASS nikon_e995()
{
  int i, histo[256];
  const uchar often[] = { 0x00, 0x55, 0xaa, 0xff };

  memset (histo, 0, sizeof histo);
  fseek (ifp, -2000, SEEK_END);
  for (i=0; i < 2000; i++)
    histo[fgetc(ifp)]++;
  for (i=0; i < 4; i++)
    if (histo[often[i]] < 200)
      return 0;
  return 1;
}

/*
   Returns 1 for a Coolpix 2100, 0 for anything else.
 */
int CLASS nikon_e2100()
{
  uchar t[12];
  int i;

  fseek (ifp, 0, SEEK_SET);
  for (i=0; i < 1024; i++) {
    fread (t, 1, 12, ifp);
    if (((t[2] & t[4] & t[7] & t[9]) >> 4
	& t[1] & t[6] & t[8] & t[11] & 3) != 3)
      return 0;
  }
  return 1;
}

void CLASS nikon_3700()
{
  int bits, i;
  uchar dp[24];
  static const struct {
    int bits;
    char t_make[12], t_model[15];
  } table[] = {
    { 0x00, "PENTAX",  "Optio 33WR" },
    { 0x03, "NIKON",   "E3200" },
    { 0x32, "NIKON",   "E3700" },
    { 0x33, "OLYMPUS", "C740UZ" } };

  fseek (ifp, 3072, SEEK_SET);
  fread (dp, 1, 24, ifp);
  bits = (dp[8] & 3) << 4 | (dp[20] & 3);
  for (i=0; i < sizeof table / sizeof *table; i++)
    if (bits == table[i].bits) {
      strcpy (make,  table[i].t_make );
      strcpy (model, table[i].t_model);
    }
}

/*
   Separates a Minolta DiMAGE Z2 from a Nikon E4300.
 */
int CLASS minolta_z2()
{
  int i, nz;
  char tail[424];

  fseek (ifp, -sizeof tail, SEEK_END);
  fread (tail, 1, sizeof tail, ifp);
  for (nz=i=0; i < sizeof tail; i++)
    if (tail[i]) nz++;
  return nz > 20;
}

/*
   The Fuji Super CCD is just a Bayer grid rotated 45 degrees.
 */
void CLASS fuji_load_raw()
{
#ifndef LIBRAW_LIBRARY_BUILD
  ushort *pixel;
  int wide, row, col, r, c;

  fseek (ifp, (top_margin*raw_width + left_margin) * 2, SEEK_CUR);
  wide = fuji_width << !fuji_layout;
  pixel = (ushort *) calloc (wide, sizeof *pixel);
  merror (pixel, "fuji_load_raw()");
  for (row=0; row < raw_height; row++) {
    read_shorts (pixel, wide);
    fseek (ifp, 2*(raw_width - wide), SEEK_CUR);
    for (col=0; col < wide; col++) {
      if (fuji_layout) {
	r = fuji_width - 1 - col + (row >> 1);
	c = col + ((row+1) >> 1);
      } else {
	r = fuji_width - 1 + row - (col >> 1);
	c = row + ((col+1) >> 1);
      }
      BAYER(r,c) = pixel[col];
    }
  }
  free (pixel);
#else
  read_shorts(raw_image,raw_width*raw_height);
#endif
}
void CLASS ppm_thumb()
{
  char *thumb;
  thumb_length = thumb_width*thumb_height*3;
  thumb = (char *) malloc (thumb_length);
  merror (thumb, "ppm_thumb()");
  fprintf (ofp, "P6\n%d %d\n255\n", thumb_width, thumb_height);
  fread  (thumb, 1, thumb_length, ifp);
  fwrite (thumb, 1, thumb_length, ofp);
  free (thumb);
}

void CLASS layer_thumb()
{
  int i, c;
  char *thumb, map[][4] = { "012","102" };

  colors = thumb_misc >> 5 & 7;
  thumb_length = thumb_width*thumb_height;
  thumb = (char *) calloc (colors, thumb_length);
  merror (thumb, "layer_thumb()");
  fprintf (ofp, "P%d\n%d %d\n255\n",
	5 + (colors >> 1), thumb_width, thumb_height);
  fread (thumb, thumb_length, colors, ifp);
  for (i=0; i < thumb_length; i++)
    FORCC putc (thumb[i+thumb_length*(map[thumb_misc >> 8][c]-'0')], ofp);
  free (thumb);
}

void CLASS rollei_thumb()
{
  unsigned i;
  ushort *thumb;

  thumb_length = thumb_width * thumb_height;
  thumb = (ushort *) calloc (thumb_length, 2);
  merror (thumb, "rollei_thumb()");
  fprintf (ofp, "P6\n%d %d\n255\n", thumb_width, thumb_height);
  read_shorts (thumb, thumb_length);
  for (i=0; i < thumb_length; i++) {
    putc (thumb[i] << 3, ofp);
    putc (thumb[i] >> 5  << 2, ofp);
    putc (thumb[i] >> 11 << 3, ofp);
  }
  free (thumb);
}

void CLASS rollei_load_raw()
{
  uchar pixel[10];
  unsigned iten=0, isix, i, buffer=0, row, col, todo[16];

  isix = raw_width * raw_height * 5 / 8;
  while (fread (pixel, 1, 10, ifp) == 10) {
    for (i=0; i < 10; i+=2) {
      todo[i]   = iten++;
      todo[i+1] = pixel[i] << 8 | pixel[i+1];
      buffer    = pixel[i] >> 2 | buffer << 6;
    }
    for (   ; i < 16; i+=2) {
      todo[i]   = isix++;
      todo[i+1] = buffer >> (14-i)*5;
    }
    for (i=0; i < 16; i+=2) {
#ifndef LIBRAW_LIBRARY_BUILD
      row = todo[i] / raw_width - top_margin;
      col = todo[i] % raw_width - left_margin;
      if (row < height && col < width)
              BAYER(row,col) = (todo[i+1] & 0x3ff);
#else
      RBAYER(todo[i] / raw_width,todo[i] % raw_width) = (todo[i+1] & 0x3ff);
#endif
    }
  }
  maximum = 0x3ff;
}

int CLASS bayer (unsigned row, unsigned col)
{
  return (row < height && col < width) ? BAYER(row,col) : 0;
}

void CLASS phase_one_flat_field (int is_float, int nc)
{
  ushort head[8];
  unsigned wide, y, x, c, rend, cend, row, col;
  float *mrow, num, mult[4];

  read_shorts (head, 8);
  wide = head[2] / head[4];
  mrow = (float *) calloc (nc*wide, sizeof *mrow);
  merror (mrow, "phase_one_flat_field()");
  for (y=0; y < head[3] / head[5]; y++) {
    for (x=0; x < wide; x++)
      for (c=0; c < nc; c+=2) {
	num = is_float ? getreal(11) : get2()/32768.0;
	if (y==0) mrow[c*wide+x] = num;
	else mrow[(c+1)*wide+x] = (num - mrow[c*wide+x]) / head[5];
      }
    if (y==0) continue;
    rend = head[1]-top_margin + y*head[5];
    for (row = rend-head[5]; row < height && row < rend; row++) {
      for (x=1; x < wide; x++) {
	for (c=0; c < nc; c+=2) {
	  mult[c] = mrow[c*wide+x-1];
	  mult[c+1] = (mrow[c*wide+x] - mult[c]) / head[4];
	}
	cend = head[0]-left_margin + x*head[4];
	for (col = cend-head[4]; col < width && col < cend; col++) {
	  c = nc > 2 ? FC(row,col) : 0;
	  if (!(c & 1)) {
	    c = BAYER(row,col) * mult[c];
	    BAYER(row,col) = LIM(c,0,65535);
	  }
	  for (c=0; c < nc; c+=2)
	    mult[c] += mult[c+1];
	}
      }
      for (x=0; x < wide; x++)
	for (c=0; c < nc; c+=2)
	  mrow[c*wide+x] += mrow[(c+1)*wide+x];
    }
  }
  free (mrow);
}

void CLASS phase_one_correct()
{
  unsigned entries, tag, data, save, col, row, type;
  int len, i, j, k, cip, val[4], dev[4], sum, max;
  int head[9], diff, mindiff=INT_MAX, off_412=0;
  static const signed char dir[12][2] =
    { {-1,-1}, {-1,1}, {1,-1}, {1,1}, {-2,0}, {0,-2}, {0,2}, {2,0},
      {-2,-2}, {-2,2}, {2,-2}, {2,2} };
  float poly[8], num, cfrac, frac, mult[2], *yval[2];
  ushort *xval[2];

  if (half_size || !meta_length) return;
#ifdef DCRAW_VERBOSE
  if (verbose) fprintf (stderr,_("Phase One correction...\n"));
#endif
  fseek (ifp, meta_offset, SEEK_SET);
  order = get2();
  fseek (ifp, 6, SEEK_CUR);
  fseek (ifp, meta_offset+get4(), SEEK_SET);
  entries = get4();  get4();
  while (entries--) {
    tag  = get4();
    len  = get4();
    data = get4();
    save = ftell(ifp);
    fseek (ifp, meta_offset+data, SEEK_SET);
    if (tag == 0x419) {				/* Polynomial curve */
      for (get4(), i=0; i < 8; i++)
	poly[i] = getreal(11);
      poly[3] += (ph1.tag_210 - poly[7]) * poly[6] + 1;
      for (i=0; i < 0x10000; i++) {
	num = (poly[5]*i + poly[3])*i + poly[1];
	curve[i] = LIM(num,0,65535);
      } goto apply;				/* apply to right half */
    } else if (tag == 0x41a) {			/* Polynomial curve */
      for (i=0; i < 4; i++)
	poly[i] = getreal(11);
      for (i=0; i < 0x10000; i++) {
	for (num=0, j=4; j--; )
	  num = num * i + poly[j];
	curve[i] = LIM(num+i,0,65535);
      } apply:					/* apply to whole image */
      for (row=0; row < height; row++)
	for (col = (tag & 1)*ph1.split_col; col < width; col++)
	  BAYER(row,col) = curve[BAYER(row,col)];
    } else if (tag == 0x400) {			/* Sensor defects */
      while ((len -= 8) >= 0) {
	col  = get2() - left_margin;
	row  = get2() - top_margin;
	type = get2(); get2();
	if (col >= width) continue;
	if (type == 131)			/* Bad column */
	  for (row=0; row < height; row++)
	    if (FC(row,col) == 1) {
	      for (sum=i=0; i < 4; i++)
		sum += val[i] = bayer (row+dir[i][0], col+dir[i][1]);
	      for (max=i=0; i < 4; i++) {
		dev[i] = abs((val[i] << 2) - sum);
		if (dev[max] < dev[i]) max = i;
	      }
	      BAYER(row,col) = (sum - val[max])/3.0 + 0.5;
	    } else {
	      for (sum=0, i=8; i < 12; i++)
		sum += bayer (row+dir[i][0], col+dir[i][1]);
	      BAYER(row,col) = 0.5 + sum * 0.0732233 +
		(bayer(row,col-2) + bayer(row,col+2)) * 0.3535534;
	    }
	else if (type == 129) {			/* Bad pixel */
	  if (row >= height) continue;
	  j = (FC(row,col) != 1) * 4;
	  for (sum=0, i=j; i < j+8; i++)
	    sum += bayer (row+dir[i][0], col+dir[i][1]);
	  BAYER(row,col) = (sum + 4) >> 3;
	}
      }
    } else if (tag == 0x401) {			/* All-color flat fields */
      phase_one_flat_field (1, 2);
    } else if (tag == 0x416 || tag == 0x410) {
      phase_one_flat_field (0, 2);
    } else if (tag == 0x40b) {			/* Red+blue flat field */
      phase_one_flat_field (0, 4);
    } else if (tag == 0x412) {
      fseek (ifp, 36, SEEK_CUR);
      diff = abs (get2() - ph1.tag_21a);
      if (mindiff > diff) {
	mindiff = diff;
	off_412 = ftell(ifp) - 38;
      }
    }
    fseek (ifp, save, SEEK_SET);
  }
  if (off_412) {
    fseek (ifp, off_412, SEEK_SET);
    for (i=0; i < 9; i++) head[i] = get4() & 0x7fff;
    yval[0] = (float *) calloc (head[1]*head[3] + head[2]*head[4], 6);
    merror (yval[0], "phase_one_correct()");
    yval[1] = (float  *) (yval[0] + head[1]*head[3]);
    xval[0] = (ushort *) (yval[1] + head[2]*head[4]);
    xval[1] = (ushort *) (xval[0] + head[1]*head[3]);
    get2();
    for (i=0; i < 2; i++)
      for (j=0; j < head[i+1]*head[i+3]; j++)
	yval[i][j] = getreal(11);
    for (i=0; i < 2; i++)
      for (j=0; j < head[i+1]*head[i+3]; j++)
	xval[i][j] = get2();
    for (row=0; row < height; row++)
      for (col=0; col < width; col++) {
	cfrac = (float) col * head[3] / raw_width;
	cfrac -= cip = cfrac;
	num = BAYER(row,col) * 0.5;
	for (i=cip; i < cip+2; i++) {
	  for (k=j=0; j < head[1]; j++)
	    if (num < xval[0][k = head[1]*i+j]) break;
	  frac = (j == 0 || j == head[1]) ? 0 :
		(xval[0][k] - num) / (xval[0][k] - xval[0][k-1]);
	  mult[i-cip] = yval[0][k-1] * frac + yval…