do {
	  if (bytes_in_buffer == 0) {
	    if (! (*cinfo->src->fill_input_buffer) (cinfo))
	      return FALSE;
	    next_input_byte = cinfo->src->next_input_byte;
	    bytes_in_buffer = cinfo->src->bytes_in_buffer;
	  }
	  bytes_in_buffer--;
	  c = GETJOCTET(*next_input_byte++);
	} while (c == 0xFF);

	if (c == 0) {
	  /* Found FF/00, which represents an FF data byte */
	  c = 0xFF;
	} else {
	  /* Oops, it's actually a marker indicating end of compressed data.
	   * Save the marker code for later use.
	   * Fine point: it might appear that we should save the marker into
	   * bitread working state, not straight into permanent state.  But
	   * once we have hit a marker, we cannot need to suspend within the
	   * current MCU, because we will read no more bytes from the data
	   * source.  So it is OK to update permanent state right away.
	   */
	  cinfo->unread_marker = c;
	  /* See if we need to insert some fake zero bits. */
	  goto no_more_bytes;
	}
      }

      /* OK, load c into get_buffer */
      get_buffer = (get_buffer << 8) | c;
      bits_left += 8;
    } /* end while */
  } else {
  no_more_bytes:
    /* We get here if we've read the marker that terminates the compressed
     * data segment.  There should be enough bits in the buffer register
     * to satisfy the request; if so, no problem.
     */
    if (nbits > bits_left) {
      /* Uh-oh.  Report corrupted data to user and stuff zeroes into
       * the data stream, so that we can produce some kind of image.
       * We use a nonvolatile flag to ensure that only one warning message
       * appears per data segment.
       */
      if (! cinfo->entropy->insufficient_data) {
	WARNMS(cinfo, JWRN_HIT_MARKER);
	cinfo->entropy->insufficient_data = TRUE;
      }
      /* Fill the buffer with zero bits */
      get_buffer <<= MIN_GET_BITS - bits_left;
      bits_left = MIN_GET_BITS;
    }
  }

  /* Unload the local registers */
  state->next_input_byte = next_input_byte;
  state->bytes_in_buffer = bytes_in_buffer;
  state->get_buffer = get_buffer;
  state->bits_left = bits_left;

  return TRUE;
}


/*
 * Out-of-line code for Huffman code decoding.
 * See jdhuff.h for info about usage.
 */

GLOBAL(int)
jpeg_huff_decode (bitread_working_state * state,
		  register bit_buf_type get_buffer, register int bits_left,
		  d_derived_tbl * htbl, int min_bits)
{
  register int l = min_bits;
  register INT32 code;

  /* HUFF_DECODE has determined that the code is at least min_bits */
  /* bits long, so fetch that many bits in one swoop. */

  CHECK_BIT_BUFFER(*state, l, return -1);
  code = GET_BITS(l);

  /* Collect the rest of the Huffman code one bit at a time. */
  /* This is per Figure F.16 in the JPEG spec. */

  while (code > htbl->maxcode[l]) {
    code <<= 1;
    CHECK_BIT_BUFFER(*state, 1, return -1);
    code |= GET_BITS(1);
    l++;
  }

  /* Unload the local registers */
  state->get_buffer = get_buffer;
  state->bits_left = bits_left;

  /* With garbage input we may reach the sentinel value l = 17. */

  if (l > 16) {
    WARNMS(state->cinfo, JWRN_HUFF_BAD_CODE);
    return 0;			/* fake a zero as the safest result */
  }

  return htbl->pub->huffval[ (int) (code + htbl->valoffset[l]) ];
}


/*
 * Figure F.12: extend sign bit.
 * On some machines, a shift and add will be faster than a table lookup.
 */

#ifdef AVOID_TABLES

#define HUFF_EXTEND(x,s)  ((x) < (1<<((s)-1)) ? (x) + (((-1)<<(s)) + 1) : (x))

#else

#define HUFF_EXTEND(x,s)  ((x) < extend_test[s] ? (x) + extend_offset[s] : (x))

static const int extend_test[16] =   /* entry n is 2**(n-1) */
  { 0, 0x0001, 0x0002, 0x0004, 0x0008, 0x0010, 0x0020, 0x0040, 0x0080,
    0x0100, 0x0200, 0x0400, 0x0800, 0x1000, 0x2000, 0x4000 };

static const int extend_offset[16] = /* entry n is (-1 << n) + 1 */
  { 0, ((-1)<<1) + 1, ((-1)<<2) + 1, ((-1)<<3) + 1, ((-1)<<4) + 1,
    ((-1)<<5) + 1, ((-1)<<6) + 1, ((-1)<<7) + 1, ((-1)<<8) + 1,
    ((-1)<<9) + 1, ((-1)<<10) + 1, ((-1)<<11) + 1, ((-1)<<12) + 1,
    ((-1)<<13) + 1, ((-1)<<14) + 1, ((-1)<<15) + 1 };

#endif /* AVOID_TABLES */


/*
 * Check for a restart marker & resynchronize decoder.
 * Returns FALSE if must suspend.
 */

LOCAL(boolean)
process_restart (j_decompress_ptr cinfo)
{
  huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy;
  int ci;

  /* Throw away any unused bits remaining in bit buffer; */
  /* include any full bytes in next_marker's count of discarded bytes */
  cinfo->marker->discarded_bytes += entropy->bitstate.bits_left / 8;
  entropy->bitstate.bits_left = 0;

  /* Advance past the RSTn marker */
  if (! (*cinfo->marker->read_restart_marker) (cinfo))
    return FALSE;

  /* Re-initialize DC predictions to 0 */
  for (ci = 0; ci < cinfo->comps_in_scan; ci++)
    entropy->saved.last_dc_val[ci] = 0;

  /* Reset restart counter */
  entropy->restarts_to_go = cinfo->restart_interval;

  /* Reset out-of-data flag, unless read_restart_marker left us smack up
   * against a marker.  In that case we will end up treating the next data
   * segment as empty, and we can avoid producing bogus output pixels by
   * leaving the flag set.
   */
  if (cinfo->unread_marker == 0)
    entropy->pub.insufficient_data = FALSE;

  return TRUE;
}


/*
 * Decode and return one MCU's worth of Huffman-compressed coefficients.
 * The coefficients are reordered from zigzag order into natural array order,
 * but are not dequantized.
 *
 * The i'th block of the MCU is stored into the block pointed to by
 * MCU_data[i].  WE ASSUME THIS AREA HAS BEEN ZEROED BY THE CALLER.
 * (Wholesale zeroing is usually a little faster than retail...)
 *
 * Returns FALSE if data source requested suspension.  In that case no
 * changes have been made to permanent state.  (Exception: some output
 * coefficients may already have been assigned.  This is harmless for
 * this module, since we'll just re-assign them on the next call.)
 */

METHODDEF(boolean)
decode_mcu (j_decompress_ptr cinfo, JBLOCKROW *MCU_data)
{
  huff_entropy_ptr entropy = (huff_entropy_ptr) cinfo->entropy;
  int blkn;
  BITREAD_STATE_VARS;
  savable_state state;

  /* Process restart marker if needed; may have to suspend */
  if (cinfo->restart_interval) {
    if (entropy->restarts_to_go == 0)
      if (! process_restart(cinfo))
	return FALSE;
  }

  /* If we've run out of data, just leave the MCU set to zeroes.
   * This way, we return uniform gray for the remainder of the segment.
   */
  if (! entropy->pub.insufficient_data) {

    /* Load up working state */
    BITREAD_LOAD_STATE(cinfo,entropy->bitstate);
    ASSIGN_STATE(state, entropy->saved);

    /* Outer loop handles each block in the MCU */

    for (blkn = 0; blkn < cinfo->blocks_in_MCU; blkn++) {
      JBLOCKROW block = MCU_data[blkn];
      d_derived_tbl * dctbl = entropy->dc_cur_tbls[blkn];
      d_derived_tbl * actbl = entropy->ac_cur_tbls[blkn];
      register int s, k, r;

      /* Decode a single block's worth of coefficients */

      /* Section F.2.2.1: decode the DC coefficient difference */
      HUFF_DECODE(s, br_state, dctbl, return FALSE, label1);
      if (s) {
	CHECK_BIT_BUFFER(br_state, s, return FALSE);
	r = GET_BITS(s);
	s = HUFF_EXTEND(r, s);
      }

      if (entropy->dc_needed[blkn]) {
	/* Convert DC difference to actual value, update last_dc_val */
	int ci = cinfo->MCU_membership[blkn];
	s += state.last_dc_val[ci];
	state.last_dc_val[ci] = s;
	/* Output the DC coefficient (assumes jpeg_natural_order[0] = 0) */
	(*block)[0] = (JCOEF) s;
      }

      if (entropy->ac_needed[blkn]) {

	/* Section F.2.2.2: decode the AC coefficients */
	/* Since zeroes are skipped, output area must be cleared beforehand */
	for (k = 1; k < DCTSIZE2; k++) {
	  HUFF_DECODE(s, br_state, actbl, return FALSE, label2);
      
	  r = s >> 4;
	  s &= 15;
      
	  if (s) {
	    k += r;
	    CHECK_BIT_BUFFER(br_state, s, return FALSE);
	    r = GET_BITS(s);
	    s = HUFF_EXTEND(r, s);
	    /* Output coefficient in natural (dezigzagged) order.
	     * Note: the extra entries in jpeg_natural_order[] will save us
	     * if k >= DCTSIZE2, which could happen if the data is corrupted.
	     */
	    (*block)[jpeg_natural_order[k]] = (JCOEF) s;
	  } else {
	    if (r != 15)
	      break;
	    k += 15;
	  }
	}

      } else {

	/* Section F.2.2.2: decode the AC coefficients */
	/* In this path we just discard the values */
	for (k = 1; k < DCTSIZE2; k++) {
	  HUFF_DECODE(s, br_state, actbl, return FALSE, label3);
      
	  r = s >> 4;
	  s &= 15;
      
	  if (s) {
	    k += r;
	    CHECK_BIT_BUFFER(br_state, s, return FALSE);
	    DROP_BITS(s);
	  } else {
	    if (r != 15)
	      break;
	    k += 15;
	  }
	}

      }
    }

    /* Completed MCU, so update state */
    BITREAD_SAVE_STATE(cinfo,entropy->bitstate);
    ASSIGN_STATE(entropy->saved, state);
  }

  /* Account for restart interval (no-op if not using restarts) */
  entropy->restarts_to_go--;

  return TRUE;
}


/*
 * Module initialization routine for Huffman entropy decoding.
 */

GLOBAL(void)
jinit_huff_decoder (j_decompress_ptr cinfo)
{
  huff_entropy_ptr entropy;
  int i;

  entropy = (huff_entropy_ptr)
    (*cinfo->mem->alloc_small) ((j_common_ptr) cinfo, JPOOL_IMAGE,
				SIZEOF(huff_entropy_decoder));
  cinfo->entropy = (struct jpeg_entropy_decoder *) entropy;
  entropy->pub.start_pass = start_pass_huff_decoder;
  entropy->pub.decode_mcu = decode_mcu;

  /* Mark tables unallocated */
  for (i = 0; i < NUM_HUFF_TBLS; i++) {
    entropy->dc_derived_tbls[i] = entropy->ac_derived_tbls[i] = NULL;
  }
}