t1decode.c

Go to the documentation of this file.

/****************************************************************************
 *
 * t1decode.c
 *
 *   PostScript Type 1 decoding routines (body).
 *
 * Copyright (C) 2000-2020 by
 * David Turner, Robert Wilhelm, and Werner Lemberg.
 *
 * This file is part of the FreeType project, and may only be used,
 * modified, and distributed under the terms of the FreeType project
 * license, LICENSE.TXT.  By continuing to use, modify, or distribute
 * this file you indicate that you have read the license and
 * understand and accept it fully.
 *
 */
 
 
#include <freetype/internal/ftcalc.h>
#include <freetype/internal/ftdebug.h>
#include <freetype/internal/pshints.h>
#include <freetype/internal/fthash.h>
#include <freetype/ftoutln.h>
 
#include "t1decode.h"
#include "psobjs.h"
 
#include "psauxerr.h"
 
/* ensure proper sign extension */
#define Fix2Int( f )  ( (FT_Int)(FT_Short)( (f) >> 16 ) )
 
  /**************************************************************************
   *
   * The macro FT_COMPONENT is used in trace mode.  It is an implicit
   * parameter of the FT_TRACE() and FT_ERROR() macros, used to print/log
   * messages during execution.
   */
#undef  FT_COMPONENT
#define FT_COMPONENT  t1decode
 
 
  typedef enum  T1_Operator_
  {
    op_none = 0,
    op_endchar,
    op_hsbw,
    op_seac,
    op_sbw,
    op_closepath,
    op_hlineto,
    op_hmoveto,
    op_hvcurveto,
    op_rlineto,
    op_rmoveto,
    op_rrcurveto,
    op_vhcurveto,
    op_vlineto,
    op_vmoveto,
    op_dotsection,
    op_hstem,
    op_hstem3,
    op_vstem,
    op_vstem3,
    op_div,
    op_callothersubr,
    op_callsubr,
    op_pop,
    op_return,
    op_setcurrentpoint,
    op_unknown15,
 
    op_max    /* never remove this one */
 
  } T1_Operator;
 
 
  static
  const FT_Int  t1_args_count[op_max] =
  {
    0, /* none */
    0, /* endchar */
    2, /* hsbw */
    5, /* seac */
    4, /* sbw */
    0, /* closepath */
    1, /* hlineto */
    1, /* hmoveto */
    4, /* hvcurveto */
    2, /* rlineto */
    2, /* rmoveto */
    6, /* rrcurveto */
    4, /* vhcurveto */
    1, /* vlineto */
    1, /* vmoveto */
    0, /* dotsection */
    2, /* hstem */
    6, /* hstem3 */
    2, /* vstem */
    6, /* vstem3 */
    2, /* div */
   -1, /* callothersubr */
    1, /* callsubr */
    0, /* pop */
    0, /* return */
    2, /* setcurrentpoint */
    2  /* opcode 15 (undocumented and obsolete) */
  };
 
 
  /**************************************************************************
   *
   * @Function:
   *   t1_lookup_glyph_by_stdcharcode_ps
   *
   * @Description:
   *   Looks up a given glyph by its StandardEncoding charcode.  Used to
   *   implement the SEAC Type 1 operator in the Adobe engine
   *
   * @Input:
   *   face ::
   *     The current face object.
   *
   *   charcode ::
   *     The character code to look for.
   *
   * @Return:
   *   A glyph index in the font face.  Returns -1 if the corresponding
   *   glyph wasn't found.
   */
  FT_LOCAL_DEF( FT_Int )
  t1_lookup_glyph_by_stdcharcode_ps( PS_Decoder*  decoder,
                                     FT_Int       charcode )
  {
    FT_UInt             n;
    const FT_String*    glyph_name;
    FT_Service_PsCMaps  psnames = decoder->psnames;
 
 
    /* check range of standard char code */
    if ( charcode < 0 || charcode > 255 )
      return -1;
 
    glyph_name = psnames->adobe_std_strings(
                   psnames->adobe_std_encoding[charcode]);
 
    for ( n = 0; n < decoder->num_glyphs; n++ )
    {
      FT_String*  name = (FT_String*)decoder->glyph_names[n];
 
 
      if ( name                               &&
           name[0] == glyph_name[0]           &&
           ft_strcmp( name, glyph_name ) == 0 )
        return (FT_Int)n;
    }
 
    return -1;
  }
 
 
#ifdef T1_CONFIG_OPTION_OLD_ENGINE
 
  /**************************************************************************
   *
   * @Function:
   *   t1_lookup_glyph_by_stdcharcode
   *
   * @Description:
   *   Looks up a given glyph by its StandardEncoding charcode.  Used to
   *   implement the SEAC Type 1 operator.
   *
   * @Input:
   *   face ::
   *     The current face object.
   *
   *   charcode ::
   *     The character code to look for.
   *
   * @Return:
   *   A glyph index in the font face.  Returns -1 if the corresponding
   *   glyph wasn't found.
   */
  static FT_Int
  t1_lookup_glyph_by_stdcharcode( T1_Decoder  decoder,
                                  FT_Int      charcode )
  {
    FT_UInt             n;
    const FT_String*    glyph_name;
    FT_Service_PsCMaps  psnames = decoder->psnames;
 
 
    /* check range of standard char code */
    if ( charcode < 0 || charcode > 255 )
      return -1;
 
    glyph_name = psnames->adobe_std_strings(
                   psnames->adobe_std_encoding[charcode]);
 
    for ( n = 0; n < decoder->num_glyphs; n++ )
    {
      FT_String*  name = (FT_String*)decoder->glyph_names[n];
 
 
      if ( name                               &&
           name[0] == glyph_name[0]           &&
           ft_strcmp( name, glyph_name ) == 0 )
        return (FT_Int)n;
    }
 
    return -1;
  }
 
 
  /* parse a single Type 1 glyph */
  FT_LOCAL_DEF( FT_Error )
  t1_decoder_parse_glyph( T1_Decoder  decoder,
                          FT_UInt     glyph )
  {
    return decoder->parse_callback( decoder, glyph );
  }
 
 
  /**************************************************************************
   *
   * @Function:
   *   t1operator_seac
   *
   * @Description:
   *   Implements the `seac' Type 1 operator for a Type 1 decoder.
   *
   * @Input:
   *   decoder ::
   *     The current CID decoder.
   *
   *   asb ::
   *     The accent's side bearing.
   *
   *   adx ::
   *     The horizontal offset of the accent.
   *
   *   ady ::
   *     The vertical offset of the accent.
   *
   *   bchar ::
   *     The base character's StandardEncoding charcode.
   *
   *   achar ::
   *     The accent character's StandardEncoding charcode.
   *
   * @Return:
   *   FreeType error code.  0 means success.
   */
  static FT_Error
  t1operator_seac( T1_Decoder  decoder,
                   FT_Pos      asb,
                   FT_Pos      adx,
                   FT_Pos      ady,
                   FT_Int      bchar,
                   FT_Int      achar )
  {
    FT_Error     error;
    FT_Int       bchar_index, achar_index;
#if 0
    FT_Int       n_base_points;
    FT_Outline*  base = decoder->builder.base;
#endif
    FT_Vector    left_bearing, advance;
 
#ifdef FT_CONFIG_OPTION_INCREMENTAL
    T1_Face      face  = (T1_Face)decoder->builder.face;
#endif
 
 
    if ( decoder->seac )
    {
      FT_ERROR(( "t1operator_seac: invalid nested seac\n" ));
      return FT_THROW( Syntax_Error );
    }
 
    if ( decoder->builder.metrics_only )
    {
      FT_ERROR(( "t1operator_seac: unexpected seac\n" ));
      return FT_THROW( Syntax_Error );
    }
 
    /* seac weirdness */
    adx += decoder->builder.left_bearing.x;
 
    /* `glyph_names' is set to 0 for CID fonts which do not */
    /* include an encoding.  How can we deal with these?    */
#ifdef FT_CONFIG_OPTION_INCREMENTAL
    if ( decoder->glyph_names == 0                   &&
         !face->root.internal->incremental_interface )
#else
    if ( decoder->glyph_names == 0 )
#endif /* FT_CONFIG_OPTION_INCREMENTAL */
    {
      FT_ERROR(( "t1operator_seac:"
                 " glyph names table not available in this font\n" ));
      return FT_THROW( Syntax_Error );
    }
 
#ifdef FT_CONFIG_OPTION_INCREMENTAL
    if ( face->root.internal->incremental_interface )
    {
      /* the caller must handle the font encoding also */
      bchar_index = bchar;
      achar_index = achar;
    }
    else
#endif
    {
      bchar_index = t1_lookup_glyph_by_stdcharcode( decoder, bchar );
      achar_index = t1_lookup_glyph_by_stdcharcode( decoder, achar );
    }
 
    if ( bchar_index < 0 || achar_index < 0 )
    {
      FT_ERROR(( "t1operator_seac:"
                 " invalid seac character code arguments\n" ));
      return FT_THROW( Syntax_Error );
    }
 
    /* if we are trying to load a composite glyph, do not load the */
    /* accent character and return the array of subglyphs.         */
    if ( decoder->builder.no_recurse )
    {
      FT_GlyphSlot    glyph  = (FT_GlyphSlot)decoder->builder.glyph;
      FT_GlyphLoader  loader = glyph->internal->loader;
      FT_SubGlyph     subg;
 
 
      /* reallocate subglyph array if necessary */
      error = FT_GlyphLoader_CheckSubGlyphs( loader, 2 );
      if ( error )
        goto Exit;
 
      subg = loader->current.subglyphs;
 
      /* subglyph 0 = base character */
      subg->index = bchar_index;
      subg->flags = FT_SUBGLYPH_FLAG_ARGS_ARE_XY_VALUES |
                    FT_SUBGLYPH_FLAG_USE_MY_METRICS;
      subg->arg1  = 0;
      subg->arg2  = 0;
      subg++;
 
      /* subglyph 1 = accent character */
      subg->index = achar_index;
      subg->flags = FT_SUBGLYPH_FLAG_ARGS_ARE_XY_VALUES;
      subg->arg1  = (FT_Int)FIXED_TO_INT( adx - asb );
      subg->arg2  = (FT_Int)FIXED_TO_INT( ady );
 
      /* set up remaining glyph fields */
      glyph->num_subglyphs = 2;
      glyph->subglyphs     = loader->base.subglyphs;
      glyph->format        = FT_GLYPH_FORMAT_COMPOSITE;
 
      loader->current.num_subglyphs = 2;
      goto Exit;
    }
 
    /* First load `bchar' in builder */
    /* now load the unscaled outline */
 
    FT_GlyphLoader_Prepare( decoder->builder.loader );  /* prepare loader */
 
    /* save the left bearing and width of the SEAC   */
    /* glyph as they will be erased by the next load */
 
    left_bearing = decoder->builder.left_bearing;
    advance      = decoder->builder.advance;
 
    /* the seac operator must not be nested */
    decoder->seac = TRUE;
    error = t1_decoder_parse_glyph( decoder, (FT_UInt)bchar_index );
    decoder->seac = FALSE;
    if ( error )
      goto Exit;
 
    /* If the SEAC glyph doesn't have a (H)SBW of its */
    /* own use the values from the base glyph.        */
 
    if ( decoder->builder.parse_state != T1_Parse_Have_Width )
    {
      left_bearing = decoder->builder.left_bearing;
      advance      = decoder->builder.advance;
    }
 
    decoder->builder.left_bearing.x = 0;
    decoder->builder.left_bearing.y = 0;
 
    decoder->builder.pos_x = adx - asb;
    decoder->builder.pos_y = ady;
 
    /* Now load `achar' on top of */
    /* the base outline           */
 
    /* the seac operator must not be nested */
    decoder->seac = TRUE;
    error = t1_decoder_parse_glyph( decoder, (FT_UInt)achar_index );
    decoder->seac = FALSE;
    if ( error )
      goto Exit;
 
    /* restore the left side bearing and advance width   */
    /* of the SEAC glyph or base character (saved above) */
 
    decoder->builder.left_bearing = left_bearing;
    decoder->builder.advance      = advance;
 
    decoder->builder.pos_x = 0;
    decoder->builder.pos_y = 0;
 
  Exit:
    return error;
  }
 
 
  /**************************************************************************
   *
   * @Function:
   *   t1_decoder_parse_charstrings
   *
   * @Description:
   *   Parses a given Type 1 charstrings program.
   *
   * @Input:
   *   decoder ::
   *     The current Type 1 decoder.
   *
   *   charstring_base ::
   *     The base address of the charstring stream.
   *
   *   charstring_len ::
   *     The length in bytes of the charstring stream.
   *
   * @Return:
   *   FreeType error code.  0 means success.
   */
  FT_LOCAL_DEF( FT_Error )
  t1_decoder_parse_charstrings( T1_Decoder  decoder,
                                FT_Byte*    charstring_base,
                                FT_UInt     charstring_len )
  {
    FT_Error         error;
    T1_Decoder_Zone  zone;
    FT_Byte*         ip;
    FT_Byte*         limit;
    T1_Builder       builder = &decoder->builder;
    FT_Pos           x, y, orig_x, orig_y;
    FT_Int           known_othersubr_result_cnt   = 0;
    FT_Int           unknown_othersubr_result_cnt = 0;
    FT_Bool          large_int;
    FT_Fixed         seed;
 
    T1_Hints_Funcs   hinter;
 
#ifdef FT_DEBUG_LEVEL_TRACE
    FT_Bool          bol = TRUE;
#endif
 
 
    /* compute random seed from stack address of parameter */
    seed = (FT_Fixed)( ( (FT_Offset)(char*)&seed            ^
                         (FT_Offset)(char*)&decoder         ^
                         (FT_Offset)(char*)&charstring_base ) &
                         FT_ULONG_MAX                         );
    seed = ( seed ^ ( seed >> 10 ) ^ ( seed >> 20 ) ) & 0xFFFFL;
    if ( seed == 0 )
      seed = 0x7384;
 
    /* First of all, initialize the decoder */
    decoder->top  = decoder->stack;
    decoder->zone = decoder->zones;
    zone          = decoder->zones;
 
    builder->parse_state = T1_Parse_Start;
 
    hinter = (T1_Hints_Funcs)builder->hints_funcs;
 
    /* a font that reads BuildCharArray without setting */
    /* its values first is buggy, but ...               */
    FT_ASSERT( ( decoder->len_buildchar == 0 ) ==
               ( decoder->buildchar == NULL )  );
 
    if ( decoder->buildchar && decoder->len_buildchar > 0 )
      FT_ARRAY_ZERO( decoder->buildchar, decoder->len_buildchar );
 
    zone->base           = charstring_base;
    limit = zone->limit  = charstring_base + charstring_len;
    ip    = zone->cursor = zone->base;
 
    error = FT_Err_Ok;
 
    x = orig_x = builder->pos_x;
    y = orig_y = builder->pos_y;
 
    /* begin hints recording session, if any */
    if ( hinter )
      hinter->open( hinter->hints );
 
    large_int = FALSE;
 
    /* now, execute loop */
    while ( ip < limit )
    {
      FT_Long*     top   = decoder->top;
      T1_Operator  op    = op_none;
      FT_Int32     value = 0;
 
 
      FT_ASSERT( known_othersubr_result_cnt == 0   ||
                 unknown_othersubr_result_cnt == 0 );
 
#ifdef FT_DEBUG_LEVEL_TRACE
      if ( bol )
      {
        FT_TRACE5(( " (%d)", decoder->top - decoder->stack ));
        bol = FALSE;
      }
#endif
 
      /**********************************************************************
       *
       * Decode operator or operand
       *
       */
 
      /* first of all, decompress operator or value */
      switch ( *ip++ )
      {
      case 1:
        op = op_hstem;
        break;
 
      case 3:
        op = op_vstem;
        break;
      case 4:
        op = op_vmoveto;
        break;
      case 5:
        op = op_rlineto;
        break;
      case 6:
        op = op_hlineto;
        break;
      case 7:
        op = op_vlineto;
        break;
      case 8:
        op = op_rrcurveto;
        break;
      case 9:
        op = op_closepath;
        break;
      case 10:
        op = op_callsubr;
        break;
      case 11:
        op = op_return;
        break;
 
      case 13:
        op = op_hsbw;
        break;
      case 14:
        op = op_endchar;
        break;
 
      case 15:          /* undocumented, obsolete operator */
        op = op_unknown15;
        break;
 
      case 21:
        op = op_rmoveto;
        break;
      case 22:
        op = op_hmoveto;
        break;
 
      case 30:
        op = op_vhcurveto;
        break;
      case 31:
        op = op_hvcurveto;
        break;
 
      case 12:
        if ( ip >= limit )
        {
          FT_ERROR(( "t1_decoder_parse_charstrings:"
                     " invalid escape (12+EOF)\n" ));
          goto Syntax_Error;
        }
 
        switch ( *ip++ )
        {
        case 0:
          op = op_dotsection;
          break;
        case 1:
          op = op_vstem3;
          break;
        case 2:
          op = op_hstem3;
          break;
        case 6:
          op = op_seac;
          break;
        case 7:
          op = op_sbw;
          break;
        case 12:
          op = op_div;
          break;
        case 16:
          op = op_callothersubr;
          break;
        case 17:
          op = op_pop;
          break;
        case 33:
          op = op_setcurrentpoint;
          break;
 
        default:
          FT_ERROR(( "t1_decoder_parse_charstrings:"
                     " invalid escape (12+%d)\n",
                     ip[-1] ));
          goto Syntax_Error;
        }
        break;
 
      case 255:    /* four bytes integer */
        if ( ip + 4 > limit )
        {
          FT_ERROR(( "t1_decoder_parse_charstrings:"
                     " unexpected EOF in integer\n" ));
          goto Syntax_Error;
        }
 
        value = (FT_Int32)( ( (FT_UInt32)ip[0] << 24 ) |
                            ( (FT_UInt32)ip[1] << 16 ) |
                            ( (FT_UInt32)ip[2] << 8  ) |
                              (FT_UInt32)ip[3]         );
        ip += 4;
 
        /* According to the specification, values > 32000 or < -32000 must */
        /* be followed by a `div' operator to make the result be in the    */
        /* range [-32000;32000].  We expect that the second argument of    */
        /* `div' is not a large number.  Additionally, we don't handle     */
        /* stuff like `<large1> <large2> <num> div <num> div' or           */
        /* <large1> <large2> <num> div div'.  This is probably not allowed */
        /* anyway.                                                         */
        if ( value > 32000 || value < -32000 )
        {
          if ( large_int )
            FT_ERROR(( "t1_decoder_parse_charstrings:"
                       " no `div' after large integer\n" ));
          else
            large_int = TRUE;
        }
        else
        {
          if ( !large_int )
            value = (FT_Int32)( (FT_UInt32)value << 16 );
        }
 
        break;
 
      default:
        if ( ip[-1] >= 32 )
        {
          if ( ip[-1] < 247 )
            value = (FT_Int32)ip[-1] - 139;
          else
          {
            if ( ++ip > limit )
            {
              FT_ERROR(( "t1_decoder_parse_charstrings:"
                         " unexpected EOF in integer\n" ));
              goto Syntax_Error;
            }
 
            if ( ip[-2] < 251 )
              value =    ( ( ip[-2] - 247 ) * 256 ) + ip[-1] + 108;
            else
              value = -( ( ( ip[-2] - 251 ) * 256 ) + ip[-1] + 108 );
          }
 
          if ( !large_int )
            value = (FT_Int32)( (FT_UInt32)value << 16 );
        }
        else
        {
          FT_ERROR(( "t1_decoder_parse_charstrings:"
                     " invalid byte (%d)\n", ip[-1] ));
          goto Syntax_Error;
        }
      }
 
      if ( unknown_othersubr_result_cnt > 0 )
      {
        switch ( op )
        {
        case op_callsubr:
        case op_return:
        case op_none:
        case op_pop:
          break;
 
        default:
          /* all operands have been transferred by previous pops */
          unknown_othersubr_result_cnt = 0;
          break;
        }
      }
 
      if ( large_int && !( op == op_none || op == op_div ) )
      {
        FT_ERROR(( "t1_decoder_parse_charstrings:"
                   " no `div' after large integer\n" ));
 
        large_int = FALSE;
      }
 
      /**********************************************************************
       *
       * Push value on stack, or process operator
       *
       */
      if ( op == op_none )
      {
        if ( top - decoder->stack >= T1_MAX_CHARSTRINGS_OPERANDS )
        {
          FT_ERROR(( "t1_decoder_parse_charstrings: stack overflow\n" ));
          goto Syntax_Error;
        }
 
#ifdef FT_DEBUG_LEVEL_TRACE
        if ( large_int )
          FT_TRACE4(( " %d", value ));
        else
          FT_TRACE4(( " %d", value / 65536 ));
#endif
 
        *top++       = value;
        decoder->top = top;
      }
      else if ( op == op_callothersubr )  /* callothersubr */
      {
        FT_Int  subr_no;
        FT_Int  arg_cnt;
 
 
#ifdef FT_DEBUG_LEVEL_TRACE
        FT_TRACE4(( " callothersubr\n" ));
        bol = TRUE;
#endif
 
        if ( top - decoder->stack < 2 )
          goto Stack_Underflow;
 
        top -= 2;
 
        subr_no = Fix2Int( top[1] );
        arg_cnt = Fix2Int( top[0] );
 
        /************************************************************
         *
         * remove all operands to callothersubr from the stack
         *
         * for handled othersubrs, where we know the number of
         * arguments, we increase the stack by the value of
         * known_othersubr_result_cnt
         *
         * for unhandled othersubrs the following pops adjust the
         * stack pointer as necessary
         */
 
        if ( arg_cnt > top - decoder->stack )
          goto Stack_Underflow;
 
        top -= arg_cnt;
 
        known_othersubr_result_cnt   = 0;
        unknown_othersubr_result_cnt = 0;
 
        /* XXX TODO: The checks to `arg_count == <whatever>'       */
        /* might not be correct; an othersubr expects a certain    */
        /* number of operands on the PostScript stack (as opposed  */
        /* to the T1 stack) but it doesn't have to put them there  */
        /* by itself; previous othersubrs might have left the      */
        /* operands there if they were not followed by an          */
        /* appropriate number of pops                              */
        /*                                                         */
        /* On the other hand, Adobe Reader 7.0.8 for Linux doesn't */
        /* accept a font that contains charstrings like            */
        /*                                                         */
        /*     100 200 2 20 callothersubr                          */
        /*     300 1 20 callothersubr pop                          */
        /*                                                         */
        /* Perhaps this is the reason why BuildCharArray exists.   */
 
        switch ( subr_no )
        {
        case 0:                     /* end flex feature */
          if ( arg_cnt != 3 )
            goto Unexpected_OtherSubr;
 
          if ( !decoder->flex_state           ||
               decoder->num_flex_vectors != 7 )
          {
            FT_ERROR(( "t1_decoder_parse_charstrings:"
                       " unexpected flex end\n" ));
            goto Syntax_Error;
          }
 
          /* the two `results' are popped by the following setcurrentpoint */
          top[0] = x;
          top[1] = y;
          known_othersubr_result_cnt = 2;
          break;
 
        case 1:                     /* start flex feature */
          if ( arg_cnt != 0 )
            goto Unexpected_OtherSubr;
 
          if ( FT_SET_ERROR( t1_builder_start_point( builder, x, y ) ) ||
               FT_SET_ERROR( t1_builder_check_points( builder, 6 ) )   )
            goto Fail;
 
          decoder->flex_state        = 1;
          decoder->num_flex_vectors  = 0;
          break;
 
        case 2:                     /* add flex vectors */
          {
            FT_Int  idx;
 
 
            if ( arg_cnt != 0 )
              goto Unexpected_OtherSubr;
 
            if ( !decoder->flex_state )
            {
              FT_ERROR(( "t1_decoder_parse_charstrings:"
                         " missing flex start\n" ));
              goto Syntax_Error;
            }
 
            /* note that we should not add a point for index 0; */
            /* this will move our current position to the flex  */
            /* point without adding any point to the outline    */
            idx = decoder->num_flex_vectors++;
            if ( idx > 0 && idx < 7 )
            {
              /* in malformed fonts it is possible to have other */
              /* opcodes in the middle of a flex (which don't    */
              /* increase `num_flex_vectors'); we thus have to   */
              /* check whether we can add a point                */
              if ( FT_SET_ERROR( t1_builder_check_points( builder, 1 ) ) )
                goto Syntax_Error;
 
              t1_builder_add_point( builder,
                                    x,
                                    y,
                                    (FT_Byte)( idx == 3 || idx == 6 ) );
            }
          }
          break;
 
        case 3:                     /* change hints */
          if ( arg_cnt != 1 )
            goto Unexpected_OtherSubr;
 
          known_othersubr_result_cnt = 1;
 
          if ( hinter )
            hinter->reset( hinter->hints,
                           (FT_UInt)builder->current->n_points );
          break;
 
        case 12:
        case 13:
          /* counter control hints, clear stack */
          top = decoder->stack;
          break;
 
        case 14:
        case 15:
        case 16:
        case 17:
        case 18:                    /* multiple masters */
          {
            PS_Blend  blend = decoder->blend;
            FT_UInt   num_points, nn, mm;
            FT_Long*  delta;
            FT_Long*  values;
 
 
            if ( !blend )
            {
              FT_ERROR(( "t1_decoder_parse_charstrings:"
                         " unexpected multiple masters operator\n" ));
              goto Syntax_Error;
            }
 
            num_points = (FT_UInt)subr_no - 13 + ( subr_no == 18 );
            if ( arg_cnt != (FT_Int)( num_points * blend->num_designs ) )
            {
              FT_ERROR(( "t1_decoder_parse_charstrings:"
                         " incorrect number of multiple masters arguments\n" ));
              goto Syntax_Error;
            }
 
            /* We want to compute                                    */
            /*                                                       */
            /*   a0*w0 + a1*w1 + ... + ak*wk                         */
            /*                                                       */
            /* but we only have a0, a1-a0, a2-a0, ..., ak-a0.        */
            /*                                                       */
            /* However, given that w0 + w1 + ... + wk == 1, we can   */
            /* rewrite it easily as                                  */
            /*                                                       */
            /*   a0 + (a1-a0)*w1 + (a2-a0)*w2 + ... + (ak-a0)*wk     */
            /*                                                       */
            /* where k == num_designs-1.                             */
            /*                                                       */
            /* I guess that's why it's written in this `compact'     */
            /* form.                                                 */
            /*                                                       */
            delta  = top + num_points;
            values = top;
            for ( nn = 0; nn < num_points; nn++ )
            {
              FT_Long  tmp = values[0];
 
 
              for ( mm = 1; mm < blend->num_designs; mm++ )
                tmp = ADD_LONG( tmp,
                                FT_MulFix( *delta++,
                                           blend->weight_vector[mm] ) );
 
              *values++ = tmp;
            }
 
            known_othersubr_result_cnt = (FT_Int)num_points;
            break;
          }
 
        case 19:
          /* <idx> 1 19 callothersubr                             */
          /* => replace elements starting from index cvi( <idx> ) */
          /*    of BuildCharArray with WeightVector               */
          {
            FT_Int    idx;
            PS_Blend  blend = decoder->blend;
 
 
            if ( arg_cnt != 1 || !blend )
              goto Unexpected_OtherSubr;
 
            idx = Fix2Int( top[0] );
 
            if ( idx < 0                                                    ||
                 (FT_UInt)idx + blend->num_designs > decoder->len_buildchar )
              goto Unexpected_OtherSubr;
 
            ft_memcpy( &decoder->buildchar[idx],
                       blend->weight_vector,
                       blend->num_designs *
                         sizeof ( blend->weight_vector[0] ) );
          }
          break;
 
        case 20:
          /* <arg1> <arg2> 2 20 callothersubr pop   */
          /* ==> push <arg1> + <arg2> onto T1 stack */
          if ( arg_cnt != 2 )
            goto Unexpected_OtherSubr;
 
          top[0] = ADD_LONG( top[0], top[1] );
 
          known_othersubr_result_cnt = 1;
          break;
 
        case 21:
          /* <arg1> <arg2> 2 21 callothersubr pop   */
          /* ==> push <arg1> - <arg2> onto T1 stack */
          if ( arg_cnt != 2 )
            goto Unexpected_OtherSubr;
 
          top[0] = SUB_LONG( top[0], top[1] );
 
          known_othersubr_result_cnt = 1;
          break;
 
        case 22:
          /* <arg1> <arg2> 2 22 callothersubr pop   */
          /* ==> push <arg1> * <arg2> onto T1 stack */
          if ( arg_cnt != 2 )
            goto Unexpected_OtherSubr;
 
          top[0] = FT_MulFix( top[0], top[1] );
 
          known_othersubr_result_cnt = 1;
          break;
 
        case 23:
          /* <arg1> <arg2> 2 23 callothersubr pop   */
          /* ==> push <arg1> / <arg2> onto T1 stack */
          if ( arg_cnt != 2 || top[1] == 0 )
            goto Unexpected_OtherSubr;
 
          top[0] = FT_DivFix( top[0], top[1] );
 
          known_othersubr_result_cnt = 1;
          break;
 
        case 24:
          /* <val> <idx> 2 24 callothersubr               */
          /* ==> set BuildCharArray[cvi( <idx> )] = <val> */
          {
            FT_Int    idx;
            PS_Blend  blend = decoder->blend;
 
 
            if ( arg_cnt != 2 || !blend )
              goto Unexpected_OtherSubr;
 
            idx = Fix2Int( top[1] );
 
            if ( idx < 0 || (FT_UInt) idx >= decoder->len_buildchar )
              goto Unexpected_OtherSubr;
 
            decoder->buildchar[idx] = top[0];
          }
          break;
 
        case 25:
          /* <idx> 1 25 callothersubr pop        */
          /* ==> push BuildCharArray[cvi( idx )] */
          /*     onto T1 stack                   */
          {
            FT_Int    idx;
            PS_Blend  blend = decoder->blend;
 
 
            if ( arg_cnt != 1 || !blend )
              goto Unexpected_OtherSubr;
 
            idx = Fix2Int( top[0] );
 
            if ( idx < 0 || (FT_UInt) idx >= decoder->len_buildchar )
              goto Unexpected_OtherSubr;
 
            top[0] = decoder->buildchar[idx];
          }
 
          known_othersubr_result_cnt = 1;
          break;
 
#if 0
        case 26:
          /* <val> mark <idx> ==> set BuildCharArray[cvi( <idx> )] = <val>, */
          /*                      leave mark on T1 stack                    */
          /* <val> <idx>      ==> set BuildCharArray[cvi( <idx> )] = <val>  */
          XXX which routine has left its mark on the (PostScript) stack?;
          break;
#endif
 
        case 27:
          /* <res1> <res2> <val1> <val2> 4 27 callothersubr pop */
          /* ==> push <res1> onto T1 stack if <val1> <= <val2>, */
          /*     otherwise push <res2>                          */
          if ( arg_cnt != 4 )
            goto Unexpected_OtherSubr;
 
          if ( top[2] > top[3] )
            top[0] = top[1];
 
          known_othersubr_result_cnt = 1;
          break;
 
        case 28:
          /* 0 28 callothersubr pop                               */
          /* => push random value from interval [0, 1) onto stack */
          if ( arg_cnt != 0 )
            goto Unexpected_OtherSubr;
 
          {
            FT_Fixed  Rand;
 
 
            Rand = seed;
            if ( Rand >= 0x8000L )
              Rand++;
 
            top[0] = Rand;
 
            seed = FT_MulFix( seed, 0x10000L - seed );
            if ( seed == 0 )
              seed += 0x2873;
          }
 
          known_othersubr_result_cnt = 1;
          break;
 
        default:
          if ( arg_cnt >= 0 && subr_no >= 0 )
          {
            FT_ERROR(( "t1_decoder_parse_charstrings:"
                       " unknown othersubr [%d %d], wish me luck\n",
                       arg_cnt, subr_no ));
            unknown_othersubr_result_cnt = arg_cnt;
            break;
          }
          /* fall through */
 
        Unexpected_OtherSubr:
          FT_ERROR(( "t1_decoder_parse_charstrings:"
                     " invalid othersubr [%d %d]\n", arg_cnt, subr_no ));
          goto Syntax_Error;
        }
 
        top += known_othersubr_result_cnt;
 
        decoder->top = top;
      }
      else  /* general operator */
      {
        FT_Int  num_args = t1_args_count[op];
 
 
        FT_ASSERT( num_args >= 0 );
 
        if ( top - decoder->stack < num_args )
          goto Stack_Underflow;
 
        /* XXX Operators usually take their operands from the        */
        /*     bottom of the stack, i.e., the operands are           */
        /*     decoder->stack[0], ..., decoder->stack[num_args - 1]; */
        /*     only div, callsubr, and callothersubr are different.  */
        /*     In practice it doesn't matter (?).                    */
 
#ifdef FT_DEBUG_LEVEL_TRACE
 
        switch ( op )
        {
        case op_callsubr:
        case op_div:
        case op_callothersubr:
        case op_pop:
        case op_return:
          break;
 
        default:
          if ( top - decoder->stack != num_args )
            FT_TRACE0(( "t1_decoder_parse_charstrings:"
                        " too much operands on the stack"
                        " (seen %d, expected %d)\n",
                        top - decoder->stack, num_args ));
            break;
        }
 
#endif /* FT_DEBUG_LEVEL_TRACE */
 
        top -= num_args;
 
        switch ( op )
        {
        case op_endchar:
          FT_TRACE4(( " endchar\n" ));
 
          t1_builder_close_contour( builder );
 
          /* close hints recording session */
          if ( hinter )
          {
            if ( hinter->close( hinter->hints,
                                (FT_UInt)builder->current->n_points ) )
              goto Syntax_Error;
 
            /* apply hints to the loaded glyph outline now */
            error = hinter->apply( hinter->hints,
                                   builder->current,
                                   (PSH_Globals)builder->hints_globals,
                                   decoder->hint_mode );
            if ( error )
              goto Fail;
          }
 
          /* add current outline to the glyph slot */
          FT_GlyphLoader_Add( builder->loader );
 
          /* the compiler should optimize away this empty loop but ... */
 
#ifdef FT_DEBUG_LEVEL_TRACE
 
          if ( decoder->len_buildchar > 0 )
          {
            FT_UInt  i;
 
 
            FT_TRACE4(( "BuildCharArray = [ " ));
 
            for ( i = 0; i < decoder->len_buildchar; i++ )
              FT_TRACE4(( "%d ", decoder->buildchar[i] ));
 
            FT_TRACE4(( "]\n" ));
          }
 
#endif /* FT_DEBUG_LEVEL_TRACE */
 
          FT_TRACE4(( "\n" ));
 
          /* return now! */
          return FT_Err_Ok;
 
        case op_hsbw:
          FT_TRACE4(( " hsbw" ));
 
          builder->parse_state = T1_Parse_Have_Width;
 
          builder->left_bearing.x = ADD_LONG( builder->left_bearing.x,
                                              top[0] );
 
          builder->advance.x = top[1];
          builder->advance.y = 0;
 
          orig_x = x = ADD_LONG( builder->pos_x, top[0] );
          orig_y = y = builder->pos_y;
 
          FT_UNUSED( orig_y );
 
          /* `metrics_only' indicates that we only want to compute the */
          /* glyph's metrics (lsb + advance width) without loading the */
          /* rest of it; so exit immediately                           */
          if ( builder->metrics_only )
          {
            FT_TRACE4(( "\n" ));
            return FT_Err_Ok;
          }
 
          break;
 
        case op_seac:
          return t1operator_seac( decoder,
                                  top[0],
                                  top[1],
                                  top[2],
                                  Fix2Int( top[3] ),
                                  Fix2Int( top[4] ) );
 
        case op_sbw:
          FT_TRACE4(( " sbw" ));
 
          builder->parse_state = T1_Parse_Have_Width;
 
          builder->left_bearing.x = ADD_LONG( builder->left_bearing.x,
                                              top[0] );
          builder->left_bearing.y = ADD_LONG( builder->left_bearing.y,
                                              top[1] );
 
          builder->advance.x = top[2];
          builder->advance.y = top[3];
 
          x = ADD_LONG( builder->pos_x, top[0] );
          y = ADD_LONG( builder->pos_y, top[1] );
 
          /* `metrics_only' indicates that we only want to compute the */
          /* glyph's metrics (lsb + advance width) without loading the */
          /* rest of it; so exit immediately                           */
          if ( builder->metrics_only )
          {
            FT_TRACE4(( "\n" ));
            return FT_Err_Ok;
          }
 
          break;
 
        case op_closepath:
          FT_TRACE4(( " closepath" ));
 
          /* if there is no path, `closepath' is a no-op */
          if ( builder->parse_state == T1_Parse_Have_Path   ||
               builder->parse_state == T1_Parse_Have_Moveto )
            t1_builder_close_contour( builder );
 
          builder->parse_state = T1_Parse_Have_Width;
          break;
 
        case op_hlineto:
          FT_TRACE4(( " hlineto" ));
 
          if ( FT_SET_ERROR( t1_builder_start_point( builder, x, y ) ) )
            goto Fail;
 
          x = ADD_LONG( x, top[0] );
          goto Add_Line;
 
        case op_hmoveto:
          FT_TRACE4(( " hmoveto" ));
 
          x = ADD_LONG( x, top[0] );
 
          if ( !decoder->flex_state )
          {
            if ( builder->parse_state == T1_Parse_Start )
              goto Syntax_Error;
            builder->parse_state = T1_Parse_Have_Moveto;
          }
          break;
 
        case op_hvcurveto:
          FT_TRACE4(( " hvcurveto" ));
 
          if ( FT_SET_ERROR( t1_builder_start_point( builder, x, y ) ) ||
               FT_SET_ERROR( t1_builder_check_points( builder, 3 ) )   )
            goto Fail;
 
          x = ADD_LONG( x, top[0] );
          t1_builder_add_point( builder, x, y, 0 );
 
          x = ADD_LONG( x, top[1] );
          y = ADD_LONG( y, top[2] );
          t1_builder_add_point( builder, x, y, 0 );
 
          y = ADD_LONG( y, top[3] );
          t1_builder_add_point( builder, x, y, 1 );
          break;
 
        case op_rlineto:
          FT_TRACE4(( " rlineto" ));
 
          if ( FT_SET_ERROR( t1_builder_start_point( builder, x, y ) ) )
            goto Fail;
 
          x = ADD_LONG( x, top[0] );
          y = ADD_LONG( y, top[1] );
 
        Add_Line:
          if ( FT_SET_ERROR( t1_builder_add_point1( builder, x, y ) ) )
            goto Fail;
          break;
 
        case op_rmoveto:
          FT_TRACE4(( " rmoveto" ));
 
          x = ADD_LONG( x, top[0] );
          y = ADD_LONG( y, top[1] );
 
          if ( !decoder->flex_state )
          {
            if ( builder->parse_state == T1_Parse_Start )
              goto Syntax_Error;
            builder->parse_state = T1_Parse_Have_Moveto;
          }
          break;
 
        case op_rrcurveto:
          FT_TRACE4(( " rrcurveto" ));
 
          if ( FT_SET_ERROR( t1_builder_start_point( builder, x, y ) ) ||
               FT_SET_ERROR( t1_builder_check_points( builder, 3 ) )   )
            goto Fail;
 
          x = ADD_LONG( x, top[0] );
          y = ADD_LONG( y, top[1] );
          t1_builder_add_point( builder, x, y, 0 );
 
          x = ADD_LONG( x, top[2] );
          y = ADD_LONG( y, top[3] );
          t1_builder_add_point( builder, x, y, 0 );
 
          x = ADD_LONG( x, top[4] );
          y = ADD_LONG( y, top[5] );
          t1_builder_add_point( builder, x, y, 1 );
          break;
 
        case op_vhcurveto:
          FT_TRACE4(( " vhcurveto" ));
 
          if ( FT_SET_ERROR( t1_builder_start_point( builder, x, y ) ) ||
               FT_SET_ERROR( t1_builder_check_points( builder, 3 ) )   )
            goto Fail;
 
          y = ADD_LONG( y, top[0] );
          t1_builder_add_point( builder, x, y, 0 );
 
          x = ADD_LONG( x, top[1] );
          y = ADD_LONG( y, top[2] );
          t1_builder_add_point( builder, x, y, 0 );
 
          x = ADD_LONG( x, top[3] );
          t1_builder_add_point( builder, x, y, 1 );
          break;
 
        case op_vlineto:
          FT_TRACE4(( " vlineto" ));
 
          if ( FT_SET_ERROR( t1_builder_start_point( builder, x, y ) ) )
            goto Fail;
 
          y = ADD_LONG( y, top[0] );
          goto Add_Line;
 
        case op_vmoveto:
          FT_TRACE4(( " vmoveto" ));
 
          y = ADD_LONG( y, top[0] );
 
          if ( !decoder->flex_state )
          {
            if ( builder->parse_state == T1_Parse_Start )
              goto Syntax_Error;
            builder->parse_state = T1_Parse_Have_Moveto;
          }
          break;
 
        case op_div:
          FT_TRACE4(( " div" ));
 
          /* if `large_int' is set, we divide unscaled numbers; */
          /* otherwise, we divide numbers in 16.16 format --    */
          /* in both cases, it is the same operation            */
          *top = FT_DivFix( top[0], top[1] );
          top++;
 
          large_int = FALSE;
          break;
 
        case op_callsubr:
          {
            FT_Int  idx;
 
 
            FT_TRACE4(( " callsubr" ));
 
            idx = Fix2Int( top[0] );
 
            if ( decoder->subrs_hash )
            {
              size_t*  val = ft_hash_num_lookup( idx,
                                                 decoder->subrs_hash );
 
 
              if ( val )
                idx = *val;
              else
                idx = -1;
            }
 
            if ( idx < 0 || idx >= decoder->num_subrs )
            {
              FT_ERROR(( "t1_decoder_parse_charstrings:"
                         " invalid subrs index\n" ));
              goto Syntax_Error;
            }
 
            if ( zone - decoder->zones >= T1_MAX_SUBRS_CALLS )
            {
              FT_ERROR(( "t1_decoder_parse_charstrings:"
                         " too many nested subrs\n" ));
              goto Syntax_Error;
            }
 
            zone->cursor = ip;  /* save current instruction pointer */
 
            zone++;
 
            /* The Type 1 driver stores subroutines without the seed bytes. */
            /* The CID driver stores subroutines with seed bytes.  This     */
            /* case is taken care of when decoder->subrs_len == 0.          */
            zone->base = decoder->subrs[idx];
 
            if ( decoder->subrs_len )
              zone->limit = zone->base + decoder->subrs_len[idx];
            else
            {
              /* We are using subroutines from a CID font.  We must adjust */
              /* for the seed bytes.                                       */
              zone->base  += ( decoder->lenIV >= 0 ? decoder->lenIV : 0 );
              zone->limit  = decoder->subrs[idx + 1];
            }
 
            zone->cursor = zone->base;
 
            if ( !zone->base )
            {
              FT_ERROR(( "t1_decoder_parse_charstrings:"
                         " invoking empty subrs\n" ));
              goto Syntax_Error;
            }
 
            decoder->zone = zone;
            ip            = zone->base;
            limit         = zone->limit;
            break;
          }
 
        case op_pop:
          FT_TRACE4(( " pop" ));
 
          if ( known_othersubr_result_cnt > 0 )
          {
            known_othersubr_result_cnt--;
            /* ignore, we pushed the operands ourselves */
            break;
          }
 
          if ( unknown_othersubr_result_cnt == 0 )
          {
            FT_ERROR(( "t1_decoder_parse_charstrings:"
                       " no more operands for othersubr\n" ));
            goto Syntax_Error;
          }
 
          unknown_othersubr_result_cnt--;
          top++;   /* `push' the operand to callothersubr onto the stack */
          break;
 
        case op_return:
          FT_TRACE4(( " return" ));
 
          if ( zone <= decoder->zones )
          {
            FT_ERROR(( "t1_decoder_parse_charstrings:"
                       " unexpected return\n" ));
            goto Syntax_Error;
          }
 
          zone--;
          ip            = zone->cursor;
          limit         = zone->limit;
          decoder->zone = zone;
          break;
 
        case op_dotsection:
          FT_TRACE4(( " dotsection" ));
 
          break;
 
        case op_hstem:
          FT_TRACE4(( " hstem" ));
 
          /* record horizontal hint */
          if ( hinter )
          {
            /* top[0] += builder->left_bearing.y; */
            hinter->stem( hinter->hints, 1, top );
          }
          break;
 
        case op_hstem3:
          FT_TRACE4(( " hstem3" ));
 
          /* record horizontal counter-controlled hints */
          if ( hinter )
            hinter->stem3( hinter->hints, 1, top );
          break;
 
        case op_vstem:
          FT_TRACE4(( " vstem" ));
 
          /* record vertical hint */
          if ( hinter )
          {
            top[0] = ADD_LONG( top[0], orig_x );
            hinter->stem( hinter->hints, 0, top );
          }
          break;
 
        case op_vstem3:
          FT_TRACE4(( " vstem3" ));
 
          /* record vertical counter-controlled hints */
          if ( hinter )
          {
            FT_Pos  dx = orig_x;
 
 
            top[0] = ADD_LONG( top[0], dx );
            top[2] = ADD_LONG( top[2], dx );
            top[4] = ADD_LONG( top[4], dx );
            hinter->stem3( hinter->hints, 0, top );
          }
          break;
 
        case op_setcurrentpoint:
          FT_TRACE4(( " setcurrentpoint" ));
 
          /* From the T1 specification, section 6.4:                */
          /*                                                        */
          /*   The setcurrentpoint command is used only in          */
          /*   conjunction with results from OtherSubrs procedures. */
 
          /* known_othersubr_result_cnt != 0 is already handled     */
          /* above.                                                 */
 
          /* Note, however, that both Ghostscript and Adobe         */
          /* Distiller handle this situation by silently ignoring   */
          /* the inappropriate `setcurrentpoint' instruction.  So   */
          /* we do the same.                                        */
#if 0
 
          if ( decoder->flex_state != 1 )
          {
            FT_ERROR(( "t1_decoder_parse_charstrings:"
                       " unexpected `setcurrentpoint'\n" ));
            goto Syntax_Error;
          }
          else
            ...
#endif
 
          x = top[0];
          y = top[1];
          decoder->flex_state = 0;
          break;
 
        case op_unknown15:
          FT_TRACE4(( " opcode_15" ));
          /* nothing to do except to pop the two arguments */
          break;
 
        default:
          FT_ERROR(( "t1_decoder_parse_charstrings:"
                     " unhandled opcode %d\n", op ));
          goto Syntax_Error;
        }
 
        /* XXX Operators usually clear the operand stack;  */
        /*     only div, callsubr, callothersubr, pop, and */
        /*     return are different.                       */
        /*     In practice it doesn't matter (?).          */
 
        decoder->top = top;
 
#ifdef FT_DEBUG_LEVEL_TRACE
        FT_TRACE4(( "\n" ));
        bol = TRUE;
#endif
 
      } /* general operator processing */
 
    } /* while ip < limit */
 
    FT_TRACE4(( "..end..\n\n" ));
 
  Fail:
    return error;
 
  Syntax_Error:
    return FT_THROW( Syntax_Error );
 
  Stack_Underflow:
    return FT_THROW( Stack_Underflow );
  }
 
 
#else /* !T1_CONFIG_OPTION_OLD_ENGINE */
 
 
  /**************************************************************************
   *
   * @Function:
   *   t1_decoder_parse_metrics
   *
   * @Description:
   *   Parses a given Type 1 charstrings program to extract width
   *
   * @Input:
   *   decoder ::
   *     The current Type 1 decoder.
   *
   *   charstring_base ::
   *     The base address of the charstring stream.
   *
   *   charstring_len ::
   *     The length in bytes of the charstring stream.
   *
   * @Return:
   *   FreeType error code.  0 means success.
   */
  FT_LOCAL_DEF( FT_Error )
  t1_decoder_parse_metrics( T1_Decoder  decoder,
                            FT_Byte*    charstring_base,
                            FT_UInt     charstring_len )
  {
    T1_Decoder_Zone  zone;
    FT_Byte*         ip;
    FT_Byte*         limit;
    T1_Builder       builder = &decoder->builder;
    FT_Bool          large_int;
 
#ifdef FT_DEBUG_LEVEL_TRACE
    FT_Bool          bol = TRUE;
#endif
 
 
    /* First of all, initialize the decoder */
    decoder->top  = decoder->stack;
    decoder->zone = decoder->zones;
    zone          = decoder->zones;
 
    builder->parse_state = T1_Parse_Start;
 
    zone->base           = charstring_base;
    limit = zone->limit  = charstring_base + charstring_len;
    ip    = zone->cursor = zone->base;
 
    large_int = FALSE;
 
    /* now, execute loop */
    while ( ip < limit )
    {
      FT_Long*     top   = decoder->top;
      T1_Operator  op    = op_none;
      FT_Int32     value = 0;
 
 
#ifdef FT_DEBUG_LEVEL_TRACE
      if ( bol )
      {
        FT_TRACE5(( " (%ld)", decoder->top - decoder->stack ));
        bol = FALSE;
      }
#endif
 
      /**********************************************************************
       *
       * Decode operator or operand
       *
       */
 
      /* first of all, decompress operator or value */
      switch ( *ip++ )
      {
      case 1:
      case 3:
      case 4:
      case 5:
      case 6:
      case 7:
      case 8:
      case 9:
      case 14:
      case 15:
      case 21:
      case 22:
      case 30:
      case 31:
        goto No_Width;
 
      case 10:
        op = op_callsubr;
        break;
      case 11:
        op = op_return;
        break;
 
      case 13:
        op = op_hsbw;
        break;
 
      case 12:
        if ( ip >= limit )
        {
          FT_ERROR(( "t1_decoder_parse_metrics:"
                     " invalid escape (12+EOF)\n" ));
          goto Syntax_Error;
        }
 
        switch ( *ip++ )
        {
        case 7:
          op = op_sbw;
          break;
        case 12:
          op = op_div;
          break;
 
        default:
          goto No_Width;
        }
        break;
 
      case 255:    /* four bytes integer */
        if ( ip + 4 > limit )
        {
          FT_ERROR(( "t1_decoder_parse_metrics:"
                     " unexpected EOF in integer\n" ));
          goto Syntax_Error;
        }
 
        value = (FT_Int32)( ( (FT_UInt32)ip[0] << 24 ) |
                            ( (FT_UInt32)ip[1] << 16 ) |
                            ( (FT_UInt32)ip[2] << 8  ) |
                              (FT_UInt32)ip[3]         );
        ip += 4;
 
        /* According to the specification, values > 32000 or < -32000 must */
        /* be followed by a `div' operator to make the result be in the    */
        /* range [-32000;32000].  We expect that the second argument of    */
        /* `div' is not a large number.  Additionally, we don't handle     */
        /* stuff like `<large1> <large2> <num> div <num> div' or           */
        /* <large1> <large2> <num> div div'.  This is probably not allowed */
        /* anyway.                                                         */
        if ( value > 32000 || value < -32000 )
        {
          if ( large_int )
          {
            FT_ERROR(( "t1_decoder_parse_metrics:"
                       " no `div' after large integer\n" ));
            goto Syntax_Error;
          }
          else
            large_int = TRUE;
        }
        else
        {
          if ( !large_int )
            value = (FT_Int32)( (FT_UInt32)value << 16 );
        }
 
        break;
 
      default:
        if ( ip[-1] >= 32 )
        {
          if ( ip[-1] < 247 )
            value = (FT_Int32)ip[-1] - 139;
          else
          {
            if ( ++ip > limit )
            {
              FT_ERROR(( "t1_decoder_parse_metrics:"
                         " unexpected EOF in integer\n" ));
              goto Syntax_Error;
            }
 
            if ( ip[-2] < 251 )
              value =    ( ( ip[-2] - 247 ) * 256 ) + ip[-1] + 108;
            else
              value = -( ( ( ip[-2] - 251 ) * 256 ) + ip[-1] + 108 );
          }
 
          if ( !large_int )
            value = (FT_Int32)( (FT_UInt32)value << 16 );
        }
        else
        {
          FT_ERROR(( "t1_decoder_parse_metrics:"
                     " invalid byte (%d)\n", ip[-1] ));
          goto Syntax_Error;
        }
      }
 
      if ( large_int && !( op == op_none || op == op_div ) )
      {
        FT_ERROR(( "t1_decoder_parse_metrics:"
                   " no `div' after large integer\n" ));
        goto Syntax_Error;
      }
 
      /**********************************************************************
       *
       * Push value on stack, or process operator
       *
       */
      if ( op == op_none )
      {
        if ( top - decoder->stack >= T1_MAX_CHARSTRINGS_OPERANDS )
        {
          FT_ERROR(( "t1_decoder_parse_metrics: stack overflow\n" ));
          goto Syntax_Error;
        }
 
#ifdef FT_DEBUG_LEVEL_TRACE
        if ( large_int )
          FT_TRACE4(( " %d", value ));
        else
          FT_TRACE4(( " %d", value / 65536 ));
#endif
 
        *top++       = value;
        decoder->top = top;
      }
      else  /* general operator */
      {
        FT_Int  num_args = t1_args_count[op];
 
 
        FT_ASSERT( num_args >= 0 );
 
        if ( top - decoder->stack < num_args )
          goto Stack_Underflow;
 
#ifdef FT_DEBUG_LEVEL_TRACE
 
        switch ( op )
        {
        case op_callsubr:
        case op_div:
        case op_return:
          break;
 
        default:
          if ( top - decoder->stack != num_args )
            FT_TRACE0(( "t1_decoder_parse_metrics:"
                        " too much operands on the stack"
                        " (seen %ld, expected %d)\n",
                        top - decoder->stack, num_args ));
          break;
        }
 
#endif /* FT_DEBUG_LEVEL_TRACE */
 
        top -= num_args;
 
        switch ( op )
        {
        case op_hsbw:
          FT_TRACE4(( " hsbw" ));
 
          builder->parse_state = T1_Parse_Have_Width;
 
          builder->left_bearing.x = ADD_LONG( builder->left_bearing.x,
                                              top[0] );
 
          builder->advance.x = top[1];
          builder->advance.y = 0;
 
          /* we only want to compute the glyph's metrics */
          /* (lsb + advance width) without loading the   */
          /* rest of it; so exit immediately             */
          FT_TRACE4(( "\n" ));
          return FT_Err_Ok;
 
        case op_sbw:
          FT_TRACE4(( " sbw" ));
 
          builder->parse_state = T1_Parse_Have_Width;
 
          builder->left_bearing.x = ADD_LONG( builder->left_bearing.x,
                                              top[0] );
          builder->left_bearing.y = ADD_LONG( builder->left_bearing.y,
                                              top[1] );
 
          builder->advance.x = top[2];
          builder->advance.y = top[3];
 
          /* we only want to compute the glyph's metrics */
          /* (lsb + advance width), without loading the  */
          /* rest of it; so exit immediately             */
          FT_TRACE4(( "\n" ));
          return FT_Err_Ok;
 
        case op_div:
          FT_TRACE4(( " div" ));
 
          /* if `large_int' is set, we divide unscaled numbers; */
          /* otherwise, we divide numbers in 16.16 format --    */
          /* in both cases, it is the same operation            */
          *top = FT_DivFix( top[0], top[1] );
          top++;
 
          large_int = FALSE;
          break;
 
        case op_callsubr:
          {
            FT_Int  idx;
 
 
            FT_TRACE4(( " callsubr" ));
 
            idx = Fix2Int( top[0] );
 
            if ( decoder->subrs_hash )
            {
              size_t*  val = ft_hash_num_lookup( idx,
                                                 decoder->subrs_hash );
 
 
              if ( val )
                idx = *val;
              else
                idx = -1;
            }
 
            if ( idx < 0 || idx >= decoder->num_subrs )
            {
              FT_ERROR(( "t1_decoder_parse_metrics:"
                         " invalid subrs index\n" ));
              goto Syntax_Error;
            }
 
            if ( zone - decoder->zones >= T1_MAX_SUBRS_CALLS )
            {
              FT_ERROR(( "t1_decoder_parse_metrics:"
                         " too many nested subrs\n" ));
              goto Syntax_Error;
            }
 
            zone->cursor = ip;  /* save current instruction pointer */
 
            zone++;
 
            /* The Type 1 driver stores subroutines without the seed bytes. */
            /* The CID driver stores subroutines with seed bytes.  This     */
            /* case is taken care of when decoder->subrs_len == 0.          */
            zone->base = decoder->subrs[idx];
 
            if ( decoder->subrs_len )
              zone->limit = zone->base + decoder->subrs_len[idx];
            else
            {
              /* We are using subroutines from a CID font.  We must adjust */
              /* for the seed bytes.                                       */
              zone->base  += ( decoder->lenIV >= 0 ? decoder->lenIV : 0 );
              zone->limit  = decoder->subrs[idx + 1];
            }
 
            zone->cursor = zone->base;
 
            if ( !zone->base )
            {
              FT_ERROR(( "t1_decoder_parse_metrics:"
                         " invoking empty subrs\n" ));
              goto Syntax_Error;
            }
 
            decoder->zone = zone;
            ip            = zone->base;
            limit         = zone->limit;
            break;
          }
 
        case op_return:
          FT_TRACE4(( " return" ));
 
          if ( zone <= decoder->zones )
          {
            FT_ERROR(( "t1_decoder_parse_metrics:"
                       " unexpected return\n" ));
            goto Syntax_Error;
          }
 
          zone--;
          ip            = zone->cursor;
          limit         = zone->limit;
          decoder->zone = zone;
          break;
 
        default:
          FT_ERROR(( "t1_decoder_parse_metrics:"
                     " unhandled opcode %d\n", op ));
          goto Syntax_Error;
        }
 
        decoder->top = top;
 
      } /* general operator processing */
 
    } /* while ip < limit */
 
    FT_TRACE4(( "..end..\n\n" ));
 
  No_Width:
    FT_ERROR(( "t1_decoder_parse_metrics:"
               " no width, found op %d instead\n",
               ip[-1] ));
  Syntax_Error:
    return FT_THROW( Syntax_Error );
 
  Stack_Underflow:
    return FT_THROW( Stack_Underflow );
  }
 
#endif /* !T1_CONFIG_OPTION_OLD_ENGINE */
 
 
  /* initialize T1 decoder */
  FT_LOCAL_DEF( FT_Error )
  t1_decoder_init( T1_Decoder           decoder,
                   FT_Face              face,
                   FT_Size              size,
                   FT_GlyphSlot         slot,
                   FT_Byte**            glyph_names,
                   PS_Blend             blend,
                   FT_Bool              hinting,
                   FT_Render_Mode       hint_mode,
                   T1_Decoder_Callback  parse_callback )
  {
    FT_ZERO( decoder );
 
    /* retrieve `psnames' interface from list of current modules */
    {
      FT_Service_PsCMaps  psnames;
 
 
      FT_FACE_FIND_GLOBAL_SERVICE( face, psnames, POSTSCRIPT_CMAPS );
      if ( !psnames )
      {
        FT_ERROR(( "t1_decoder_init:"
                   " the `psnames' module is not available\n" ));
        return FT_THROW( Unimplemented_Feature );
      }
 
      decoder->psnames = psnames;
    }
 
    t1_builder_init( &decoder->builder, face, size, slot, hinting );
 
    /* decoder->buildchar and decoder->len_buildchar have to be  */
    /* initialized by the caller since we cannot know the length */
    /* of the BuildCharArray                                     */
 
    decoder->num_glyphs     = (FT_UInt)face->num_glyphs;
    decoder->glyph_names    = glyph_names;
    decoder->hint_mode      = hint_mode;
    decoder->blend          = blend;
    decoder->parse_callback = parse_callback;
 
    decoder->funcs          = t1_decoder_funcs;
 
    return FT_Err_Ok;
  }
 
 
  /* finalize T1 decoder */
  FT_LOCAL_DEF( void )
  t1_decoder_done( T1_Decoder  decoder )
  {
    FT_Memory  memory = decoder->builder.memory;
 
 
    t1_builder_done( &decoder->builder );
 
    if ( decoder->cf2_instance.finalizer )
    {
      decoder->cf2_instance.finalizer( decoder->cf2_instance.data );
      FT_FREE( decoder->cf2_instance.data );
    }
  }
 
 
/* END */