git.sesse.net Git - vlc/blob - modules/codec/subsdec.c

   1 /*****************************************************************************
   2  * subsdec.c : text subtitles decoder
   3  *****************************************************************************
   4  * Copyright (C) 2000-2006 the VideoLAN team
   5  * $Id$
   6  *
   7  * Authors: Gildas Bazin <gbazin@videolan.org>
   8  *          Samuel Hocevar <sam@zoy.org>
   9  *          Derk-Jan Hartman <hartman at videolan dot org>
  10  *          Bernie Purcell <bitmap@videolan.org>
  11  *
  12  * This program is free software; you can redistribute it and/or modify
  13  * it under the terms of the GNU General Public License as published by
  14  * the Free Software Foundation; either version 2 of the License, or
  15  * (at your option) any later version.
  16  *
  17  * This program is distributed in the hope that it will be useful,
  18  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  19  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  20  * GNU General Public License for more details.
  21  *
  22  * You should have received a copy of the GNU General Public License
  23  * along with this program; if not, write to the Free Software
  24  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston MA 02110-1301, USA.
  25  *****************************************************************************/
  26
  27 /*****************************************************************************
  28  * Preamble
  29  *****************************************************************************/
  30 #ifdef HAVE_CONFIG_H
  31 # include "config.h"
  32 #endif
  33
  34 #include "subsdec.h"
  35 #include <vlc_plugin.h>
  36
  37 /*****************************************************************************
  38  * Local prototypes
  39  *****************************************************************************/
  40 static int  OpenDecoder   ( vlc_object_t * );
  41 static void CloseDecoder  ( vlc_object_t * );
  42
  43 static subpicture_t   *DecodeBlock   ( decoder_t *, block_t ** );
  44 static subpicture_t   *ParseText     ( decoder_t *, block_t * );
  45 static char           *StripTags      ( char * );
  46 static char           *CreateHtmlSubtitle( int *pi_align, char * );
  47
  48
  49 /*****************************************************************************
  50  * Module descriptor.
  51  *****************************************************************************/
  52 static const char *const ppsz_encodings[] = {
  53     "",
  54     "system",
  55     "UTF-8",
  56     "UTF-16",
  57     "UTF-16BE",
  58     "UTF-16LE",
  59     "GB18030",
  60     "ISO-8859-15",
  61     "Windows-1252",
  62     "ISO-8859-2",
  63     "Windows-1250",
  64     "ISO-8859-3",
  65     "ISO-8859-10",
  66     "Windows-1251",
  67     "KOI8-R",
  68     "KOI8-U",
  69     "ISO-8859-6",
  70     "Windows-1256",
  71     "ISO-8859-7",
  72     "Windows-1253",
  73     "ISO-8859-8",
  74     "Windows-1255",
  75     "ISO-8859-9",
  76     "Windows-1254",
  77     "ISO-8859-11",
  78     "Windows-874",
  79     "ISO-8859-13",
  80     "Windows-1257",
  81     "ISO-8859-14",
  82     "ISO-8859-16",
  83     "ISO-2022-CN-EXT",
  84     "EUC-CN",
  85     "ISO-2022-JP-2",
  86     "EUC-JP",
  87     "Shift_JIS",
  88     "CP949",
  89     "ISO-2022-KR",
  90     "Big5",
  91     "ISO-2022-TW",
  92     "Big5-HKSCS",
  93     "VISCII",
  94     "Windows-1258",
  95 };
  96
  97 static const char *const ppsz_encoding_names[] = {
  98     /* xgettext:
  99       The character encoding name in parenthesis corresponds to that used for
 100       the GetACP translation. "Windows-1252" applies to Western European
 101       languages using the Latin alphabet. */
 102     N_("Default (Windows-1252)"),
 103     N_("System codeset"),
 104     N_("Universal (UTF-8)"),
 105     N_("Universal (UTF-16)"),
 106     N_("Universal (big endian UTF-16)"),
 107     N_("Universal (little endian UTF-16)"),
 108     N_("Universal, Chinese (GB18030)"),
 109
 110   /* ISO 8859 and the likes */
 111     /* 1 */
 112     N_("Western European (Latin-9)"), /* mostly superset of Latin-1 */
 113     N_("Western European (Windows-1252)"),
 114     /* 2 */
 115     N_("Eastern European (Latin-2)"),
 116     N_("Eastern European (Windows-1250)"),
 117     /* 3 */
 118     N_("Esperanto (Latin-3)"),
 119     /* 4 */
 120     N_("Nordic (Latin-6)"), /* Latin 6 supersedes Latin 4 */
 121     /* 5 */
 122     N_("Cyrillic (Windows-1251)"), /* ISO 8859-5 is not practically used */
 123     N_("Russian (KOI8-R)"),
 124     N_("Ukrainian (KOI8-U)"),
 125     /* 6 */
 126     N_("Arabic (ISO 8859-6)"),
 127     N_("Arabic (Windows-1256)"),
 128     /* 7 */
 129     N_("Greek (ISO 8859-7)"),
 130     N_("Greek (Windows-1253)"),
 131     /* 8 */
 132     N_("Hebrew (ISO 8859-8)"),
 133     N_("Hebrew (Windows-1255)"),
 134     /* 9 */
 135     N_("Turkish (ISO 8859-9)"),
 136     N_("Turkish (Windows-1254)"),
 137     /* 10 -> 4 */
 138     /* 11 */
 139     N_("Thai (TIS 620-2533/ISO 8859-11)"),
 140     N_("Thai (Windows-874)"),
 141     /* 13 */
 142     N_("Baltic (Latin-7)"),
 143     N_("Baltic (Windows-1257)"),
 144     /* 12 -> /dev/null */
 145     /* 14 */
 146     N_("Celtic (Latin-8)"),
 147     /* 15 -> 1 */
 148     /* 16 */
 149     N_("South-Eastern European (Latin-10)"),
 150   /* CJK families */
 151     N_("Simplified Chinese (ISO-2022-CN-EXT)"),
 152     N_("Simplified Chinese Unix (EUC-CN)"),
 153     N_("Japanese (7-bits JIS/ISO-2022-JP-2)"),
 154     N_("Japanese Unix (EUC-JP)"),
 155     N_("Japanese (Shift JIS)"),
 156     N_("Korean (EUC-KR/CP949)"),
 157     N_("Korean (ISO-2022-KR)"),
 158     N_("Traditional Chinese (Big5)"),
 159     N_("Traditional Chinese Unix (EUC-TW)"),
 160     N_("Hong-Kong Supplementary (HKSCS)"),
 161   /* Other */
 162     N_("Vietnamese (VISCII)"),
 163     N_("Vietnamese (Windows-1258)"),
 164 };
 165 /*
 166 SSA supports charset selection.
 167 The following known charsets are used:
 168
 169 0 = Ansi - Western European
 170 1 = default
 171 2 = symbol
 172 3 = invalid
 173 77 = Mac
 174 128 = Japanese (Shift JIS)
 175 129 = Hangul
 176 130 = Johab
 177 134 = GB2312 Simplified Chinese
 178 136 = Big5 Traditional Chinese
 179 161 = Greek
 180 162 = Turkish
 181 163 = Vietnamese
 182 177 = Hebrew
 183 178 = Arabic
 184 186 = Baltic
 185 204 = Russian (Cyrillic)
 186 222 = Thai
 187 238 = Eastern European
 188 254 = PC 437
 189 */
 190
 191 static const int  pi_justification[] = { 0, 1, 2 };
 192 static const char *const ppsz_justification_text[] = {
 193     N_("Center"),N_("Left"),N_("Right")};
 194
 195 #define ENCODING_TEXT N_("Subtitles text encoding")
 196 #define ENCODING_LONGTEXT N_("Set the encoding used in text subtitles")
 197 #define ALIGN_TEXT N_("Subtitles justification")
 198 #define ALIGN_LONGTEXT N_("Set the justification of subtitles")
 199 #define AUTODETECT_UTF8_TEXT N_("UTF-8 subtitles autodetection")
 200 #define AUTODETECT_UTF8_LONGTEXT N_("This enables automatic detection of " \
 201             "UTF-8 encoding within subtitles files.")
 202 #define FORMAT_TEXT N_("Formatted Subtitles")
 203 #define FORMAT_LONGTEXT N_("Some subtitle formats allow for text formatting. " \
 204  "VLC partly implements this, but you can choose to disable all formatting.")
 205
 206
 207 vlc_module_begin ()
 208     set_shortname( N_("Subtitles"))
 209     set_description( N_("Text subtitles decoder") )
 210     set_capability( "decoder", 50 )
 211     set_callbacks( OpenDecoder, CloseDecoder )
 212     set_category( CAT_INPUT )
 213     set_subcategory( SUBCAT_INPUT_SCODEC )
 214
 215     add_integer( "subsdec-align", 0, NULL, ALIGN_TEXT, ALIGN_LONGTEXT,
 216                  false )
 217         change_integer_list( pi_justification, ppsz_justification_text, NULL )
 218     add_string( "subsdec-encoding", "", NULL,
 219                 ENCODING_TEXT, ENCODING_LONGTEXT, false )
 220         change_string_list( ppsz_encodings, ppsz_encoding_names, 0 )
 221     add_bool( "subsdec-autodetect-utf8", true, NULL,
 222               AUTODETECT_UTF8_TEXT, AUTODETECT_UTF8_LONGTEXT, false )
 223     add_bool( "subsdec-formatted", true, NULL, FORMAT_TEXT, FORMAT_LONGTEXT,
 224                  false )
 225 vlc_module_end ()
 226
 227 /*****************************************************************************
 228  * OpenDecoder: probe the decoder and return score
 229  *****************************************************************************
 230  * Tries to launch a decoder and return score so that the interface is able
 231  * to chose.
 232  *****************************************************************************/
 233 static int OpenDecoder( vlc_object_t *p_this )
 234 {
 235     decoder_t     *p_dec = (decoder_t*)p_this;
 236     decoder_sys_t *p_sys;
 237
 238     switch( p_dec->fmt_in.i_codec )
 239     {
 240         case VLC_CODEC_SUBT:
 241         case VLC_CODEC_SSA:
 242         case VLC_CODEC_ITU_T140:
 243             break;
 244         default:
 245             return VLC_EGENERIC;
 246     }
 247
 248     p_dec->pf_decode_sub = DecodeBlock;
 249     p_dec->fmt_out.i_cat = SPU_ES;
 250     p_dec->fmt_out.i_codec = 0;
 251
 252     /* Allocate the memory needed to store the decoder's structure */
 253     p_dec->p_sys = p_sys = calloc( 1, sizeof( *p_sys ) );
 254     if( p_sys == NULL )
 255         return VLC_ENOMEM;
 256
 257     /* init of p_sys */
 258     p_sys->i_align = 0;
 259     p_sys->iconv_handle = (vlc_iconv_t)-1;
 260     p_sys->b_autodetect_utf8 = false;
 261     p_sys->b_ass = false;
 262     p_sys->i_original_height = -1;
 263     p_sys->i_original_width = -1;
 264     TAB_INIT( p_sys->i_ssa_styles, p_sys->pp_ssa_styles );
 265     TAB_INIT( p_sys->i_images, p_sys->pp_images );
 266
 267     char *psz_charset = NULL;
 268
 269     /* First try demux-specified encoding */
 270     if( p_dec->fmt_in.i_codec == VLC_CODEC_ITU_T140 )
 271         psz_charset = strdup( "UTF-8" ); /* IUT T.140 is always using UTF-8 */
 272     else
 273     if( p_dec->fmt_in.subs.psz_encoding && *p_dec->fmt_in.subs.psz_encoding )
 274     {
 275         psz_charset = strdup (p_dec->fmt_in.subs.psz_encoding);
 276         msg_Dbg (p_dec, "trying demuxer-specified character encoding: %s",
 277                  p_dec->fmt_in.subs.psz_encoding ?
 278                  p_dec->fmt_in.subs.psz_encoding : "not specified");
 279     }
 280
 281     /* Second, try configured encoding */
 282     if (psz_charset == NULL)
 283     {
 284         psz_charset = var_InheritString (p_dec, "subsdec-encoding");
 285         msg_Dbg (p_dec, "trying configured character encoding: %s",
 286                  psz_charset ? psz_charset : "not specified");
 287         if (psz_charset != NULL && !strcmp (psz_charset, "system"))
 288         {
 289             free (psz_charset);
 290             psz_charset = strdup ("");
 291             /* ^ iconv() treats "" as nl_langinfo(CODESET) */
 292         }
 293     }
 294
 295     /* Third, try "local" encoding with optional UTF-8 autodetection */
 296     if (psz_charset == NULL)
 297     {
 298         /* xgettext:
 299            The Windows ANSI code page most commonly used for this language.
 300            VLC uses this as a guess of the subtitle files character set
 301            (if UTF-8 and UTF-16 autodetection fails).
 302            Western European languages normally use "CP1252", which is a
 303            Microsoft-variant of ISO 8859-1. That suits the Latin alphabet.
 304            Other scripts use other code pages.
 305
 306            This MUST be a valid iconv character set. If unsure, please refer
 307            the VideoLAN translators mailing list. */
 308         const char *acp = vlc_pgettext("GetACP", "CP1252");
 309
 310         psz_charset = strdup (acp);
 311         msg_Dbg (p_dec, "trying default character encoding: %s",
 312                  psz_charset ? psz_charset : "not specified");
 313
 314         if (var_InheritBool (p_dec, "subsdec-autodetect-utf8"))
 315         {
 316             msg_Dbg (p_dec, "using automatic UTF-8 detection");
 317             p_sys->b_autodetect_utf8 = true;
 318         }
 319     }
 320
 321     /* Forth, don't do character decoding, i.e. assume UTF-8 */
 322     if (psz_charset == NULL)
 323     {
 324         psz_charset = strdup ("UTF-8");
 325         msg_Dbg (p_dec, "using UTF-8 character encoding" );
 326     }
 327
 328     if ((psz_charset != NULL)
 329      && strcasecmp (psz_charset, "UTF-8")
 330      && strcasecmp (psz_charset, "utf8"))
 331     {
 332         p_sys->iconv_handle = vlc_iconv_open ("UTF-8", psz_charset);
 333         if (p_sys->iconv_handle == (vlc_iconv_t)(-1))
 334             msg_Err (p_dec, "cannot convert from %s: %m", psz_charset);
 335     }
 336     free (psz_charset);
 337
 338     p_sys->i_align = var_InheritInteger( p_dec, "subsdec-align" );
 339
 340     if( p_dec->fmt_in.i_codec == VLC_CODEC_SSA
 341      && var_InheritBool( p_dec, "subsdec-formatted" ) )
 342     {
 343         if( p_dec->fmt_in.i_extra > 0 )
 344             ParseSSAHeader( p_dec );
 345     }
 346
 347     return VLC_SUCCESS;
 348 }
 349
 350 /****************************************************************************
 351  * DecodeBlock: the whole thing
 352  ****************************************************************************
 353  * This function must be fed with complete subtitles units.
 354  ****************************************************************************/
 355 static subpicture_t *DecodeBlock( decoder_t *p_dec, block_t **pp_block )
 356 {
 357     subpicture_t *p_spu;
 358     block_t *p_block;
 359
 360     if( !pp_block || *pp_block == NULL )
 361         return NULL;
 362
 363     p_block = *pp_block;
 364     if( p_block->i_flags & (BLOCK_FLAG_DISCONTINUITY|BLOCK_FLAG_CORRUPTED) )
 365     {
 366         block_Release( p_block );
 367         return NULL;
 368     }
 369
 370     p_spu = ParseText( p_dec, p_block );
 371
 372     block_Release( p_block );
 373     *pp_block = NULL;
 374
 375     return p_spu;
 376 }
 377
 378 /*****************************************************************************
 379  * CloseDecoder: clean up the decoder
 380  *****************************************************************************/
 381 static void CloseDecoder( vlc_object_t *p_this )
 382 {
 383     decoder_t *p_dec = (decoder_t *)p_this;
 384     decoder_sys_t *p_sys = p_dec->p_sys;
 385
 386     if( p_sys->iconv_handle != (vlc_iconv_t)-1 )
 387         vlc_iconv_close( p_sys->iconv_handle );
 388
 389     if( p_sys->pp_ssa_styles )
 390     {
 391         int i;
 392         for( i = 0; i < p_sys->i_ssa_styles; i++ )
 393         {
 394             if( !p_sys->pp_ssa_styles[i] )
 395                 continue;
 396
 397             free( p_sys->pp_ssa_styles[i]->psz_stylename );
 398             free( p_sys->pp_ssa_styles[i]->font_style.psz_fontname );
 399             free( p_sys->pp_ssa_styles[i] );
 400         }
 401         TAB_CLEAN( p_sys->i_ssa_styles, p_sys->pp_ssa_styles );
 402     }
 403     if( p_sys->pp_images )
 404     {
 405         int i;
 406         for( i = 0; i < p_sys->i_images; i++ )
 407         {
 408             if( !p_sys->pp_images[i] )
 409                 continue;
 410
 411             if( p_sys->pp_images[i]->p_pic )
 412                 picture_Release( p_sys->pp_images[i]->p_pic );
 413             free( p_sys->pp_images[i]->psz_filename );
 414
 415             free( p_sys->pp_images[i] );
 416         }
 417         TAB_CLEAN( p_sys->i_images, p_sys->pp_images );
 418     }
 419
 420     free( p_sys );
 421 }
 422
 423 /*****************************************************************************
 424  * ParseText: parse an text subtitle packet and send it to the video output
 425  *****************************************************************************/
 426 static subpicture_t *ParseText( decoder_t *p_dec, block_t *p_block )
 427 {
 428     decoder_sys_t *p_sys = p_dec->p_sys;
 429     subpicture_t *p_spu = NULL;
 430     char *psz_subtitle = NULL;
 431     video_format_t fmt;
 432
 433     /* We cannot display a subpicture with no date */
 434     if( p_block->i_pts <= VLC_TS_INVALID )
 435     {
 436         msg_Warn( p_dec, "subtitle without a date" );
 437         return NULL;
 438     }
 439
 440     /* Check validity of packet data */
 441     /* An "empty" line containing only \0 can be used to force
 442        and ephemer picture from the screen */
 443     if( p_block->i_buffer < 1 )
 444     {
 445         msg_Warn( p_dec, "no subtitle data" );
 446         return NULL;
 447     }
 448
 449     /* Should be resiliant against bad subtitles */
 450     psz_subtitle = malloc( p_block->i_buffer + 1 );
 451     if( psz_subtitle == NULL )
 452         return NULL;
 453     memcpy( psz_subtitle, p_block->p_buffer, p_block->i_buffer );
 454     psz_subtitle[p_block->i_buffer] = '\0';
 455
 456     if( p_sys->iconv_handle == (vlc_iconv_t)-1 )
 457     {
 458         if (EnsureUTF8( psz_subtitle ) == NULL)
 459         {
 460             msg_Err( p_dec, "failed to convert subtitle encoding.\n"
 461                      "Try manually setting a character-encoding "
 462                      "before you open the file." );
 463         }
 464     }
 465     else
 466     {
 467
 468         if( p_sys->b_autodetect_utf8 )
 469         {
 470             if( IsUTF8( psz_subtitle ) == NULL )
 471             {
 472                 msg_Dbg( p_dec, "invalid UTF-8 sequence: "
 473                          "disabling UTF-8 subtitles autodetection" );
 474                 p_sys->b_autodetect_utf8 = false;
 475             }
 476         }
 477
 478         if( !p_sys->b_autodetect_utf8 )
 479         {
 480             size_t inbytes_left = strlen( psz_subtitle );
 481             size_t outbytes_left = 6 * inbytes_left;
 482             char *psz_new_subtitle = xmalloc( outbytes_left + 1 );
 483             char *psz_convert_buffer_out = psz_new_subtitle;
 484             const char *psz_convert_buffer_in = psz_subtitle;
 485
 486             size_t ret = vlc_iconv( p_sys->iconv_handle,
 487                                     &psz_convert_buffer_in, &inbytes_left,
 488                                     &psz_convert_buffer_out, &outbytes_left );
 489
 490             *psz_convert_buffer_out++ = '\0';
 491             free( psz_subtitle );
 492
 493             if( ( ret == (size_t)(-1) ) || inbytes_left )
 494             {
 495                 free( psz_new_subtitle );
 496                 msg_Err( p_dec, "failed to convert subtitle encoding.\n"
 497                         "Try manually setting a character-encoding "
 498                                 "before you open the file." );
 499                 return NULL;
 500             }
 501
 502             psz_subtitle = realloc( psz_new_subtitle,
 503                                     psz_convert_buffer_out - psz_new_subtitle );
 504             if( !psz_subtitle )
 505                 psz_subtitle = psz_new_subtitle;
 506         }
 507     }
 508
 509     /* Create the subpicture unit */
 510     p_spu = decoder_NewSubpicture( p_dec, NULL );
 511     if( !p_spu )
 512     {
 513         msg_Warn( p_dec, "can't get spu buffer" );
 514         free( psz_subtitle );
 515         return NULL;
 516     }
 517
 518     /* Create a new subpicture region */
 519     memset( &fmt, 0, sizeof(video_format_t) );
 520     fmt.i_chroma = VLC_CODEC_TEXT;
 521     fmt.i_width = fmt.i_height = 0;
 522     fmt.i_x_offset = fmt.i_y_offset = 0;
 523     p_spu->p_region = subpicture_region_New( &fmt );
 524     if( !p_spu->p_region )
 525     {
 526         msg_Err( p_dec, "cannot allocate SPU region" );
 527         free( psz_subtitle );
 528         decoder_DeleteSubpicture( p_dec, p_spu );
 529         return NULL;
 530     }
 531
 532     /* Decode and format the subpicture unit */
 533     if( p_dec->fmt_in.i_codec != VLC_CODEC_SSA )
 534     {
 535         /* Normal text subs, easy markup */
 536         p_spu->p_region->i_align = SUBPICTURE_ALIGN_BOTTOM | p_sys->i_align;
 537         p_spu->p_region->i_x = p_sys->i_align ? 20 : 0;
 538         p_spu->p_region->i_y = 10;
 539
 540         /* Remove formatting from string */
 541
 542         p_spu->p_region->psz_text = StripTags( psz_subtitle );
 543         if( var_InheritBool( p_dec, "subsdec-formatted" ) )
 544         {
 545             p_spu->p_region->psz_html = CreateHtmlSubtitle( &p_spu->p_region->i_align, psz_subtitle );
 546         }
 547
 548         p_spu->i_start = p_block->i_pts;
 549         p_spu->i_stop = p_block->i_pts + p_block->i_length;
 550         p_spu->b_ephemer = (p_block->i_length == 0);
 551         p_spu->b_absolute = false;
 552     }
 553     else
 554     {
 555         /* Decode SSA/USF strings */
 556         ParseSSAString( p_dec, psz_subtitle, p_spu );
 557
 558         p_spu->i_start = p_block->i_pts;
 559         p_spu->i_stop = p_block->i_pts + p_block->i_length;
 560         p_spu->b_ephemer = (p_block->i_length == 0);
 561         p_spu->b_absolute = false;
 562         p_spu->i_original_picture_width = p_sys->i_original_width;
 563         p_spu->i_original_picture_height = p_sys->i_original_height;
 564     }
 565     free( psz_subtitle );
 566
 567     return p_spu;
 568 }
 569
 570 char* GotoNextLine( char *psz_text )
 571 {
 572     char *p_newline = psz_text;
 573
 574     while( p_newline[0] != '\0' )
 575     {
 576         if( p_newline[0] == '\n' || p_newline[0] == '\r' )
 577         {
 578             p_newline++;
 579             while( p_newline[0] == '\n' || p_newline[0] == '\r' )
 580                 p_newline++;
 581             break;
 582         }
 583         else p_newline++;
 584     }
 585     return p_newline;
 586 }
 587
 588 /* Function now handles tags with attribute values, and tries
 589  * to deal with &' commands too. It no longer modifies the string
 590  * in place, so that the original text can be reused
 591  */
 592 static char *StripTags( char *psz_subtitle )
 593 {
 594     char *psz_text_start;
 595     char *psz_text;
 596
 597     psz_text = psz_text_start = malloc( strlen( psz_subtitle ) + 1 );
 598     if( !psz_text_start )
 599         return NULL;
 600
 601     while( *psz_subtitle )
 602     {
 603         if( *psz_subtitle == '<' )
 604         {
 605             if( strncasecmp( psz_subtitle, "<br/>", 5 ) == 0 )
 606                 *psz_text++ = '\n';
 607
 608             psz_subtitle += strcspn( psz_subtitle, ">" );
 609         }
 610         else if( *psz_subtitle == '&' )
 611         {
 612             if( !strncasecmp( psz_subtitle, "&lt;", 4 ))
 613             {
 614                 *psz_text++ = '<';
 615                 psz_subtitle += strcspn( psz_subtitle, ";" );
 616             }
 617             else if( !strncasecmp( psz_subtitle, "&gt;", 4 ))
 618             {
 619                 *psz_text++ = '>';
 620                 psz_subtitle += strcspn( psz_subtitle, ";" );
 621             }
 622             else if( !strncasecmp( psz_subtitle, "&amp;", 5 ))
 623             {
 624                 *psz_text++ = '&';
 625                 psz_subtitle += strcspn( psz_subtitle, ";" );
 626             }
 627             else if( !strncasecmp( psz_subtitle, "&quot;", 6 ))
 628             {
 629                 *psz_text++ = '\"';
 630                 psz_subtitle += strcspn( psz_subtitle, ";" );
 631             }
 632             else
 633             {
 634                 /* Assume it is just a normal ampersand */
 635                 *psz_text++ = '&';
 636             }
 637         }
 638         else
 639         {
 640             *psz_text++ = *psz_subtitle;
 641         }
 642
 643         psz_subtitle++;
 644     }
 645     *psz_text = '\0';
 646     char *psz = realloc( psz_text_start, strlen( psz_text_start ) + 1 );
 647     if( psz ) psz_text_start = psz;
 648
 649     return psz_text_start;
 650 }
 651
 652 /* Try to respect any style tags present in the subtitle string. The main
 653  * problem here is a lack of adequate specs for the subtitle formats.
 654  * SSA/ASS and USF are both detail spec'ed -- but they are handled elsewhere.
 655  * SAMI has a detailed spec, but extensive rework is needed in the demux
 656  * code to prevent all this style information being excised, as it presently
 657  * does.
 658  * That leaves the others - none of which were (I guess) originally intended
 659  * to be carrying style information. Over time people have used them that way.
 660  * In the absence of specifications from which to work, the tags supported
 661  * have been restricted to the simple set permitted by the USF DTD, ie. :
 662  *  Basic: <br>, <i>, <b>, <u>, <s>
 663  *  Extended: <font>
 664  *    Attributes: face
 665  *                family
 666  *                size
 667  *                color
 668  *                outline-color
 669  *                shadow-color
 670  *                outline-level
 671  *                shadow-level
 672  *                back-color
 673  *                alpha
 674  * There is also the further restriction that the subtitle be well-formed
 675  * as an XML entity, ie. the HTML sentence:
 676  *        <b><i>Bold and Italics</b></i>
 677  * doesn't qualify because the tags aren't nested one inside the other.
 678  * <text> tags are automatically added to the output to ensure
 679  * well-formedness.
 680  * If the text doesn't qualify for any reason, a NULL string is
 681  * returned, and the rendering engine will fall back to the
 682  * plain text version of the subtitle.
 683  */
 684 static void HtmlNPut( char **ppsz_html, const char *psz_text, int i_max )
 685 {
 686     const int i_len = strlen(psz_text);
 687
 688     strncpy( *ppsz_html, psz_text, i_max );
 689     *ppsz_html += __MIN(i_max,i_len);
 690 }
 691
 692 static void HtmlPut( char **ppsz_html, const char *psz_text )
 693 {
 694     strcpy( *ppsz_html, psz_text );
 695     *ppsz_html += strlen(psz_text);
 696 }
 697 static void HtmlCopy( char **ppsz_html, char **ppsz_subtitle, const char *psz_text )
 698 {
 699     HtmlPut( ppsz_html, psz_text );
 700     *ppsz_subtitle += strlen(psz_text);
 701 }
 702
 703 static char *CreateHtmlSubtitle( int *pi_align, char *psz_subtitle )
 704 {
 705     /* */
 706     char *psz_tag = malloc( ( strlen( psz_subtitle ) / 3 ) + 1 );
 707     if( !psz_tag )
 708         return NULL;
 709     psz_tag[ 0 ] = '\0';
 710
 711     /* */
 712     //Oo + 100 ???
 713     size_t i_buf_size = strlen( psz_subtitle ) + 100;
 714     char   *psz_html_start = malloc( i_buf_size );
 715     char   *psz_html = psz_html_start;
 716     if( psz_html_start == NULL )
 717     {
 718         free( psz_tag );
 719         return NULL;
 720     }
 721     psz_html[0] = '\0';
 722
 723     bool b_has_align = false;
 724
 725     HtmlPut( &psz_html, "<text>" );
 726
 727     /* */
 728     while( *psz_subtitle )
 729     {
 730         if( *psz_subtitle == '\n' )
 731         {
 732             HtmlPut( &psz_html, "<br/>" );
 733             psz_subtitle++;
 734         }
 735         else if( *psz_subtitle == '<' )
 736         {
 737             if( !strncasecmp( psz_subtitle, "<br/>", 5 ))
 738             {
 739                 HtmlCopy( &psz_html, &psz_subtitle, "<br/>" );
 740             }
 741             else if( !strncasecmp( psz_subtitle, "<b>", 3 ) )
 742             {
 743                 HtmlCopy( &psz_html, &psz_subtitle, "<b>" );
 744                 strcat( psz_tag, "b" );
 745             }
 746             else if( !strncasecmp( psz_subtitle, "<i>", 3 ) )
 747             {
 748                 HtmlCopy( &psz_html, &psz_subtitle, "<i>" );
 749                 strcat( psz_tag, "i" );
 750             }
 751             else if( !strncasecmp( psz_subtitle, "<u>", 3 ) )
 752             {
 753                 HtmlCopy( &psz_html, &psz_subtitle, "<u>" );
 754                 strcat( psz_tag, "u" );
 755             }
 756             else if( !strncasecmp( psz_subtitle, "<s>", 3 ) )
 757             {
 758                 HtmlCopy( &psz_html, &psz_subtitle, "<s>" );
 759                 strcat( psz_tag, "s" );
 760             }
 761             else if( !strncasecmp( psz_subtitle, "<font ", 6 ))
 762             {
 763                 const char *psz_attribs[] = { "face=", "family=", "size=",
 764                         "color=", "outline-color=", "shadow-color=",
 765                         "outline-level=", "shadow-level=", "back-color=",
 766                         "alpha=", NULL };
 767
 768                 HtmlCopy( &psz_html, &psz_subtitle, "<font " );
 769                 strcat( psz_tag, "f" );
 770
 771                 while( *psz_subtitle != '>' )
 772                 {
 773                     int  k;
 774
 775                     for( k=0; psz_attribs[ k ]; k++ )
 776                     {
 777                         int i_len = strlen( psz_attribs[ k ] );
 778
 779                         if( !strncasecmp( psz_subtitle, psz_attribs[k], i_len ) )
 780                         {
 781                             /* */
 782                             HtmlPut( &psz_html, psz_attribs[k] );
 783                             psz_subtitle += i_len;
 784
 785                             /* */
 786                             if( *psz_subtitle == '"' )
 787                             {
 788                                 psz_subtitle++;
 789                                 i_len = strcspn( psz_subtitle, "\"" );
 790                             }
 791                             else
 792                             {
 793                                 i_len = strcspn( psz_subtitle, " \t>" );
 794                             }
 795                             HtmlPut( &psz_html, "\"" );
 796                             if( !strcmp( psz_attribs[ k ], "color=" ) && *psz_subtitle >= '0' && *psz_subtitle <= '9' )
 797                                 HtmlPut( &psz_html, "#" );
 798                             HtmlNPut( &psz_html, psz_subtitle, i_len );
 799                             HtmlPut( &psz_html, "\"" );
 800
 801                             psz_subtitle += i_len;
 802                             if( *psz_subtitle == '\"' )
 803                                 psz_subtitle++;
 804                             break;
 805                         }
 806                     }
 807                     if( psz_attribs[ k ] == NULL )
 808                     {
 809                         /* Jump over unrecognised tag */
 810                         int i_len = strcspn( psz_subtitle, "\"" );
 811                         if( psz_subtitle[i_len] == '\"' )
 812                         {
 813                             i_len += 1 + strcspn( &psz_subtitle[i_len + 1], "\"" );
 814                             if( psz_subtitle[i_len] == '\"' )
 815                                 i_len++;
 816                         }
 817                         psz_subtitle += i_len;
 818                     }
 819                     while (*psz_subtitle == ' ')
 820                         *psz_html++ = *psz_subtitle++;
 821                 }
 822                 *psz_html++ = *psz_subtitle++;
 823             }
 824             else if( !strncmp( psz_subtitle, "</", 2 ))
 825             {
 826                 bool   b_match     = false;
 827                 bool   b_ignore    = false;
 828                 int    i_len       = strlen( psz_tag ) - 1;
 829                 char  *psz_lastTag = NULL;
 830
 831                 if( i_len >= 0 )
 832                 {
 833                     psz_lastTag = psz_tag + i_len;
 834                     i_len = 0;
 835
 836                     switch( *psz_lastTag )
 837                     {
 838                     case 'b':
 839                         b_match = !strncasecmp( psz_subtitle, "</b>", 4 );
 840                         i_len   = 4;
 841                         break;
 842                     case 'i':
 843                         b_match = !strncasecmp( psz_subtitle, "</i>", 4 );
 844                         i_len   = 4;
 845                         break;
 846                     case 'u':
 847                         b_match = !strncasecmp( psz_subtitle, "</u>", 4 );
 848                         i_len   = 4;
 849                         break;
 850                     case 's':
 851                         b_match = !strncasecmp( psz_subtitle, "</s>", 4 );
 852                         i_len   = 4;
 853                         break;
 854                     case 'f':
 855                         b_match = !strncasecmp( psz_subtitle, "</font>", 7 );
 856                         i_len   = 7;
 857                         break;
 858                     case 'I':
 859                         i_len = strcspn( psz_subtitle, ">" );
 860                         b_match = psz_subtitle[i_len] == '>';
 861                         b_ignore = true;
 862                         if( b_match )
 863                             i_len++;
 864                         break;
 865                     }
 866                 }
 867                 if( !b_match )
 868                 {
 869                     /* Not well formed -- kill everything */
 870                     free( psz_html_start );
 871                     psz_html_start = NULL;
 872                     break;
 873                 }
 874                 *psz_lastTag = '\0';
 875                 if( !b_ignore )
 876                     HtmlNPut( &psz_html, psz_subtitle, i_len );
 877
 878                 psz_subtitle += i_len;
 879             }
 880             else if( ( psz_subtitle[1] < 'a' || psz_subtitle[1] > 'z' ) &&
 881                      ( psz_subtitle[1] < 'A' || psz_subtitle[1] > 'Z' ) )
 882             {
 883                 /* We have a single < */
 884                 HtmlPut( &psz_html, "&lt;" );
 885                 psz_subtitle++;
 886             }
 887             else
 888             {
 889                 /* We have an unknown tag or a single < */
 890
 891                 /* Search for the next tag or end of tag or end of string */
 892                 char *psz_stop = psz_subtitle + 1 + strcspn( &psz_subtitle[1], "<>" );
 893                 char *psz_closing = strstr( psz_subtitle, "/>" );
 894
 895                 if( psz_closing && psz_closing < psz_stop )
 896                 {
 897                     /* We have a self closed tag, remove it */
 898                     psz_subtitle = &psz_closing[2];
 899                 }
 900                 else if( *psz_stop == '>' )
 901                 {
 902                     char psz_match[256];
 903
 904                     snprintf( psz_match, sizeof(psz_match), "</%s", &psz_subtitle[1] );
 905                     psz_match[strcspn( psz_match, " \t>" )] = '\0';
 906
 907                     if( strstr( psz_subtitle, psz_match ) )
 908                     {
 909                         /* We have the closing tag, ignore it TODO */
 910                         psz_subtitle = &psz_stop[1];
 911                         strcat( psz_tag, "I" );
 912                     }
 913                     else
 914                     {
 915                         int i_len = psz_stop + 1 - psz_subtitle;
 916
 917                         /* Copy the whole data */
 918                         for( ; i_len > 0; i_len--, psz_subtitle++ )
 919                         {
 920                             if( *psz_subtitle == '<' )
 921                                 HtmlPut( &psz_html, "&lt;" );
 922                             else if( *psz_subtitle == '>' )
 923                                 HtmlPut( &psz_html, "&gt;" );
 924                             else
 925                                 *psz_html++ = *psz_subtitle;
 926                         }
 927                     }
 928                 }
 929                 else
 930                 {
 931                     /* We have a single < */
 932                     HtmlPut( &psz_html, "&lt;" );
 933                     psz_subtitle++;
 934                 }
 935             }
 936         }
 937         else if( *psz_subtitle == '&' )
 938         {
 939             if( !strncasecmp( psz_subtitle, "&lt;", 4 ))
 940             {
 941                 HtmlCopy( &psz_html, &psz_subtitle, "&lt;" );
 942             }
 943             else if( !strncasecmp( psz_subtitle, "&gt;", 4 ))
 944             {
 945                 HtmlCopy( &psz_html, &psz_subtitle, "&gt;" );
 946             }
 947             else if( !strncasecmp( psz_subtitle, "&amp;", 5 ))
 948             {
 949                 HtmlCopy( &psz_html, &psz_subtitle, "&amp;" );
 950             }
 951             else
 952             {
 953                 HtmlPut( &psz_html, "&amp;" );
 954                 psz_subtitle++;
 955             }
 956         }
 957         else if( *psz_subtitle == '>' )
 958         {
 959             HtmlPut( &psz_html, "&gt;" );
 960             psz_subtitle++;
 961         }
 962         else if( psz_subtitle[0] == '{' && psz_subtitle[1] == '\\' &&
 963                  strchr( psz_subtitle, '}' ) )
 964         {
 965             /* Check for forced alignment */
 966             if( !b_has_align &&
 967                 !strncmp( psz_subtitle, "{\\an", 4 ) && psz_subtitle[4] >= '1' && psz_subtitle[4] <= '9' && psz_subtitle[5] == '}' )
 968             {
 969                 static const int pi_vertical[3] = { SUBPICTURE_ALIGN_BOTTOM, 0, SUBPICTURE_ALIGN_TOP };
 970                 static const int pi_horizontal[3] = { SUBPICTURE_ALIGN_LEFT, 0, SUBPICTURE_ALIGN_RIGHT };
 971                 const int i_id = psz_subtitle[4] - '1';
 972
 973                 b_has_align = true;
 974                 *pi_align = pi_vertical[i_id/3] | pi_horizontal[i_id%3];
 975             }
 976             /* TODO fr -> rotation */
 977
 978             /* Hide {\stupidity} */
 979             psz_subtitle = strchr( psz_subtitle, '}' ) + 1;
 980         }
 981         else if( psz_subtitle[0] == '{' && psz_subtitle[1] == 'Y'
 982                 && psz_subtitle[2] == ':' && strchr( psz_subtitle, '}' ) )
 983         {
 984             /* Hide {Y:stupidity} */
 985             psz_subtitle = strchr( psz_subtitle, '}' ) + 1;
 986         }
 987         else if( psz_subtitle[0] == '\\' && psz_subtitle[1] )
 988         {
 989             if( psz_subtitle[1] == 'N' || psz_subtitle[1] == 'n' )
 990             {
 991                 HtmlPut( &psz_html, "<br/>" );
 992                 psz_subtitle += 2;
 993             }
 994             else if( psz_subtitle[1] == 'h' )
 995             {
 996                 /* Non breakable space */
 997                 HtmlPut( &psz_html, NO_BREAKING_SPACE );
 998                 psz_subtitle += 2;
 999             }
1000             else
1001             {
1002                 HtmlPut( &psz_html, "\\" );
1003                 psz_subtitle++;
1004             }
1005         }
1006         else
1007         {
1008             *psz_html = *psz_subtitle;
1009             if( psz_html > psz_html_start )
1010             {
1011                 /* Check for double whitespace */
1012                 if( ( *psz_html == ' '  || *psz_html == '\t' ) &&
1013                     ( *(psz_html-1) == ' ' || *(psz_html-1) == '\t' ) )
1014                 {
1015                     HtmlPut( &psz_html, NO_BREAKING_SPACE );
1016                     psz_html--;
1017                 }
1018             }
1019             psz_html++;
1020             psz_subtitle++;
1021         }
1022
1023         if( ( size_t )( psz_html - psz_html_start ) > i_buf_size - 50 )
1024         {
1025             const int i_len = psz_html - psz_html_start;
1026
1027             i_buf_size += 200;
1028             char *psz_new = realloc( psz_html_start, i_buf_size );
1029             if( !psz_new )
1030                 break;
1031             psz_html_start = psz_new;
1032             psz_html = &psz_new[i_len];
1033         }
1034     }
1035     if( psz_html_start )
1036     {
1037         static const char *psz_text_close = "</text>";
1038         static const char *psz_tag_long = "/font>";
1039
1040         /* Realloc for closing tags and shrink memory */
1041         const size_t i_length = (size_t)( psz_html - psz_html_start );
1042
1043         const size_t i_size = i_length + strlen(psz_tag_long) * strlen(psz_tag) + strlen(psz_text_close) + 1;
1044         char *psz_new = realloc( psz_html_start, i_size );
1045         if( psz_new )
1046         {
1047             psz_html_start = psz_new;
1048             psz_html = &psz_new[i_length];
1049
1050             /* Close not well formed subtitle */
1051             while( *psz_tag )
1052             {
1053                 /* */
1054                 char *psz_last = &psz_tag[strlen(psz_tag)-1];
1055                 switch( *psz_last )
1056                 {
1057                 case 'b':
1058                     HtmlPut( &psz_html, "</b>" );
1059                     break;
1060                 case 'i':
1061                     HtmlPut( &psz_html, "</i>" );
1062                     break;
1063                 case 'u':
1064                     HtmlPut( &psz_html, "</u>" );
1065                     break;
1066                 case 's':
1067                     HtmlPut( &psz_html, "</s>" );
1068                     break;
1069                 case 'f':
1070                     HtmlPut( &psz_html, "/font>" );
1071                     break;
1072                 case 'I':
1073                     break;
1074                 }
1075
1076                 *psz_last = '\0';
1077             }
1078             HtmlPut( &psz_html, psz_text_close );
1079         }
1080     }
1081     free( psz_tag );
1082
1083     return psz_html_start;
1084 }
1085