1211 lines
36 KiB
C
1211 lines
36 KiB
C
/*
|
|
* Methods for dealing with opentype font tables
|
|
*
|
|
* Copyright 2014 Aric Stewart for CodeWeavers
|
|
*
|
|
* This library is free software; you can redistribute it and/or
|
|
* modify it under the terms of the GNU Lesser General Public
|
|
* License as published by the Free Software Foundation; either
|
|
* version 2.1 of the License, or (at your option) any later version.
|
|
*
|
|
* This library is distributed in the hope that it will be useful,
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
|
* Lesser General Public License for more details.
|
|
*
|
|
* You should have received a copy of the GNU Lesser General Public
|
|
* License along with this library; if not, write to the Free Software
|
|
* Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
|
|
*/
|
|
|
|
#define COBJMACROS
|
|
|
|
#include "dwrite_private.h"
|
|
|
|
WINE_DEFAULT_DEBUG_CHANNEL(dwrite);
|
|
|
|
#define MS_HEAD_TAG DWRITE_MAKE_OPENTYPE_TAG('h','e','a','d')
|
|
#define MS_HHEA_TAG DWRITE_MAKE_OPENTYPE_TAG('h','h','e','a')
|
|
#define MS_OTTO_TAG DWRITE_MAKE_OPENTYPE_TAG('O','T','T','O')
|
|
#define MS_OS2_TAG DWRITE_MAKE_OPENTYPE_TAG('O','S','/','2')
|
|
#define MS_POST_TAG DWRITE_MAKE_OPENTYPE_TAG('p','o','s','t')
|
|
#define MS_TTCF_TAG DWRITE_MAKE_OPENTYPE_TAG('t','t','c','f')
|
|
|
|
#ifdef WORDS_BIGENDIAN
|
|
#define GET_BE_WORD(x) (x)
|
|
#define GET_BE_DWORD(x) (x)
|
|
#else
|
|
#define GET_BE_WORD(x) MAKEWORD(HIBYTE(x), LOBYTE(x))
|
|
#define GET_BE_DWORD(x) MAKELONG(GET_BE_WORD(HIWORD(x)), GET_BE_WORD(LOWORD(x)))
|
|
#endif
|
|
|
|
typedef struct {
|
|
CHAR TTCTag[4];
|
|
DWORD Version;
|
|
DWORD numFonts;
|
|
DWORD OffsetTable[1];
|
|
} TTC_Header_V1;
|
|
|
|
typedef struct {
|
|
DWORD version;
|
|
WORD numTables;
|
|
WORD searchRange;
|
|
WORD entrySelector;
|
|
WORD rangeShift;
|
|
} TTC_SFNT_V1;
|
|
|
|
typedef struct {
|
|
CHAR tag[4];
|
|
DWORD checkSum;
|
|
DWORD offset;
|
|
DWORD length;
|
|
} TT_TableRecord;
|
|
|
|
typedef struct {
|
|
WORD platformID;
|
|
WORD encodingID;
|
|
DWORD offset;
|
|
} CMAP_EncodingRecord;
|
|
|
|
typedef struct {
|
|
WORD version;
|
|
WORD numTables;
|
|
CMAP_EncodingRecord tables[1];
|
|
} CMAP_Header;
|
|
|
|
typedef struct {
|
|
DWORD startCharCode;
|
|
DWORD endCharCode;
|
|
DWORD startGlyphID;
|
|
} CMAP_SegmentedCoverage_group;
|
|
|
|
typedef struct {
|
|
WORD format;
|
|
WORD reserved;
|
|
DWORD length;
|
|
DWORD language;
|
|
DWORD nGroups;
|
|
CMAP_SegmentedCoverage_group groups[1];
|
|
} CMAP_SegmentedCoverage;
|
|
|
|
typedef struct {
|
|
WORD format;
|
|
WORD length;
|
|
WORD language;
|
|
WORD segCountX2;
|
|
WORD searchRange;
|
|
WORD entrySelector;
|
|
WORD rangeShift;
|
|
WORD endCode[1];
|
|
} CMAP_SegmentMapping_0;
|
|
|
|
enum OPENTYPE_CMAP_TABLE_FORMAT
|
|
{
|
|
OPENTYPE_CMAP_TABLE_SEGMENT_MAPPING = 4,
|
|
OPENTYPE_CMAP_TABLE_SEGMENTED_COVERAGE = 12
|
|
};
|
|
|
|
/* PANOSE is 10 bytes in size, need to pack the structure properly */
|
|
#include "pshpack2.h"
|
|
typedef struct
|
|
{
|
|
ULONG version;
|
|
ULONG revision;
|
|
ULONG checksumadj;
|
|
ULONG magic;
|
|
USHORT flags;
|
|
USHORT unitsPerEm;
|
|
ULONGLONG created;
|
|
ULONGLONG modified;
|
|
SHORT xMin;
|
|
SHORT yMin;
|
|
SHORT xMax;
|
|
SHORT yMax;
|
|
USHORT macStyle;
|
|
USHORT lowestRecPPEM;
|
|
SHORT direction_hint;
|
|
SHORT index_format;
|
|
SHORT glyphdata_format;
|
|
} TT_HEAD;
|
|
|
|
typedef struct
|
|
{
|
|
ULONG Version;
|
|
ULONG italicAngle;
|
|
SHORT underlinePosition;
|
|
SHORT underlineThickness;
|
|
ULONG fixed_pitch;
|
|
ULONG minmemType42;
|
|
ULONG maxmemType42;
|
|
ULONG minmemType1;
|
|
ULONG maxmemType1;
|
|
} TT_POST;
|
|
|
|
typedef struct
|
|
{
|
|
USHORT version;
|
|
SHORT xAvgCharWidth;
|
|
USHORT usWeightClass;
|
|
USHORT usWidthClass;
|
|
SHORT fsType;
|
|
SHORT ySubscriptXSize;
|
|
SHORT ySubscriptYSize;
|
|
SHORT ySubscriptXOffset;
|
|
SHORT ySubscriptYOffset;
|
|
SHORT ySuperscriptXSize;
|
|
SHORT ySuperscriptYSize;
|
|
SHORT ySuperscriptXOffset;
|
|
SHORT ySuperscriptYOffset;
|
|
SHORT yStrikeoutSize;
|
|
SHORT yStrikeoutPosition;
|
|
SHORT sFamilyClass;
|
|
PANOSE panose;
|
|
ULONG ulUnicodeRange1;
|
|
ULONG ulUnicodeRange2;
|
|
ULONG ulUnicodeRange3;
|
|
ULONG ulUnicodeRange4;
|
|
CHAR achVendID[4];
|
|
USHORT fsSelection;
|
|
USHORT usFirstCharIndex;
|
|
USHORT usLastCharIndex;
|
|
/* According to the Apple spec, original version didn't have the below fields,
|
|
* version numbers were taken from the OpenType spec.
|
|
*/
|
|
/* version 0 (TrueType 1.5) */
|
|
USHORT sTypoAscender;
|
|
USHORT sTypoDescender;
|
|
USHORT sTypoLineGap;
|
|
USHORT usWinAscent;
|
|
USHORT usWinDescent;
|
|
/* version 1 (TrueType 1.66) */
|
|
ULONG ulCodePageRange1;
|
|
ULONG ulCodePageRange2;
|
|
/* version 2 (OpenType 1.2) */
|
|
SHORT sxHeight;
|
|
SHORT sCapHeight;
|
|
USHORT usDefaultChar;
|
|
USHORT usBreakChar;
|
|
USHORT usMaxContext;
|
|
} TT_OS2_V2;
|
|
|
|
typedef struct {
|
|
ULONG version;
|
|
SHORT ascender;
|
|
SHORT descender;
|
|
SHORT linegap;
|
|
USHORT advanceWidthMax;
|
|
SHORT minLeftSideBearing;
|
|
SHORT minRightSideBearing;
|
|
SHORT xMaxExtent;
|
|
SHORT caretSlopeRise;
|
|
SHORT caretSlopeRun;
|
|
SHORT caretOffset;
|
|
SHORT reserved[4];
|
|
SHORT metricDataFormat;
|
|
USHORT numberOfHMetrics;
|
|
} TT_HHEA;
|
|
|
|
#include "poppack.h"
|
|
|
|
enum OS2_FSSELECTION {
|
|
OS2_FSSELECTION_ITALIC = 1 << 0,
|
|
OS2_FSSELECTION_UNDERSCORE = 1 << 1,
|
|
OS2_FSSELECTION_NEGATIVE = 1 << 2,
|
|
OS2_FSSELECTION_OUTLINED = 1 << 3,
|
|
OS2_FSSELECTION_STRIKEOUT = 1 << 4,
|
|
OS2_FSSELECTION_BOLD = 1 << 5,
|
|
OS2_FSSELECTION_REGULAR = 1 << 6,
|
|
OS2_FSSELECTION_USE_TYPO_METRICS = 1 << 7,
|
|
OS2_FSSELECTION_WWS = 1 << 8,
|
|
OS2_FSSELECTION_OBLIQUE = 1 << 9
|
|
};
|
|
|
|
typedef struct {
|
|
WORD platformID;
|
|
WORD encodingID;
|
|
WORD languageID;
|
|
WORD nameID;
|
|
WORD length;
|
|
WORD offset;
|
|
} TT_NameRecord;
|
|
|
|
typedef struct {
|
|
WORD format;
|
|
WORD count;
|
|
WORD stringOffset;
|
|
TT_NameRecord nameRecord[1];
|
|
} TT_NAME_V0;
|
|
|
|
enum OPENTYPE_PLATFORM_ID
|
|
{
|
|
OPENTYPE_PLATFORM_UNICODE = 0,
|
|
OPENTYPE_PLATFORM_MAC,
|
|
OPENTYPE_PLATFORM_ISO,
|
|
OPENTYPE_PLATFORM_WIN,
|
|
OPENTYPE_PLATFORM_CUSTOM
|
|
};
|
|
|
|
enum TT_NAME_WINDOWS_ENCODING_ID
|
|
{
|
|
TT_NAME_WINDOWS_ENCODING_SYMBOL = 0,
|
|
TT_NAME_WINDOWS_ENCODING_UCS2,
|
|
TT_NAME_WINDOWS_ENCODING_SJIS,
|
|
TT_NAME_WINDOWS_ENCODING_PRC,
|
|
TT_NAME_WINDOWS_ENCODING_BIG5,
|
|
TT_NAME_WINDOWS_ENCODING_WANSUNG,
|
|
TT_NAME_WINDOWS_ENCODING_JOHAB,
|
|
TT_NAME_WINDOWS_ENCODING_RESERVED1,
|
|
TT_NAME_WINDOWS_ENCODING_RESERVED2,
|
|
TT_NAME_WINDOWS_ENCODING_RESERVED3,
|
|
TT_NAME_WINDOWS_ENCODING_UCS4
|
|
};
|
|
|
|
enum TT_NAME_MAC_ENCODING_ID
|
|
{
|
|
TT_NAME_MAC_ENCODING_ROMAN = 0,
|
|
TT_NAME_MAC_ENCODING_JAPANESE,
|
|
TT_NAME_MAC_ENCODING_TRAD_CHINESE,
|
|
TT_NAME_MAC_ENCODING_KOREAN,
|
|
TT_NAME_MAC_ENCODING_ARABIC,
|
|
TT_NAME_MAC_ENCODING_HEBREW,
|
|
TT_NAME_MAC_ENCODING_GREEK,
|
|
TT_NAME_MAC_ENCODING_RUSSIAN,
|
|
TT_NAME_MAC_ENCODING_RSYMBOL,
|
|
TT_NAME_MAC_ENCODING_DEVANAGARI,
|
|
TT_NAME_MAC_ENCODING_GURMUKHI,
|
|
TT_NAME_MAC_ENCODING_GUJARATI,
|
|
TT_NAME_MAC_ENCODING_ORIYA,
|
|
TT_NAME_MAC_ENCODING_BENGALI,
|
|
TT_NAME_MAC_ENCODING_TAMIL,
|
|
TT_NAME_MAC_ENCODING_TELUGU,
|
|
TT_NAME_MAC_ENCODING_KANNADA,
|
|
TT_NAME_MAC_ENCODING_MALAYALAM,
|
|
TT_NAME_MAC_ENCODING_SINHALESE,
|
|
TT_NAME_MAC_ENCODING_BURMESE,
|
|
TT_NAME_MAC_ENCODING_KHMER,
|
|
TT_NAME_MAC_ENCODING_THAI,
|
|
TT_NAME_MAC_ENCODING_LAOTIAN,
|
|
TT_NAME_MAC_ENCODING_GEORGIAN,
|
|
TT_NAME_MAC_ENCODING_ARMENIAN,
|
|
TT_NAME_MAC_ENCODING_SIMPL_CHINESE,
|
|
TT_NAME_MAC_ENCODING_TIBETAN,
|
|
TT_NAME_MAC_ENCODING_MONGOLIAN,
|
|
TT_NAME_MAC_ENCODING_GEEZ,
|
|
TT_NAME_MAC_ENCODING_SLAVIC,
|
|
TT_NAME_MAC_ENCODING_VIETNAMESE,
|
|
TT_NAME_MAC_ENCODING_SINDHI,
|
|
TT_NAME_MAC_ENCODING_UNINTERPRETED
|
|
};
|
|
|
|
enum TT_NAME_MAC_LANGUAGE_ID
|
|
{
|
|
TT_NAME_MAC_LANGID_ENGLISH = 0,
|
|
TT_NAME_MAC_LANGID_FRENCH,
|
|
TT_NAME_MAC_LANGID_GERMAN,
|
|
TT_NAME_MAC_LANGID_ITALIAN,
|
|
TT_NAME_MAC_LANGID_DUTCH,
|
|
TT_NAME_MAC_LANGID_SWEDISH,
|
|
TT_NAME_MAC_LANGID_SPANISH,
|
|
TT_NAME_MAC_LANGID_DANISH,
|
|
TT_NAME_MAC_LANGID_PORTUGUESE,
|
|
TT_NAME_MAC_LANGID_NORWEGIAN,
|
|
TT_NAME_MAC_LANGID_HEBREW,
|
|
TT_NAME_MAC_LANGID_JAPANESE,
|
|
TT_NAME_MAC_LANGID_ARABIC,
|
|
TT_NAME_MAC_LANGID_FINNISH,
|
|
TT_NAME_MAC_LANGID_GREEK,
|
|
TT_NAME_MAC_LANGID_ICELANDIC,
|
|
TT_NAME_MAC_LANGID_MALTESE,
|
|
TT_NAME_MAC_LANGID_TURKISH,
|
|
TT_NAME_MAC_LANGID_CROATIAN,
|
|
TT_NAME_MAC_LANGID_TRAD_CHINESE,
|
|
TT_NAME_MAC_LANGID_URDU,
|
|
TT_NAME_MAC_LANGID_HINDI,
|
|
TT_NAME_MAC_LANGID_THAI,
|
|
TT_NAME_MAC_LANGID_KOREAN,
|
|
TT_NAME_MAC_LANGID_LITHUANIAN,
|
|
TT_NAME_MAC_LANGID_POLISH,
|
|
TT_NAME_MAC_LANGID_HUNGARIAN,
|
|
TT_NAME_MAC_LANGID_ESTONIAN,
|
|
TT_NAME_MAC_LANGID_LATVIAN,
|
|
TT_NAME_MAC_LANGID_SAMI,
|
|
TT_NAME_MAC_LANGID_FAROESE,
|
|
TT_NAME_MAC_LANGID_FARSI,
|
|
TT_NAME_MAC_LANGID_RUSSIAN,
|
|
TT_NAME_MAC_LANGID_SIMPL_CHINESE,
|
|
TT_NAME_MAC_LANGID_FLEMISH,
|
|
TT_NAME_MAC_LANGID_GAELIC,
|
|
TT_NAME_MAC_LANGID_ALBANIAN,
|
|
TT_NAME_MAC_LANGID_ROMANIAN,
|
|
TT_NAME_MAC_LANGID_CZECH,
|
|
TT_NAME_MAC_LANGID_SLOVAK,
|
|
TT_NAME_MAC_LANGID_SLOVENIAN,
|
|
TT_NAME_MAC_LANGID_YIDDISH,
|
|
TT_NAME_MAC_LANGID_SERBIAN,
|
|
TT_NAME_MAC_LANGID_MACEDONIAN,
|
|
TT_NAME_MAC_LANGID_BULGARIAN,
|
|
TT_NAME_MAC_LANGID_UKRAINIAN,
|
|
TT_NAME_MAC_LANGID_BYELORUSSIAN,
|
|
TT_NAME_MAC_LANGID_UZBEK,
|
|
TT_NAME_MAC_LANGID_KAZAKH,
|
|
TT_NAME_MAC_LANGID_AZERB_CYR,
|
|
TT_NAME_MAC_LANGID_AZERB_ARABIC,
|
|
TT_NAME_MAC_LANGID_ARMENIAN,
|
|
TT_NAME_MAC_LANGID_GEORGIAN,
|
|
TT_NAME_MAC_LANGID_MOLDAVIAN,
|
|
TT_NAME_MAC_LANGID_KIRGHIZ,
|
|
TT_NAME_MAC_LANGID_TAJIKI,
|
|
TT_NAME_MAC_LANGID_TURKMEN,
|
|
TT_NAME_MAC_LANGID_MONGOLIAN,
|
|
TT_NAME_MAC_LANGID_MONGOLIAN_CYR,
|
|
TT_NAME_MAC_LANGID_PASHTO,
|
|
TT_NAME_MAC_LANGID_KURDISH,
|
|
TT_NAME_MAC_LANGID_KASHMIRI,
|
|
TT_NAME_MAC_LANGID_SINDHI,
|
|
TT_NAME_MAC_LANGID_TIBETAN,
|
|
TT_NAME_MAC_LANGID_NEPALI,
|
|
TT_NAME_MAC_LANGID_SANSKRIT,
|
|
TT_NAME_MAC_LANGID_MARATHI,
|
|
TT_NAME_MAC_LANGID_BENGALI,
|
|
TT_NAME_MAC_LANGID_ASSAMESE,
|
|
TT_NAME_MAC_LANGID_GUJARATI,
|
|
TT_NAME_MAC_LANGID_PUNJABI,
|
|
TT_NAME_MAC_LANGID_ORIYA,
|
|
TT_NAME_MAC_LANGID_MALAYALAM,
|
|
TT_NAME_MAC_LANGID_KANNADA,
|
|
TT_NAME_MAC_LANGID_TAMIL,
|
|
TT_NAME_MAC_LANGID_TELUGU,
|
|
TT_NAME_MAC_LANGID_SINHALESE,
|
|
TT_NAME_MAC_LANGID_BURMESE,
|
|
TT_NAME_MAC_LANGID_KHMER,
|
|
TT_NAME_MAC_LANGID_LAO,
|
|
TT_NAME_MAC_LANGID_VIETNAMESE,
|
|
TT_NAME_MAC_LANGID_INDONESIAN,
|
|
TT_NAME_MAC_LANGID_TAGALONG,
|
|
TT_NAME_MAC_LANGID_MALAY_ROMAN,
|
|
TT_NAME_MAC_LANGID_MALAY_ARABIC,
|
|
TT_NAME_MAC_LANGID_AMHARIC,
|
|
TT_NAME_MAC_LANGID_TIGRINYA,
|
|
TT_NAME_MAC_LANGID_GALLA,
|
|
TT_NAME_MAC_LANGID_SOMALI,
|
|
TT_NAME_MAC_LANGID_SWAHILI,
|
|
TT_NAME_MAC_LANGID_KINYARWANDA,
|
|
TT_NAME_MAC_LANGID_RUNDI,
|
|
TT_NAME_MAC_LANGID_NYANJA,
|
|
TT_NAME_MAC_LANGID_MALAGASY,
|
|
TT_NAME_MAC_LANGID_ESPERANTO,
|
|
TT_NAME_MAC_LANGID_WELSH,
|
|
TT_NAME_MAC_LANGID_BASQUE,
|
|
TT_NAME_MAC_LANGID_CATALAN,
|
|
TT_NAME_MAC_LANGID_LATIN,
|
|
TT_NAME_MAC_LANGID_QUENCHUA,
|
|
TT_NAME_MAC_LANGID_GUARANI,
|
|
TT_NAME_MAC_LANGID_AYMARA,
|
|
TT_NAME_MAC_LANGID_TATAR,
|
|
TT_NAME_MAC_LANGID_UIGHUR,
|
|
TT_NAME_MAC_LANGID_DZONGKHA,
|
|
TT_NAME_MAC_LANGID_JAVANESE,
|
|
TT_NAME_MAC_LANGID_SUNDANESE,
|
|
TT_NAME_MAC_LANGID_GALICIAN,
|
|
TT_NAME_MAC_LANGID_AFRIKAANS,
|
|
TT_NAME_MAC_LANGID_BRETON,
|
|
TT_NAME_MAC_LANGID_INUKTITUT,
|
|
TT_NAME_MAC_LANGID_SCOTTISH_GAELIC,
|
|
TT_NAME_MAC_LANGID_MANX_GAELIC,
|
|
TT_NAME_MAC_LANGID_IRISH_GAELIC,
|
|
TT_NAME_MAC_LANGID_TONGAN,
|
|
TT_NAME_MAC_LANGID_GREEK_POLYTONIC,
|
|
TT_NAME_MAC_LANGID_GREENLANDIC,
|
|
TT_NAME_MAC_LANGID_AZER_ROMAN
|
|
};
|
|
|
|
/* Names are indexed with TT_NAME_MAC_LANGUAGE_ID values */
|
|
static const char name_mac_langid_to_locale[][10] = {
|
|
"en-US",
|
|
"fr-FR",
|
|
"de-DE",
|
|
"it-IT",
|
|
"nl-NL",
|
|
"sv-SE",
|
|
"es-ES",
|
|
"da-DA",
|
|
"pt-PT",
|
|
"no-NO",
|
|
"he-IL",
|
|
"ja-JP",
|
|
"ar-AR",
|
|
"fi-FI",
|
|
"el-GR",
|
|
"is-IS",
|
|
"mt-MT",
|
|
"tr-TR",
|
|
"hr-HR",
|
|
"zh-HK",
|
|
"ur-PK",
|
|
"hi-IN",
|
|
"th-TH",
|
|
"ko-KR",
|
|
"lt-LT",
|
|
"pl-PL",
|
|
"hu-HU",
|
|
"et-EE",
|
|
"lv-LV",
|
|
"se-NO",
|
|
"fo-FO",
|
|
"fa-IR",
|
|
"ru-RU",
|
|
"zh-CN",
|
|
"nl-BE",
|
|
"gd-GB",
|
|
"sq-AL",
|
|
"ro-RO",
|
|
"cs-CZ",
|
|
"sk-SK",
|
|
"sl-SI",
|
|
"",
|
|
"sr-Latn",
|
|
"mk-MK",
|
|
"bg-BG",
|
|
"uk-UA",
|
|
"be-BY",
|
|
"uz-Latn",
|
|
"kk-KZ",
|
|
"az-Cyrl-AZ",
|
|
"az-AZ",
|
|
"hy-AM",
|
|
"ka-GE",
|
|
"",
|
|
"",
|
|
"tg-TJ",
|
|
"tk-TM",
|
|
"mn-Mong",
|
|
"mn-MN",
|
|
"ps-AF",
|
|
"ku-Arab",
|
|
"",
|
|
"sd-Arab",
|
|
"bo-CN",
|
|
"ne-NP",
|
|
"sa-IN",
|
|
"mr-IN",
|
|
"bn-IN",
|
|
"as-IN",
|
|
"gu-IN",
|
|
"pa-Arab",
|
|
"or-IN",
|
|
"ml-IN",
|
|
"kn-IN",
|
|
"ta-LK",
|
|
"te-IN",
|
|
"si-LK",
|
|
"",
|
|
"km-KH",
|
|
"lo-LA",
|
|
"vi-VN",
|
|
"id-ID",
|
|
"",
|
|
"ms-MY",
|
|
"ms-Arab",
|
|
"am-ET",
|
|
"ti-ET",
|
|
"",
|
|
"",
|
|
"sw-KE",
|
|
"rw-RW",
|
|
"",
|
|
"",
|
|
"",
|
|
"",
|
|
"cy-GB",
|
|
"eu-ES",
|
|
"ca-ES",
|
|
"",
|
|
"",
|
|
"",
|
|
"",
|
|
"tt-RU",
|
|
"ug-CN",
|
|
"",
|
|
"",
|
|
"",
|
|
"gl-ES",
|
|
"af-ZA",
|
|
"br-FR",
|
|
"iu-Latn-CA",
|
|
"gd-GB",
|
|
"",
|
|
"ga-IE",
|
|
"",
|
|
"",
|
|
"kl-GL",
|
|
"az-Latn"
|
|
};
|
|
|
|
enum OPENTYPE_STRING_ID
|
|
{
|
|
OPENTYPE_STRING_COPYRIGHT_NOTICE = 0,
|
|
OPENTYPE_STRING_FAMILY_NAME,
|
|
OPENTYPE_STRING_SUBFAMILY_NAME,
|
|
OPENTYPE_STRING_UNIQUE_IDENTIFIER,
|
|
OPENTYPE_STRING_FULL_FONTNAME,
|
|
OPENTYPE_STRING_VERSION_STRING,
|
|
OPENTYPE_STRING_POSTSCRIPT_FONTNAME,
|
|
OPENTYPE_STRING_TRADEMARK,
|
|
OPENTYPE_STRING_MANUFACTURER,
|
|
OPENTYPE_STRING_DESIGNER,
|
|
OPENTYPE_STRING_DESCRIPTION,
|
|
OPENTYPE_STRING_VENDOR_URL,
|
|
OPENTYPE_STRING_DESIGNER_URL,
|
|
OPENTYPE_STRING_LICENSE_DESCRIPTION,
|
|
OPENTYPE_STRING_LICENSE_INFO_URL,
|
|
OPENTYPE_STRING_RESERVED_ID15,
|
|
OPENTYPE_STRING_PREFERRED_FAMILY_NAME,
|
|
OPENTYPE_STRING_PREFERRED_SUBFAMILY_NAME,
|
|
OPENTYPE_STRING_COMPATIBLE_FULLNAME,
|
|
OPENTYPE_STRING_SAMPLE_TEXT,
|
|
OPENTYPE_STRING_POSTSCRIPT_CID_NAME,
|
|
OPENTYPE_STRING_WWS_FAMILY_NAME,
|
|
OPENTYPE_STRING_WWS_SUBFAMILY_NAME
|
|
};
|
|
|
|
static const UINT16 dwriteid_to_opentypeid[DWRITE_INFORMATIONAL_STRING_POSTSCRIPT_CID_NAME+1] =
|
|
{
|
|
(UINT16)-1, /* DWRITE_INFORMATIONAL_STRING_NONE is not used */
|
|
OPENTYPE_STRING_COPYRIGHT_NOTICE,
|
|
OPENTYPE_STRING_VERSION_STRING,
|
|
OPENTYPE_STRING_TRADEMARK,
|
|
OPENTYPE_STRING_MANUFACTURER,
|
|
OPENTYPE_STRING_DESIGNER,
|
|
OPENTYPE_STRING_DESIGNER_URL,
|
|
OPENTYPE_STRING_DESCRIPTION,
|
|
OPENTYPE_STRING_VENDOR_URL,
|
|
OPENTYPE_STRING_LICENSE_DESCRIPTION,
|
|
OPENTYPE_STRING_LICENSE_INFO_URL,
|
|
OPENTYPE_STRING_FAMILY_NAME,
|
|
OPENTYPE_STRING_SUBFAMILY_NAME,
|
|
OPENTYPE_STRING_PREFERRED_FAMILY_NAME,
|
|
OPENTYPE_STRING_PREFERRED_SUBFAMILY_NAME,
|
|
OPENTYPE_STRING_SAMPLE_TEXT,
|
|
OPENTYPE_STRING_FULL_FONTNAME,
|
|
OPENTYPE_STRING_POSTSCRIPT_FONTNAME,
|
|
OPENTYPE_STRING_POSTSCRIPT_CID_NAME
|
|
};
|
|
|
|
HRESULT opentype_analyze_font(IDWriteFontFileStream *stream, UINT32* font_count, DWRITE_FONT_FILE_TYPE *file_type, DWRITE_FONT_FACE_TYPE *face_type, BOOL *supported)
|
|
{
|
|
/* TODO: Do font validation */
|
|
const void *font_data;
|
|
const char* tag;
|
|
void *context;
|
|
HRESULT hr;
|
|
|
|
hr = IDWriteFontFileStream_ReadFileFragment(stream, &font_data, 0, sizeof(TTC_Header_V1), &context);
|
|
if (FAILED(hr))
|
|
return hr;
|
|
|
|
tag = font_data;
|
|
*supported = FALSE;
|
|
*file_type = DWRITE_FONT_FILE_TYPE_UNKNOWN;
|
|
if (face_type)
|
|
*face_type = DWRITE_FONT_FACE_TYPE_UNKNOWN;
|
|
*font_count = 0;
|
|
|
|
if (DWRITE_MAKE_OPENTYPE_TAG(tag[0], tag[1], tag[2], tag[3]) == MS_TTCF_TAG)
|
|
{
|
|
const TTC_Header_V1 *header = font_data;
|
|
*font_count = GET_BE_DWORD(header->numFonts);
|
|
*file_type = DWRITE_FONT_FILE_TYPE_TRUETYPE_COLLECTION;
|
|
if (face_type)
|
|
*face_type = DWRITE_FONT_FACE_TYPE_TRUETYPE_COLLECTION;
|
|
*supported = TRUE;
|
|
}
|
|
else if (GET_BE_DWORD(*(DWORD*)font_data) == 0x10000)
|
|
{
|
|
*font_count = 1;
|
|
*file_type = DWRITE_FONT_FILE_TYPE_TRUETYPE;
|
|
if (face_type)
|
|
*face_type = DWRITE_FONT_FACE_TYPE_TRUETYPE;
|
|
*supported = TRUE;
|
|
}
|
|
else if (DWRITE_MAKE_OPENTYPE_TAG(tag[0], tag[1], tag[2], tag[3]) == MS_OTTO_TAG)
|
|
{
|
|
*file_type = DWRITE_FONT_FILE_TYPE_CFF;
|
|
}
|
|
|
|
IDWriteFontFileStream_ReleaseFileFragment(stream, context);
|
|
return S_OK;
|
|
}
|
|
|
|
HRESULT opentype_get_font_table(IDWriteFontFileStream *stream, DWRITE_FONT_FACE_TYPE type, UINT32 font_index, UINT32 tag,
|
|
const void **table_data, void **table_context, UINT32 *table_size, BOOL *found)
|
|
{
|
|
HRESULT hr;
|
|
TTC_SFNT_V1 *font_header = NULL;
|
|
void *sfnt_context;
|
|
TT_TableRecord *table_record = NULL;
|
|
void *table_record_context;
|
|
int table_count, table_offset = 0;
|
|
int i;
|
|
|
|
if (found) *found = FALSE;
|
|
if (table_size) *table_size = 0;
|
|
|
|
if (type == DWRITE_FONT_FACE_TYPE_TRUETYPE_COLLECTION) {
|
|
const TTC_Header_V1 *ttc_header;
|
|
void * ttc_context;
|
|
hr = IDWriteFontFileStream_ReadFileFragment(stream, (const void**)&ttc_header, 0, sizeof(*ttc_header), &ttc_context);
|
|
if (SUCCEEDED(hr)) {
|
|
table_offset = GET_BE_DWORD(ttc_header->OffsetTable[0]);
|
|
if (font_index >= GET_BE_DWORD(ttc_header->numFonts))
|
|
hr = E_INVALIDARG;
|
|
else
|
|
hr = IDWriteFontFileStream_ReadFileFragment(stream, (const void**)&font_header, table_offset, sizeof(*font_header), &sfnt_context);
|
|
IDWriteFontFileStream_ReleaseFileFragment(stream, ttc_context);
|
|
}
|
|
}
|
|
else
|
|
hr = IDWriteFontFileStream_ReadFileFragment(stream, (const void**)&font_header, 0, sizeof(*font_header), &sfnt_context);
|
|
|
|
if (FAILED(hr))
|
|
return hr;
|
|
|
|
table_count = GET_BE_WORD(font_header->numTables);
|
|
table_offset += sizeof(*font_header);
|
|
for (i = 0; i < table_count; i++)
|
|
{
|
|
hr = IDWriteFontFileStream_ReadFileFragment(stream, (const void**)&table_record, table_offset, sizeof(*table_record), &table_record_context);
|
|
if (FAILED(hr))
|
|
break;
|
|
if (DWRITE_MAKE_OPENTYPE_TAG(table_record->tag[0], table_record->tag[1], table_record->tag[2], table_record->tag[3]) == tag)
|
|
break;
|
|
IDWriteFontFileStream_ReleaseFileFragment(stream, table_record_context);
|
|
table_offset += sizeof(*table_record);
|
|
}
|
|
|
|
IDWriteFontFileStream_ReleaseFileFragment(stream, sfnt_context);
|
|
if (SUCCEEDED(hr) && i < table_count)
|
|
{
|
|
int offset = GET_BE_DWORD(table_record->offset);
|
|
int length = GET_BE_DWORD(table_record->length);
|
|
IDWriteFontFileStream_ReleaseFileFragment(stream, table_record_context);
|
|
|
|
if (found) *found = TRUE;
|
|
if (table_size) *table_size = length;
|
|
hr = IDWriteFontFileStream_ReadFileFragment(stream, table_data, offset, length, table_context);
|
|
}
|
|
|
|
return hr;
|
|
}
|
|
|
|
/**********
|
|
* CMAP
|
|
**********/
|
|
|
|
static int compare_group(const void *a, const void* b)
|
|
{
|
|
const DWORD *chr = a;
|
|
const CMAP_SegmentedCoverage_group *group = b;
|
|
|
|
if (*chr < GET_BE_DWORD(group->startCharCode))
|
|
return -1;
|
|
if (*chr > GET_BE_DWORD(group->endCharCode))
|
|
return 1;
|
|
return 0;
|
|
}
|
|
|
|
static void CMAP4_GetGlyphIndex(CMAP_SegmentMapping_0* format, UINT32 utf32c, UINT16 *pgi)
|
|
{
|
|
WORD *startCode;
|
|
SHORT *idDelta;
|
|
WORD *idRangeOffset;
|
|
int segment;
|
|
|
|
int segment_count = GET_BE_WORD(format->segCountX2)/2;
|
|
/* This is correct because of the padding before startCode */
|
|
startCode = (WORD*)((BYTE*)format + sizeof(CMAP_SegmentMapping_0) + (sizeof(WORD) * segment_count));
|
|
idDelta = (SHORT*)(((BYTE*)startCode) + (sizeof(WORD) * segment_count));
|
|
idRangeOffset = (WORD*)(((BYTE*)idDelta) + (sizeof(WORD) * segment_count));
|
|
|
|
segment = 0;
|
|
while(GET_BE_WORD(format->endCode[segment]) < 0xffff)
|
|
{
|
|
if (utf32c <= GET_BE_WORD(format->endCode[segment]))
|
|
break;
|
|
segment++;
|
|
}
|
|
if (segment >= segment_count)
|
|
return;
|
|
TRACE("Segment %i of %i\n",segment, segment_count);
|
|
if (GET_BE_WORD(startCode[segment]) > utf32c)
|
|
return;
|
|
TRACE("In range %i -> %i\n", GET_BE_WORD(startCode[segment]), GET_BE_WORD(format->endCode[segment]));
|
|
if (GET_BE_WORD(idRangeOffset[segment]) == 0)
|
|
{
|
|
*pgi = (SHORT)(GET_BE_WORD(idDelta[segment])) + utf32c;
|
|
}
|
|
else
|
|
{
|
|
WORD ro = GET_BE_WORD(idRangeOffset[segment])/2;
|
|
WORD co = (utf32c - GET_BE_WORD(startCode[segment]));
|
|
WORD *index = (WORD*)((BYTE*)&idRangeOffset[segment] + (ro + co));
|
|
*pgi = GET_BE_WORD(*index);
|
|
}
|
|
}
|
|
|
|
static void CMAP12_GetGlyphIndex(CMAP_SegmentedCoverage* format, UINT32 utf32c, UINT16 *pgi)
|
|
{
|
|
CMAP_SegmentedCoverage_group *group;
|
|
|
|
group = bsearch(&utf32c, format->groups, GET_BE_DWORD(format->nGroups),
|
|
sizeof(CMAP_SegmentedCoverage_group), compare_group);
|
|
|
|
if (group)
|
|
{
|
|
DWORD offset = utf32c - GET_BE_DWORD(group->startCharCode);
|
|
*pgi = GET_BE_DWORD(group->startGlyphID) + offset;
|
|
}
|
|
}
|
|
|
|
void opentype_cmap_get_glyphindex(void *data, UINT32 utf32c, UINT16 *pgi)
|
|
{
|
|
CMAP_Header *CMAP_Table = data;
|
|
int i;
|
|
|
|
*pgi = 0;
|
|
|
|
for (i = 0; i < GET_BE_WORD(CMAP_Table->numTables); i++)
|
|
{
|
|
WORD type;
|
|
WORD *table;
|
|
|
|
if (GET_BE_WORD(CMAP_Table->tables[i].platformID) != 3)
|
|
continue;
|
|
|
|
table = (WORD*)(((BYTE*)CMAP_Table) + GET_BE_DWORD(CMAP_Table->tables[i].offset));
|
|
type = GET_BE_WORD(*table);
|
|
TRACE("table type %i\n", type);
|
|
|
|
switch (type)
|
|
{
|
|
case OPENTYPE_CMAP_TABLE_SEGMENT_MAPPING:
|
|
CMAP4_GetGlyphIndex((CMAP_SegmentMapping_0*) table, utf32c, pgi);
|
|
break;
|
|
case OPENTYPE_CMAP_TABLE_SEGMENTED_COVERAGE:
|
|
CMAP12_GetGlyphIndex((CMAP_SegmentedCoverage*) table, utf32c, pgi);
|
|
break;
|
|
default:
|
|
TRACE("table type %i unhandled.\n", type);
|
|
}
|
|
|
|
if (*pgi) return;
|
|
}
|
|
}
|
|
|
|
static UINT32 opentype_cmap_get_unicode_ranges_count(const CMAP_Header *CMAP_Table)
|
|
{
|
|
UINT32 count = 0;
|
|
int i;
|
|
|
|
for (i = 0; i < GET_BE_WORD(CMAP_Table->numTables); i++) {
|
|
WORD type;
|
|
WORD *table;
|
|
|
|
if (GET_BE_WORD(CMAP_Table->tables[i].platformID) != 3)
|
|
continue;
|
|
|
|
table = (WORD*)(((BYTE*)CMAP_Table) + GET_BE_DWORD(CMAP_Table->tables[i].offset));
|
|
type = GET_BE_WORD(*table);
|
|
TRACE("table type %i\n", type);
|
|
|
|
switch (type)
|
|
{
|
|
case OPENTYPE_CMAP_TABLE_SEGMENT_MAPPING:
|
|
{
|
|
CMAP_SegmentMapping_0 *format = (CMAP_SegmentMapping_0*)table;
|
|
count += GET_BE_WORD(format->segCountX2)/2;
|
|
break;
|
|
}
|
|
case OPENTYPE_CMAP_TABLE_SEGMENTED_COVERAGE:
|
|
{
|
|
CMAP_SegmentedCoverage *format = (CMAP_SegmentedCoverage*)table;
|
|
count += GET_BE_DWORD(format->nGroups);
|
|
break;
|
|
}
|
|
default:
|
|
FIXME("table type %i unhandled.\n", type);
|
|
}
|
|
}
|
|
|
|
return count;
|
|
}
|
|
|
|
HRESULT opentype_cmap_get_unicode_ranges(void *data, UINT32 max_count, DWRITE_UNICODE_RANGE *ranges, UINT32 *count)
|
|
{
|
|
CMAP_Header *CMAP_Table = data;
|
|
int i, k = 0;
|
|
|
|
if (!CMAP_Table)
|
|
return E_FAIL;
|
|
|
|
*count = opentype_cmap_get_unicode_ranges_count(CMAP_Table);
|
|
|
|
for (i = 0; i < GET_BE_WORD(CMAP_Table->numTables) && k < max_count; i++)
|
|
{
|
|
WORD type;
|
|
WORD *table;
|
|
int j;
|
|
|
|
if (GET_BE_WORD(CMAP_Table->tables[i].platformID) != 3)
|
|
continue;
|
|
|
|
table = (WORD*)(((BYTE*)CMAP_Table) + GET_BE_DWORD(CMAP_Table->tables[i].offset));
|
|
type = GET_BE_WORD(*table);
|
|
TRACE("table type %i\n", type);
|
|
|
|
switch (type)
|
|
{
|
|
case OPENTYPE_CMAP_TABLE_SEGMENT_MAPPING:
|
|
{
|
|
CMAP_SegmentMapping_0 *format = (CMAP_SegmentMapping_0*)table;
|
|
UINT16 segment_count = GET_BE_WORD(format->segCountX2)/2;
|
|
UINT16 *startCode = (WORD*)((BYTE*)format + sizeof(CMAP_SegmentMapping_0) + (sizeof(WORD) * segment_count));
|
|
|
|
for (j = 0; j < segment_count && GET_BE_WORD(format->endCode[j]) < 0xffff && k < max_count; j++, k++) {
|
|
ranges[k].first = GET_BE_WORD(startCode[j]);
|
|
ranges[k].last = GET_BE_WORD(format->endCode[j]);
|
|
}
|
|
break;
|
|
}
|
|
case OPENTYPE_CMAP_TABLE_SEGMENTED_COVERAGE:
|
|
{
|
|
CMAP_SegmentedCoverage *format = (CMAP_SegmentedCoverage*)table;
|
|
for (j = 0; j < GET_BE_DWORD(format->nGroups) && k < max_count; j++, k++) {
|
|
ranges[k].first = GET_BE_DWORD(format->groups[j].startCharCode);
|
|
ranges[k].last = GET_BE_DWORD(format->groups[j].endCharCode);
|
|
}
|
|
break;
|
|
}
|
|
default:
|
|
FIXME("table type %i unhandled.\n", type);
|
|
}
|
|
}
|
|
|
|
return *count > max_count ? E_NOT_SUFFICIENT_BUFFER : S_OK;
|
|
}
|
|
|
|
void opentype_get_font_metrics(IDWriteFontFileStream *stream, DWRITE_FONT_FACE_TYPE face_type, UINT32 face_index,
|
|
DWRITE_FONT_METRICS1 *metrics)
|
|
{
|
|
void *os2_context, *head_context, *post_context, *hhea_context;
|
|
const TT_OS2_V2 *tt_os2;
|
|
const TT_HEAD *tt_head;
|
|
const TT_POST *tt_post;
|
|
const TT_HHEA *tt_hhea;
|
|
|
|
memset(metrics, 0, sizeof(*metrics));
|
|
|
|
opentype_get_font_table(stream, face_type, face_index, MS_OS2_TAG, (const void**)&tt_os2, &os2_context, NULL, NULL);
|
|
opentype_get_font_table(stream, face_type, face_index, MS_HEAD_TAG, (const void**)&tt_head, &head_context, NULL, NULL);
|
|
opentype_get_font_table(stream, face_type, face_index, MS_POST_TAG, (const void**)&tt_post, &post_context, NULL, NULL);
|
|
opentype_get_font_table(stream, face_type, face_index, MS_HHEA_TAG, (const void**)&tt_hhea, &hhea_context, NULL, NULL);
|
|
|
|
if (tt_head) {
|
|
metrics->designUnitsPerEm = GET_BE_WORD(tt_head->unitsPerEm);
|
|
metrics->glyphBoxLeft = GET_BE_WORD(tt_head->xMin);
|
|
metrics->glyphBoxTop = GET_BE_WORD(tt_head->yMax);
|
|
metrics->glyphBoxRight = GET_BE_WORD(tt_head->xMax);
|
|
metrics->glyphBoxBottom = GET_BE_WORD(tt_head->yMin);
|
|
}
|
|
|
|
if (tt_os2) {
|
|
USHORT version = GET_BE_WORD(tt_os2->version);
|
|
|
|
metrics->ascent = GET_BE_WORD(tt_os2->usWinAscent);
|
|
metrics->descent = GET_BE_WORD(tt_os2->usWinDescent);
|
|
|
|
/* line gap is estimated using two sets of ascender/descender values and 'hhea' line gap */
|
|
if (tt_hhea) {
|
|
SHORT descender = (SHORT)GET_BE_WORD(tt_hhea->descender);
|
|
INT32 linegap;
|
|
|
|
linegap = GET_BE_WORD(tt_hhea->ascender) + abs(descender) + GET_BE_WORD(tt_hhea->linegap) -
|
|
metrics->ascent - metrics->descent;
|
|
metrics->lineGap = linegap > 0 ? linegap : 0;
|
|
}
|
|
|
|
metrics->strikethroughPosition = GET_BE_WORD(tt_os2->yStrikeoutPosition);
|
|
metrics->strikethroughThickness = GET_BE_WORD(tt_os2->yStrikeoutSize);
|
|
metrics->subscriptPositionX = GET_BE_WORD(tt_os2->ySubscriptXOffset);
|
|
/* Y offset is stored as positive offset below baseline */
|
|
metrics->subscriptPositionY = -GET_BE_WORD(tt_os2->ySubscriptYOffset);
|
|
metrics->subscriptSizeX = GET_BE_WORD(tt_os2->ySubscriptXSize);
|
|
metrics->subscriptSizeY = GET_BE_WORD(tt_os2->ySubscriptYSize);
|
|
metrics->superscriptPositionX = GET_BE_WORD(tt_os2->ySuperscriptXOffset);
|
|
metrics->superscriptPositionY = GET_BE_WORD(tt_os2->ySuperscriptYOffset);
|
|
metrics->superscriptSizeX = GET_BE_WORD(tt_os2->ySuperscriptXSize);
|
|
metrics->superscriptSizeY = GET_BE_WORD(tt_os2->ySuperscriptYSize);
|
|
|
|
/* version 2 fields */
|
|
if (version >= 2) {
|
|
metrics->capHeight = GET_BE_WORD(tt_os2->sCapHeight);
|
|
metrics->xHeight = GET_BE_WORD(tt_os2->sxHeight);
|
|
}
|
|
|
|
/* version 4 fields */
|
|
if (version >= 4) {
|
|
if (GET_BE_WORD(tt_os2->fsSelection) & OS2_FSSELECTION_USE_TYPO_METRICS) {
|
|
SHORT descent = GET_BE_WORD(tt_os2->sTypoDescender);
|
|
metrics->ascent = GET_BE_WORD(tt_os2->sTypoAscender);
|
|
metrics->descent = descent < 0 ? -descent : 0;
|
|
metrics->lineGap = GET_BE_WORD(tt_os2->sTypoLineGap);
|
|
metrics->hasTypographicMetrics = TRUE;
|
|
}
|
|
}
|
|
}
|
|
|
|
if (tt_post) {
|
|
metrics->underlinePosition = GET_BE_WORD(tt_post->underlinePosition);
|
|
metrics->underlineThickness = GET_BE_WORD(tt_post->underlineThickness);
|
|
}
|
|
|
|
/* estimate missing metrics */
|
|
if (metrics->xHeight == 0)
|
|
metrics->xHeight = metrics->designUnitsPerEm / 2;
|
|
if (metrics->capHeight == 0)
|
|
metrics->capHeight = metrics->designUnitsPerEm * 7 / 10;
|
|
|
|
if (tt_os2)
|
|
IDWriteFontFileStream_ReleaseFileFragment(stream, os2_context);
|
|
if (tt_head)
|
|
IDWriteFontFileStream_ReleaseFileFragment(stream, head_context);
|
|
if (tt_post)
|
|
IDWriteFontFileStream_ReleaseFileFragment(stream, post_context);
|
|
if (tt_hhea)
|
|
IDWriteFontFileStream_ReleaseFileFragment(stream, hhea_context);
|
|
}
|
|
|
|
void opentype_get_font_properties(const void *os2, const void *head, DWRITE_FONT_STRETCH *stretch, DWRITE_FONT_WEIGHT *weight, DWRITE_FONT_STYLE *style)
|
|
{
|
|
TT_OS2_V2 *tt_os2 = (TT_OS2_V2*)os2;
|
|
TT_HEAD *tt_head = (TT_HEAD*)head;
|
|
|
|
/* default stretch, weight and style to normal */
|
|
*stretch = DWRITE_FONT_STRETCH_NORMAL;
|
|
*weight = DWRITE_FONT_WEIGHT_NORMAL;
|
|
*style = DWRITE_FONT_STYLE_NORMAL;
|
|
|
|
/* DWRITE_FONT_STRETCH enumeration values directly match font data values */
|
|
if (tt_os2) {
|
|
if (GET_BE_WORD(tt_os2->usWidthClass) <= DWRITE_FONT_STRETCH_ULTRA_EXPANDED)
|
|
*stretch = GET_BE_WORD(tt_os2->usWidthClass);
|
|
|
|
*weight = GET_BE_WORD(tt_os2->usWeightClass);
|
|
TRACE("stretch=%d, weight=%d\n", *stretch, *weight);
|
|
}
|
|
|
|
if (tt_head) {
|
|
USHORT macStyle = GET_BE_WORD(tt_head->macStyle);
|
|
if (macStyle & 0x0002)
|
|
*style = DWRITE_FONT_STYLE_ITALIC;
|
|
}
|
|
}
|
|
|
|
static UINT get_name_record_codepage(enum OPENTYPE_PLATFORM_ID platform, USHORT encoding)
|
|
{
|
|
UINT codepage = 0;
|
|
|
|
switch (platform) {
|
|
case OPENTYPE_PLATFORM_UNICODE:
|
|
break;
|
|
case OPENTYPE_PLATFORM_MAC:
|
|
switch (encoding)
|
|
{
|
|
case TT_NAME_MAC_ENCODING_ROMAN:
|
|
codepage = 10000;
|
|
break;
|
|
case TT_NAME_MAC_ENCODING_JAPANESE:
|
|
codepage = 10001;
|
|
break;
|
|
case TT_NAME_MAC_ENCODING_TRAD_CHINESE:
|
|
codepage = 10002;
|
|
break;
|
|
case TT_NAME_MAC_ENCODING_KOREAN:
|
|
codepage = 10003;
|
|
break;
|
|
case TT_NAME_MAC_ENCODING_ARABIC:
|
|
codepage = 10004;
|
|
break;
|
|
case TT_NAME_MAC_ENCODING_HEBREW:
|
|
codepage = 10005;
|
|
break;
|
|
case TT_NAME_MAC_ENCODING_GREEK:
|
|
codepage = 10006;
|
|
break;
|
|
case TT_NAME_MAC_ENCODING_RUSSIAN:
|
|
codepage = 10007;
|
|
break;
|
|
case TT_NAME_MAC_ENCODING_SIMPL_CHINESE:
|
|
codepage = 10008;
|
|
break;
|
|
case TT_NAME_MAC_ENCODING_THAI:
|
|
codepage = 10021;
|
|
break;
|
|
default:
|
|
FIXME("encoding %u not handled, platform %d.\n", encoding, platform);
|
|
break;
|
|
}
|
|
break;
|
|
case OPENTYPE_PLATFORM_WIN:
|
|
switch (encoding)
|
|
{
|
|
case TT_NAME_WINDOWS_ENCODING_SYMBOL:
|
|
case TT_NAME_WINDOWS_ENCODING_UCS2:
|
|
break;
|
|
case TT_NAME_WINDOWS_ENCODING_SJIS:
|
|
codepage = 932;
|
|
break;
|
|
case TT_NAME_WINDOWS_ENCODING_PRC:
|
|
codepage = 936;
|
|
break;
|
|
case TT_NAME_WINDOWS_ENCODING_BIG5:
|
|
codepage = 950;
|
|
break;
|
|
case TT_NAME_WINDOWS_ENCODING_WANSUNG:
|
|
codepage = 20949;
|
|
break;
|
|
case TT_NAME_WINDOWS_ENCODING_JOHAB:
|
|
codepage = 1361;
|
|
break;
|
|
default:
|
|
FIXME("encoding %u not handled, platform %d.\n", encoding, platform);
|
|
break;
|
|
}
|
|
break;
|
|
default:
|
|
FIXME("unknown platform %d\n", platform);
|
|
}
|
|
|
|
return codepage;
|
|
}
|
|
|
|
static void get_name_record_locale(enum OPENTYPE_PLATFORM_ID platform, USHORT lang_id, WCHAR *locale, USHORT locale_len)
|
|
{
|
|
static const WCHAR enusW[] = {'e','n','-','U','S',0};
|
|
|
|
switch (platform) {
|
|
case OPENTYPE_PLATFORM_MAC:
|
|
{
|
|
const char *locale_name = NULL;
|
|
|
|
if (lang_id > TT_NAME_MAC_LANGID_AZER_ROMAN)
|
|
ERR("invalid mac lang id %d\n", lang_id);
|
|
else if (!name_mac_langid_to_locale[lang_id][0])
|
|
FIXME("failed to map mac lang id %d to locale name\n", lang_id);
|
|
else
|
|
locale_name = name_mac_langid_to_locale[lang_id];
|
|
|
|
if (locale_name)
|
|
MultiByteToWideChar(CP_ACP, 0, name_mac_langid_to_locale[lang_id], -1, locale, locale_len);
|
|
else
|
|
strcpyW(locale, enusW);
|
|
break;
|
|
}
|
|
case OPENTYPE_PLATFORM_WIN:
|
|
if (!LCIDToLocaleName(MAKELCID(lang_id, SORT_DEFAULT), locale, locale_len, 0)) {
|
|
FIXME("failed to get locale name for lcid=0x%08x\n", MAKELCID(lang_id, SORT_DEFAULT));
|
|
strcpyW(locale, enusW);
|
|
}
|
|
break;
|
|
default:
|
|
FIXME("unknown platform %d\n", platform);
|
|
}
|
|
}
|
|
|
|
HRESULT opentype_get_font_strings_from_id(const void *table_data, DWRITE_INFORMATIONAL_STRING_ID id, IDWriteLocalizedStrings **strings)
|
|
{
|
|
const TT_NAME_V0 *header;
|
|
BYTE *storage_area = 0;
|
|
USHORT count = 0;
|
|
UINT16 name_id;
|
|
BOOL exists;
|
|
HRESULT hr;
|
|
int i;
|
|
|
|
if (!table_data)
|
|
return E_FAIL;
|
|
|
|
hr = create_localizedstrings(strings);
|
|
if (FAILED(hr)) return hr;
|
|
|
|
header = table_data;
|
|
storage_area = (LPBYTE)table_data + GET_BE_WORD(header->stringOffset);
|
|
count = GET_BE_WORD(header->count);
|
|
|
|
name_id = dwriteid_to_opentypeid[id];
|
|
|
|
exists = FALSE;
|
|
for (i = 0; i < count; i++) {
|
|
const TT_NameRecord *record = &header->nameRecord[i];
|
|
USHORT lang_id, length, offset, encoding, platform;
|
|
|
|
if (GET_BE_WORD(record->nameID) != name_id)
|
|
continue;
|
|
|
|
exists = TRUE;
|
|
|
|
/* Right now only accept unicode and windows encoded fonts */
|
|
platform = GET_BE_WORD(record->platformID);
|
|
if (platform != OPENTYPE_PLATFORM_UNICODE &&
|
|
platform != OPENTYPE_PLATFORM_MAC &&
|
|
platform != OPENTYPE_PLATFORM_WIN)
|
|
{
|
|
FIXME("platform %i not supported\n", platform);
|
|
continue;
|
|
}
|
|
|
|
lang_id = GET_BE_WORD(record->languageID);
|
|
length = GET_BE_WORD(record->length);
|
|
offset = GET_BE_WORD(record->offset);
|
|
encoding = GET_BE_WORD(record->encodingID);
|
|
|
|
if (lang_id < 0x8000) {
|
|
WCHAR locale[LOCALE_NAME_MAX_LENGTH];
|
|
WCHAR *name_string;
|
|
UINT codepage;
|
|
|
|
codepage = get_name_record_codepage(platform, encoding);
|
|
get_name_record_locale(platform, lang_id, locale, sizeof(locale)/sizeof(WCHAR));
|
|
|
|
if (codepage) {
|
|
DWORD len = MultiByteToWideChar(codepage, 0, (LPSTR)(storage_area + offset), length, NULL, 0);
|
|
name_string = heap_alloc(sizeof(WCHAR) * (len+1));
|
|
MultiByteToWideChar(codepage, 0, (LPSTR)(storage_area + offset), length, name_string, len);
|
|
name_string[len] = 0;
|
|
}
|
|
else {
|
|
int i;
|
|
|
|
length /= sizeof(WCHAR);
|
|
name_string = heap_strdupnW((LPWSTR)(storage_area + offset), length);
|
|
for (i = 0; i < length; i++)
|
|
name_string[i] = GET_BE_WORD(name_string[i]);
|
|
}
|
|
|
|
TRACE("string %s for locale %s found\n", debugstr_w(name_string), debugstr_w(locale));
|
|
add_localizedstring(*strings, locale, name_string);
|
|
heap_free(name_string);
|
|
}
|
|
else {
|
|
FIXME("handle NAME format 1");
|
|
continue;
|
|
}
|
|
}
|
|
|
|
if (!exists) {
|
|
IDWriteLocalizedStrings_Release(*strings);
|
|
*strings = NULL;
|
|
}
|
|
|
|
return hr;
|
|
}
|