/* Copyright (C) 1999, 2000, 2002 Aladdin Enterprises. All rights reserved. This software is provided AS-IS with no warranty, either express or implied. This software is distributed under license and may not be copied, modified or distributed except as expressly authorized under the terms of the license contained in the file LICENSE in this distribution. For more information about licensing, please refer to http://www.ghostscript.com/licensing/. For information on commercial licensing, go to http://www.artifex.com/licensing/ or contact Artifex Software, Inc., 101 Lucas Valley Road #110, San Rafael, CA 94903, U.S.A., +1(415)492-9861. */ /* $Id: gdevpsft.c,v 1.35 2004/12/08 21:35:13 stefan Exp $ */ /* Write an embedded TrueType font */ #include "memory_.h" #include #include /* for qsort */ #include "gx.h" #include "gscencs.h" #include "gserrors.h" #include "gsmatrix.h" #include "gsutil.h" #include "gxfcid.h" #include "gxfont.h" #include "gxfont42.h" #include "gxttf.h" #include "stream.h" #include "spprint.h" #include "gdevpsf.h" /* Internally used options */ #define WRITE_TRUETYPE_STRIPPED 0x1000 /* internal */ #define WRITE_TRUETYPE_CID 0x2000 /* internal */ #define MAX_COMPOSITE_PIECES 3 /* adhoc */ /* * The following are only for debugging. They force various format choices * in the output. The normal (non-debugging) values for all of these are * as indicated in the comments. * * Note that these options interact. Here is the complete list of settings * that make sense. 0 -1,0,1 N/A 0,1 0,1 0xf000 -1 N/A 1 0,1 0xf000 0,1 0,1 1 0,1 */ /* Define whether to use the 0xf000 character bias for generated tables. */ #define TT_BIAS 0xf000 /* 0xf000 */ /* Define whether to use cmap format 6 never(-1), sometimes(0), always(1). */ #define TT_FORCE_CMAP_6 0 /* 0 */ /* Define whether to use the bias for the cmap format 6 "first code". */ #define TT_BIAS_CMAP_6 0 /* 0 */ /* Define whether to generate an OS/2 table if none is supplied. */ #define TT_GENERATE_OS_2 1 /* 1 */ /* Define whether to adjust the OS/2 range bits. */ #define TT_ADJUST_OS_2 1 /* 1 */ /* * End of options. */ /* ---------------- Utilities ---------------- */ #define ACCESS(base, length, vptr)\ BEGIN\ code = string_proc(pfont, (ulong)(base), length, &vptr);\ if (code < 0) return code;\ END /* Pad to a multiple of 4 bytes. */ private void put_pad(stream *s, uint length) { static const byte pad_to_4[3] = {0, 0, 0}; stream_write(s, pad_to_4, (uint)(-(int)length & 3)); } /* Put short and long values on a stream. */ private void put_ushort(stream *s, uint v) { stream_putc(s, (byte)(v >> 8)); stream_putc(s, (byte)v); } private void put_ulong(stream *s, ulong v) { put_ushort(s, (uint)(v >> 16)); put_ushort(s, (uint)v); } private void put_loca(stream *s, ulong offset, int indexToLocFormat) { if (indexToLocFormat) put_ulong(s, offset); else put_ushort(s, (uint)(offset >> 1)); } /* Get or put 2- or 4-byte quantities from/into a table. */ #define U8(p) ((uint)((p)[0])) #define S8(p) (int)((U8(p) ^ 0x80) - 0x80) #define U16(p) (((uint)((p)[0]) << 8) + (p)[1]) #define S16(p) (int)((U16(p) ^ 0x8000) - 0x8000) #define u32(p) get_u32_msb(p) private void put_u16(byte *p, uint v) { p[0] = (byte)(v >> 8); p[1] = (byte)v; } private void put_u32(byte *p, ulong v) { put_u16(p, (ushort)(v >> 16)); put_u16(p + 2, (ushort)v); } private ulong put_table(byte tab[16], const char *tname, ulong checksum, ulong offset, uint length) { memcpy(tab, (const byte *)tname, 4); put_u32(tab + 4, checksum); put_u32(tab + 8, offset + 0x40000000); put_u32(tab + 12, (ulong)length); return offset + round_up(length, 4); } /* Write one range of a TrueType font. */ private int write_range(stream *s, gs_font_type42 *pfont, ulong start, uint length) { ulong base = start; ulong limit = base + length; if_debug3('l', "[l]write_range pos = %ld, start = %lu, length = %u\n", stell(s), start, length); while (base < limit) { uint size = limit - base; const byte *ptr; int code; /* Write the largest block we can access consecutively. */ while ((code = pfont->data.string_proc(pfont, base, size, &ptr)) < 0) { if (size <= 1) return code; size >>= 1; } if (code > 0 && size > code) size = code; /* Segmented data - see z42_string_proc. */ stream_write(s, ptr, size); base += size; } return 0; } /* * Determine the Macintosh glyph number for a given character, if any. * If no glyph can be found, return -1 and store the name in *pstr. */ private int mac_glyph_index(gs_font *font, int ch, gs_const_string *pstr) { gs_glyph glyph = font->procs.encode_char(font, (gs_char)ch, GLYPH_SPACE_NAME); int code; if (glyph == gs_no_glyph) return 0; /* .notdef */ code = font->procs.glyph_name(font, glyph, pstr); assert(code >= 0); if (glyph < gs_min_cid_glyph) { gs_char mac_char; gs_glyph mac_glyph; gs_const_string mstr; /* Look (not very hard) for a match in the Mac glyph space. */ if (ch >= 32 && ch <= 126) mac_char = ch - 29; else if (ch >= 128 && ch <= 255) mac_char = ch - 30; else return -1; mac_glyph = gs_c_known_encode(mac_char, ENCODING_INDEX_MACGLYPH); if (mac_glyph == gs_no_glyph) return -1; code = gs_c_glyph_name(mac_glyph, &mstr); assert(code >= 0); if (!bytes_compare(pstr->data, pstr->size, mstr.data, mstr.size)) return (int)mac_char; } return -1; } /* ---------------- Individual tables ---------------- */ /* ------ cmap ------ */ /* Write a generated cmap table. */ static const byte cmap_initial_0[] = { 0, 0, /* table version # = 0 */ 0, 2, /* # of encoding tables = 2 */ /* First table, Macintosh */ 0, 1, /* platform ID = Macintosh */ 0, 0, /* platform encoding ID = ??? */ 0, 0, 0, 4+8+8, /* offset to table start */ /* Second table, Windows */ 0, 3, /* platform ID = Microsoft */ 0, 0, /* platform encoding ID = unknown */ 0, 0, 1, 4+8+8+6, /* offset to table start */ /* Start of Macintosh format 0 table */ 0, 0, /* format = 0, byte encoding table */ 1, 6, /* length */ 0, 0 /* version number */ }; static const byte cmap_initial_6[] = { 0, 0, /* table version # = 0 */ 0, 2, /* # of encoding tables = 2 */ /* First table, Macintosh */ 0, 1, /* platform ID = Macintosh */ 0, 0, /* platform encoding ID = ??? */ 0, 0, 0, 4+8+8, /* offset to table start */ /* Second table, Windows */ 0, 3, /* platform ID = Microsoft */ 0, 0, /* platform encoding ID = unknown */ 0, 0, 0, 4+8+8+10, /* offset to table start */ /* *VARIABLE*, add 2 x # of entries */ /* Start of Macintosh format 6 table */ 0, 6, /* format = 6, trimmed table mapping */ 0, 10, /* length *VARIABLE*, add 2 x # of entries */ 0, 0, /* version number */ 0, 0, /* first character code */ 0, 0 /* # of entries *VARIABLE* */ }; static const byte cmap_initial_4[] = { 0, 0, /* table version # = 0 */ 0, 1, /* # of encoding tables = 2 */ /* Single table, Windows */ 0, 3, /* platform ID = Microsoft */ 0, 0, /* platform encoding ID = unknown */ 0, 0, 0, 4+8 /* offset to table start */ }; static const byte cmap_sub_initial[] = { 0, 4, /* format = 4, segment mapping */ 0, 32, /* length ** VARIABLE, add 2 x # of glyphs ** */ 0, 0, /* version # */ 0, 4, /* 2 x segCount */ 0, 4, /* searchRange = 2 x 2 ^ floor(log2(segCount)) */ 0, 1, /* floor(log2(segCount)) */ 0, 0, /* 2 x segCount - searchRange */ 0, 0, /* endCount[0] **VARIABLE** */ 255, 255, /* endCount[1] */ 0, 0, /* reservedPad */ 0, 0, /* startCount[0] **VARIABLE** */ 255, 255, /* startCount[1] */ 0, 0, /* idDelta[0] */ 0, 1, /* idDelta[1] */ 0, 4, /* idRangeOffset[0] */ 0, 0 /* idRangeOffset[1] */ }; /* * The following nonsense is required because C defines sizeof() * inconsistently. */ #define CMAP_ENTRIES_SIZE (256 * 2) private void write_cmap_0(stream *s, byte* entries /*[CMAP_ENTRIES_SIZE]*/, uint num_glyphs) { int i; memset(entries + 2 * num_glyphs, 0, CMAP_ENTRIES_SIZE - 2 * num_glyphs); stream_write(s, cmap_initial_0, sizeof(cmap_initial_0)); for (i = 0; i <= 0xff; ++i) sputc(s, (byte)entries[2 * i + 1]); } private void write_cmap_6(stream *s, byte *entries /*[CMAP_ENTRIES_SIZE]*/, uint first_code, uint first_entry, uint num_entries) { byte cmap_data[sizeof(cmap_initial_6)]; memcpy(cmap_data, cmap_initial_6, sizeof(cmap_initial_6)); put_u16(cmap_data + 18, U16(cmap_data + 18) + num_entries * 2); /* offset */ put_u16(cmap_data + 22, U16(cmap_data + 22) + num_entries * 2); /* length */ put_u16(cmap_data + 26, #if TT_BIAS_CMAP_6 first_code + #endif first_entry); put_u16(cmap_data + 28, num_entries); stream_write(s, cmap_data, sizeof(cmap_data)); stream_write(s, entries + first_entry * 2, num_entries * 2); } private void write_cmap(stream *s, gs_font *font, uint first_code, int num_glyphs, gs_glyph max_glyph, int options, uint cmap_length) { byte cmap_sub[sizeof(cmap_sub_initial)]; byte entries[CMAP_ENTRIES_SIZE]; int first_entry = 0, end_entry = num_glyphs; bool can_use_trimmed = !(options & WRITE_TRUETYPE_NO_TRIMMED_TABLE); uint merge = 0; uint num_entries; int i; /* Collect the table entries. */ for (i = 0; i < num_glyphs; ++i) { gs_glyph glyph = font->procs.encode_char(font, (gs_char)i, GLYPH_SPACE_INDEX); uint glyph_index; if (glyph == gs_no_glyph || glyph < GS_MIN_GLYPH_INDEX || glyph > max_glyph ) glyph = GS_MIN_GLYPH_INDEX; glyph_index = (uint)(glyph - GS_MIN_GLYPH_INDEX); merge |= glyph_index; put_u16(entries + 2 * i, glyph_index); } while (end_entry > first_entry && !U16(entries + 2 * end_entry - 2)) --end_entry; while (first_entry < end_entry && !U16(entries + 2 * first_entry)) ++first_entry; num_entries = end_entry - first_entry; /* Write the table header and Macintosh sub-table (if any). */ #if TT_FORCE_CMAP_6 > 0 /* Always use format 6. */ write_cmap_6(s, entries, first_code, first_entry, num_entries); #else # if TT_FORCE_CMAP_6 < 0 /* Never use format 6. Use format 0 if possible. */ if (merge == (byte)merge) write_cmap_0(s, entries, num_glyphs); else # else /* TT_FORCE_CMAP == 0 */ /* * Use format 0 if possible and (economical or format 6 disallowed), * otherwise format 6 if allowed. */ if (merge == (byte)merge && (num_entries <= 127 || !can_use_trimmed)) write_cmap_0(s, entries, num_glyphs); else if (can_use_trimmed) write_cmap_6(s, entries, first_code, first_entry, num_entries); else # endif { /* * Punt. Acrobat Reader 3 can't handle any other Mac table format. * (AR3 for Linux doesn't seem to be able to handle Windows format, * either, but maybe AR3 for Windows can.) */ stream_write(s, cmap_initial_4, sizeof(cmap_initial_4)); } #endif /* Write the Windows sub-table. */ memcpy(cmap_sub, cmap_sub_initial, sizeof(cmap_sub_initial)); put_u16(cmap_sub + 2, U16(cmap_sub + 2) + num_entries * 2); /* length */ put_u16(cmap_sub + 14, first_code + end_entry - 1); /* endCount[0] */ put_u16(cmap_sub + 20, first_code + first_entry); /* startCount[0] */ stream_write(s, cmap_sub, sizeof(cmap_sub)); stream_write(s, entries + first_entry * 2, num_entries * 2); put_pad(s, cmap_length); } private uint size_cmap(gs_font *font, uint first_code, int num_glyphs, gs_glyph max_glyph, int options) { stream poss; s_init(&poss, NULL); swrite_position_only(&poss); write_cmap(&poss, font, first_code, num_glyphs, max_glyph, options, 0); return stell(&poss); } /* ------ hmtx/vmtx ------ */ /* * avoid fp exceptions storing large doubles into integers by limiting * the range of the results. these are band-aids that don't fix the * root cause of the out-of-range results, but they keep gs on the rails. */ ushort limdbl2ushort(double d) { if (d < 0) return 0; else if (d > 64000) return 64000; else return d; } long limdbl2long(double d) { if (d > 2e9) return 2e9; else if (d < -2e9) return -2e9; else return d; } private void write_mtx(stream *s, gs_font_type42 *pfont, const gs_type42_mtx_t *pmtx, int wmode) { uint num_metrics = pmtx->numMetrics; uint len = num_metrics * 4; double factor = pfont->data.unitsPerEm * (wmode ? -1 : 1); float sbw[4]; uint i; sbw[0] = sbw[1] = sbw[2] = sbw[3] = 0; /* in case of failures */ for (i = 0; i < pmtx->numMetrics; ++i) { DISCARD(pfont->data.get_metrics(pfont, i, wmode, sbw)); put_ushort(s, limdbl2ushort(sbw[wmode + 2] * factor)); /* width */ put_ushort(s, limdbl2ushort(sbw[wmode] * factor)); /* lsb, may be <0 */ } for (; len < pmtx->length; ++i, len += 2) { DISCARD(pfont->data.get_metrics(pfont, i, wmode, sbw)); put_ushort(s, limdbl2ushort(sbw[wmode] * factor)); /* lsb, may be <0 */ } } /* Compute the metrics from the glyph_info. */ private uint size_mtx(gs_font_type42 *pfont, gs_type42_mtx_t *pmtx, uint max_glyph, int wmode) { int prev_width = min_int, wmode2; uint last_width = 0; /* pacify compilers */ double factor = pfont->data.unitsPerEm * (wmode ? -1 : 1); uint i, j; for (i = 0; i <= max_glyph; ++i) { float sbw[4]; int code, width; for (j = 0; j < 4; j++) sbw[j] = 0; code = pfont->data.get_metrics(pfont, i, wmode, sbw); if (code < 0) continue; wmode2 = wmode + 2; if (wmode2 < 0 || wmode2 >= 4) abort(); /* "wmode2 out of range" */ width = limdbl2long(sbw[wmode2] * factor + 0.5); if (width != prev_width) prev_width = width, last_width = i; } pmtx->numMetrics = last_width + 1; pmtx->length = pmtx->numMetrics * 4 + (max_glyph - last_width) * 2; return pmtx->length; } /* ------ name ------ */ /* Write a generated name table. */ static const byte name_initial[] = { 0, 0, /* format */ 0, 1, /* # of records = 1 */ 0, 18, /* start of string storage */ 0, 2, /* platform ID = ISO */ 0, 2, /* encoding ID = ISO 8859-1 */ 0, 0, /* language ID (none) */ 0, 6, /* name ID = PostScript name */ 0, 0, /* length *VARIABLE* */ 0, 0 /* start of string within string storage */ }; private uint size_name(const gs_const_string *font_name) { return sizeof(name_initial) + font_name->size; } private void write_name(stream *s, const gs_const_string *font_name) { byte name_bytes[sizeof(name_initial)]; memcpy(name_bytes, name_initial, sizeof(name_initial)); put_u16(name_bytes + 14, font_name->size); stream_write(s, name_bytes, sizeof(name_bytes)); stream_write(s, font_name->data, font_name->size); put_pad(s, size_name(font_name)); } /* ------ OS/2 ------ */ /* Write a generated OS/2 table. */ #define OS_2_LENGTH sizeof(ttf_OS_2_t) private void update_OS_2(ttf_OS_2_t *pos2, uint first_glyph, int num_glyphs) { put_u16(pos2->usFirstCharIndex, first_glyph); put_u16(pos2->usLastCharIndex, first_glyph + num_glyphs - 1); #if TT_ADJUST_OS_2 if (first_glyph >= 0xf000) { /* This font is being treated as a symbolic font. */ memset(pos2->ulUnicodeRanges, 0, sizeof(pos2->ulUnicodeRanges)); pos2->ulUnicodeRanges[7] = 8; /* bit 60, private use range */ memset(pos2->ulCodePageRanges, 0, sizeof(pos2->ulCodePageRanges)); pos2->ulCodePageRanges[3] = 1; /* bit 31, symbolic */ } #endif } private void write_OS_2(stream *s, gs_font *font, uint first_glyph, int num_glyphs) { ttf_OS_2_t os2; /* * We don't bother to set most of the fields. The really important * ones, which affect character mapping, are usFirst/LastCharIndex. * We also need to set usWeightClass and usWidthClass to avoid * crashing ttfdump. */ memset(&os2, 0, sizeof(os2)); put_u16(os2.version, 1); put_u16(os2.usWeightClass, 400); /* Normal */ put_u16(os2.usWidthClass, 5); /* Normal */ update_OS_2(&os2, first_glyph, num_glyphs); stream_write(s, &os2, sizeof(os2)); put_pad(s, sizeof(os2)); } /* ------ post ------ */ /* Construct and then write the post table. */ typedef struct post_glyph_s { byte char_index; byte size; ushort glyph_index; } post_glyph_t; private int compare_post_glyphs(const void *pg1, const void *pg2) { gs_glyph g1 = ((const post_glyph_t *)pg1)->glyph_index, g2 = ((const post_glyph_t *)pg2)->glyph_index; return (g1 < g2 ? -1 : g1 > g2 ? 1 : 0); } typedef struct post_s { post_glyph_t glyphs[256 + 1]; int count, glyph_count; uint length; } post_t; /* * If necessary, compute the length of the post table. Note that we * only generate post entries for characters in the Encoding. */ private void compute_post(gs_font *font, post_t *post) { int i; for (i = 0, post->length = 32 + 2; i <= 255; ++i) { gs_const_string str; gs_glyph glyph = font->procs.encode_char(font, (gs_char)i, GLYPH_SPACE_INDEX); int mac_index = mac_glyph_index(font, i, &str); if (mac_index != 0) { post->glyphs[post->count].char_index = i; post->glyphs[post->count].size = (mac_index < 0 ? str.size + 1 : 0); post->glyphs[post->count].glyph_index = glyph - GS_MIN_GLYPH_INDEX; post->count++; } } if (post->count) { int j; qsort(post->glyphs, post->count, sizeof(post->glyphs[0]), compare_post_glyphs); /* Eliminate duplicate references to the same glyph. */ for (i = j = 0; i < post->count; ++i) { if (i == 0 || post->glyphs[i].glyph_index != post->glyphs[i - 1].glyph_index ) { post->length += post->glyphs[i].size; post->glyphs[j++] = post->glyphs[i]; } } post->count = j; post->glyph_count = post->glyphs[post->count - 1].glyph_index + 1; } post->length += post->glyph_count * 2; } /* Write the post table */ private void write_post(stream *s, gs_font *font, post_t *post) { byte post_initial[32 + 2]; uint name_index; uint glyph_index; int i; memset(post_initial, 0, 32); put_u32(post_initial, 0x00020000); put_u16(post_initial + 32, post->glyph_count); stream_write(s, post_initial, sizeof(post_initial)); /* Write the name index table. */ for (i = 0, name_index = 258, glyph_index = 0; i < post->count; ++i) { gs_const_string str; int ch = post->glyphs[i].char_index; int mac_index = mac_glyph_index(font, ch, &str); for (; glyph_index < post->glyphs[i].glyph_index; ++glyph_index) put_ushort(s, 0); glyph_index++; if (mac_index >= 0) put_ushort(s, mac_index); else { put_ushort(s, name_index); name_index++; } } /* Write the string names of the glyphs. */ for (i = 0; i < post->count; ++i) { gs_const_string str; int ch = post->glyphs[i].char_index; int mac_index = mac_glyph_index(font, ch, &str); if (mac_index < 0) { spputc(s, (byte)str.size); stream_write(s, str.data, str.size); } } put_pad(s, post->length); } /* ---------------- Main program ---------------- */ /* Write the definition of a TrueType font. */ private int compare_table_tags(const void *pt1, const void *pt2) { ulong t1 = u32(pt1), t2 = u32(pt2); return (t1 < t2 ? -1 : t1 > t2 ? 1 : 0); } private int psf_write_truetype_data(stream *s, gs_font_type42 *pfont, int options, psf_glyph_enum_t *penum, bool is_subset, const gs_const_string *alt_font_name) { gs_font *const font = (gs_font *)pfont; gs_const_string font_name; int (*string_proc)(gs_font_type42 *, ulong, uint, const byte **) = pfont->data.string_proc; const byte *OffsetTable; uint numTables_stored, numTables, numTables_out; #define MAX_NUM_TABLES 40 byte tables[MAX_NUM_TABLES * 16]; uint i; ulong offset; gs_glyph glyph, glyph_prev; ulong max_glyph; uint glyf_length, loca_length; ulong glyf_checksum = 0L; /****** NO CHECKSUM ******/ ulong loca_checksum[2] = {0L,0L}; ulong glyf_alignment = 0; uint numGlyphs = 0; /* original value from maxp */ byte head[56]; /* 0 mod 4 */ gs_type42_mtx_t mtx[2]; post_t post; ulong head_checksum, file_checksum = 0; int indexToLocFormat = 0; bool writing_cid = (options & WRITE_TRUETYPE_CID) != 0, writing_stripped = (options & WRITE_TRUETYPE_STRIPPED) != 0, generate_mtx = (options & WRITE_TRUETYPE_HVMTX) != 0, no_generate = writing_cid | writing_stripped, have_cmap = no_generate, have_name = !(options & WRITE_TRUETYPE_NAME), have_OS_2 = no_generate, have_post = no_generate; int have_hvhea[2]; uint cmap_length = 0; ulong OS_2_start = 0; uint OS_2_length = OS_2_LENGTH; int code; have_hvhea[0] = have_hvhea[1] = 0; if (alt_font_name) font_name = *alt_font_name; else font_name.data = font->font_name.chars, font_name.size = font->font_name.size; /* * Count the number of tables, including the eventual glyf and loca * (which may not actually be present in the font), and copy the * table directory. */ ACCESS(0, 12, OffsetTable); numTables_stored = U16(OffsetTable + 4); for (i = numTables = 0; i < numTables_stored; ++i) { const byte *tab; const byte *data; ulong start; uint length; if (numTables == MAX_NUM_TABLES) return_error(gs_error_limitcheck); ACCESS(12 + i * 16, 16, tab); start = u32(tab + 8); length = u32(tab + 12); /* Copy the table data now, since another ACCESS may invalidate it. */ memcpy(&tables[numTables * 16], tab, 16); #define W(a,b,c,d)\ ( ((a) << 24) + ((b) << 16) + ((c) << 8) + (d)) switch (u32(tab)) { case W('h','e','a','d'): if (length != 54) return_error(gs_error_invalidfont); ACCESS(start, length, data); memcpy(head, data, length); continue; case W('g','l','y','f'): /* synthesized */ case W('g','l','y','x'): /* Adobe bogus */ case W('l','o','c','a'): /* synthesized */ case W('l','o','c','x'): /* Adobe bogus */ case W('g','d','i','r'): /* Adobe marker */ continue; case W('c','m','a','p'): if (options & (WRITE_TRUETYPE_CMAP | WRITE_TRUETYPE_CID)) continue; have_cmap = true; break; case W('m','a','x','p'): ACCESS(start, length, data); numGlyphs = U16(data + 4); break; case W('n','a','m','e'): if (writing_cid) continue; have_name = true; break; case W('O','S','/','2'): if (writing_cid) continue; have_OS_2 = true; if (length > OS_2_LENGTH) return_error(gs_error_invalidfont); OS_2_start = start; OS_2_length = length; continue; case W('p','o','s','t'): have_post = true; break; case W('h','h','e','a'): have_hvhea[0] = 1; break; case W('v','h','e','a'): have_hvhea[1] = 1; break; case W('h','m','t','x'): case W('v','m','t','x'): if (generate_mtx) continue; /* falls through */ case W('c','v','t',' '): case W('f','p','g','m'): case W('g','a','s','p'): case W('k','e','r','n'): case W('p','r','e','p'): break; /* always copy these if present */ default: if (writing_cid) continue; break; } numTables++; } /* * Enumerate the glyphs to get the size of glyf and loca, * and to compute the checksums for these tables. */ /****** NO CHECKSUMS YET ******/ for (max_glyph = 0, glyf_length = 0; (code = psf_enumerate_glyphs_next(penum, &glyph)) != 1; ) { uint glyph_index; gs_glyph_data_t glyph_data; if (glyph < gs_min_cid_glyph) return_error(gs_error_invalidfont); glyph_index = glyph & ~GS_GLYPH_TAG; if_debug1('L', "[L]glyph_index %u\n", glyph_index); glyph_data.memory = pfont->memory; if ((code = pfont->data.get_outline(pfont, glyph_index, &glyph_data)) >= 0) { /* Since indexToLocFormat==0 assumes even glyph lengths, round it up here. If later we choose indexToLocFormat==1, subtract the glyf_alignment to compensate it. */ uint l = (glyph_data.bits.size + 1) & ~1; max_glyph = max(max_glyph, glyph_index); glyf_length += l; if (l != glyph_data.bits.size) glyf_alignment++; if_debug1('L', "[L] size %u\n", glyph_data.bits.size); gs_glyph_data_free(&glyph_data, "psf_write_truetype_data"); } } /* * For subset fonts, we should trim the loca table so that it only * contains entries through max_glyph. Unfortunately, this would * require changing numGlyphs in maxp, which in turn would affect hdmx, * hhea, hmtx, vdmx, vhea, vmtx, and possibly other tables. This is way * more work than we want to do right now. */ if (writing_stripped) { glyf_length = 0; loca_length = 0; } else { /*loca_length = (max_glyph + 2) << 2;*/ loca_length = (numGlyphs + 1) << 2; indexToLocFormat = (glyf_length > 0x1fffc); if (!indexToLocFormat) loca_length >>= 1; else glyf_length -= glyf_alignment; /* Acrobat Reader won't accept fonts with empty glyfs. */ if (glyf_length == 0) glyf_length = 1; } if_debug2('l', "[l]max_glyph = %lu, glyf_length = %lu\n", (ulong)max_glyph, (ulong)glyf_length); /* * If necessary, compute the length of the post table. Note that we * only generate post entries for characters in the Encoding. */ if (!have_post) { memset(&post, 0, sizeof(post)); if (options & WRITE_TRUETYPE_POST) compute_post(font, &post); else post.length = 32; /* dummy table */ } /* Fix up the head table. */ memset(head + 8, 0, 4); head[51] = (byte)indexToLocFormat; memset(head + 54, 0, 2); for (head_checksum = 0, i = 0; i < 56; i += 4) head_checksum += u32(&head[i]); /* * Construct the table directory, except for glyf, loca, head, OS/2, * and, if necessary, generated cmap, name, and post tables. * Note that the existing directory is already sorted by tag. */ numTables_out = numTables + 1 /* head */ + !writing_stripped * 2 /* glyf, loca */ + generate_mtx * (have_hvhea[0] + have_hvhea[1]) /* hmtx, vmtx */ + !have_OS_2 /* OS/2 */ + !have_cmap + !have_name + !have_post; if (numTables_out >= MAX_NUM_TABLES) return_error(gs_error_limitcheck); offset = 12 + numTables_out * 16; for (i = 0; i < numTables; ++i) { byte *tab = &tables[i * 16]; ulong length = u32(tab + 12); offset += round_up(length, 4); } /* Make the table directory entries for generated tables. */ { byte *tab = &tables[numTables * 16]; if (!writing_stripped) { offset = put_table(tab, "glyf", glyf_checksum, offset, glyf_length); tab += 16; offset = put_table(tab, "loca", loca_checksum[indexToLocFormat], offset, loca_length); tab += 16; } if (!have_cmap) { cmap_length = size_cmap(font, TT_BIAS, 256, GS_MIN_GLYPH_INDEX + max_glyph, options); offset = put_table(tab, "cmap", 0L /****** NO CHECKSUM ******/, offset, cmap_length); tab += 16; } if (!have_name) { offset = put_table(tab, "name", 0L /****** NO CHECKSUM ******/, offset, size_name(&font_name)); tab += 16; } if (!no_generate) { offset = put_table(tab, "OS/2", 0L /****** NO CHECKSUM ******/, offset, OS_2_length); tab += 16; } if (generate_mtx) for (i = 0; i < 2; ++i) if (have_hvhea[i]) { offset = put_table(tab, (i ? "vmtx" : "hmtx"), 0L /****** NO CHECKSUM ******/, offset, size_mtx(pfont, &mtx[i], max_glyph, i)); tab += 16; } if (!have_post) { offset = put_table(tab, "post", 0L /****** NO CHECKSUM ******/, offset, post.length); tab += 16; } /* * Note that the 'head' table must have length 54, even though * it occupies 56 bytes on the file. */ offset = put_table(tab, "head", head_checksum, offset, 54); tab += 16; } numTables = numTables_out; /* Write the font header. */ { static const byte version[4] = {0, 1, 0, 0}; stream_write(s, version, 4); } put_ushort(s, numTables); for (i = 0; 1 << i <= numTables; ++i) DO_NOTHING; --i; put_ushort(s, 16 << i); /* searchRange */ put_ushort(s, i); /* entrySelectors */ put_ushort(s, numTables * 16 - (16 << i)); /* rangeShift */ /* Write the table directory. */ qsort(tables, numTables, 16, compare_table_tags); offset = 12 + numTables * 16; for (i = 0; i < numTables; ++i) { const byte *tab = &tables[i * 16]; byte entry[16]; memcpy(entry, tab, 16); if (entry[8] < 0x40) { /* Not a generated table. */ uint length = u32(tab + 12); put_u32(entry + 8, offset); offset += round_up(length, 4); } else { entry[8] -= 0x40; } stream_write(s, entry, 16); } /* Write tables other than the ones we generate here. */ for (i = 0; i < numTables; ++i) { const byte *tab = &tables[i * 16]; if (tab[8] < 0x40) { ulong start = u32(tab + 8); uint length = u32(tab + 12); switch (u32(tab)) { case W('O','S','/','2'): if (!have_cmap) { /* * Adjust the first and last character indices in the OS/2 * table to reflect the values in the generated cmap. */ const byte *pos2; ttf_OS_2_t os2; ACCESS(OS_2_start, OS_2_length, pos2); memcpy(&os2, pos2, min(OS_2_length, sizeof(os2))); update_OS_2(&os2, TT_BIAS, 256); stream_write(s, &os2, OS_2_length); put_pad(s, OS_2_length); } else { /* Just copy the existing OS/2 table. */ write_range(s, pfont, OS_2_start, OS_2_length); put_pad(s, OS_2_length); } break; case W('h','h','e','a'): case W('v','h','e','a'): if (generate_mtx) { write_range(s, pfont, start, length - 2); /* 34 */ put_ushort(s, mtx[tab[0] == 'v'].numMetrics); break; } /* falls through */ default: write_range(s, pfont, start, length); } put_pad(s, length); } } if (!writing_stripped) { /* Write glyf. */ psf_enumerate_glyphs_reset(penum); for (offset = 0; psf_enumerate_glyphs_next(penum, &glyph) != 1; ) { gs_glyph_data_t glyph_data; glyph_data.memory = pfont->memory; if ((code = pfont->data.get_outline(pfont, glyph & ~GS_GLYPH_TAG, &glyph_data)) >= 0 ) { uint l = glyph_data.bits.size, zero = 0; if (!indexToLocFormat) l = (l + 1) & ~1; stream_write(s, glyph_data.bits.data, glyph_data.bits.size); if (glyph_data.bits.size < l) stream_write(s, &zero, 1); offset += l; if_debug2('L', "[L]glyf index = %u, size = %u\n", i, glyph_data.bits.size); gs_glyph_data_free(&glyph_data, "psf_write_truetype_data"); } } if_debug1('l', "[l]glyf final offset = %lu\n", offset); /* Add a dummy byte if necessary to make glyf non-empty. */ while (offset < glyf_length) stream_putc(s, 0), ++offset; put_pad(s, (uint)offset); /* Write loca. */ psf_enumerate_glyphs_reset(penum); glyph_prev = 0; for (offset = 0; psf_enumerate_glyphs_next(penum, &glyph) != 1; ) { gs_glyph_data_t glyph_data; uint glyph_index = glyph & ~GS_GLYPH_TAG; for (; glyph_prev <= glyph_index; ++glyph_prev) put_loca(s, offset, indexToLocFormat); glyph_data.memory = pfont->memory; if ((code = pfont->data.get_outline(pfont, glyph_index, &glyph_data)) >= 0 ) { uint l = glyph_data.bits.size; if (!indexToLocFormat) l = (l + 1) & ~1; offset += l; gs_glyph_data_free(&glyph_data, "psf_write_truetype_data"); } } /* Pad to numGlyphs + 1 entries (including the trailing entry). */ for (; glyph_prev <= numGlyphs; ++glyph_prev) put_loca(s, offset, indexToLocFormat); put_pad(s, loca_length); /* If necessary, write cmap, name, and OS/2. */ if (!have_cmap) write_cmap(s, font, TT_BIAS, 256, GS_MIN_GLYPH_INDEX + max_glyph, options, cmap_length); if (!have_name) write_name(s, &font_name); if (!have_OS_2) write_OS_2(s, font, TT_BIAS, 256); /* If necessary, write [hv]mtx. */ if (generate_mtx) for (i = 0; i < 2; ++i) if (have_hvhea[i]) { write_mtx(s, pfont, &mtx[i], i); put_pad(s, mtx[i].length); } /* If necessary, write post. */ if (!have_post) { if (options & WRITE_TRUETYPE_POST) write_post(s, font, &post); else { byte post_initial[32 + 2]; memset(post_initial, 0, 32); put_u32(post_initial, 0x00030000); stream_write(s, post_initial, 32); } } } /* Write head. */ /****** CHECKSUM WAS NEVER COMPUTED ******/ /* * The following nonsense is to avoid warnings about the constant * 0xb1b0afbaL being "unsigned in ANSI C, signed with -traditional". */ #if ARCH_SIZEOF_LONG > ARCH_SIZEOF_INT # define HEAD_MAGIC 0xb1b0afbaL #else # define HEAD_MAGIC ((ulong)~0x4e4f5045) #endif put_u32(head + 8, HEAD_MAGIC - file_checksum); /* per spec */ #undef HEAD_MAGIC stream_write(s, head, 56); return 0; } /* Write a TrueType font. */ int psf_write_truetype_font(stream *s, gs_font_type42 *pfont, int options, gs_glyph *orig_subset_glyphs, uint orig_subset_size, const gs_const_string *alt_font_name) { gs_font *const font = (gs_font *)pfont; psf_glyph_enum_t genum; gs_glyph subset_data[256 * MAX_COMPOSITE_PIECES]; gs_glyph *subset_glyphs = orig_subset_glyphs; uint subset_size = orig_subset_size; /* Sort the subset glyphs, if any. */ if (subset_glyphs) { /* Add the component glyphs for composites. */ int code; memcpy(subset_data, orig_subset_glyphs, sizeof(gs_glyph) * subset_size); subset_glyphs = subset_data; code = psf_add_subset_pieces(subset_glyphs, &subset_size, countof(subset_data), countof(subset_data), font); if (code < 0) return code; subset_size = psf_sort_glyphs(subset_glyphs, subset_size); } psf_enumerate_glyphs_begin(&genum, font, subset_glyphs, (subset_glyphs ? subset_size : 0), GLYPH_SPACE_INDEX); return psf_write_truetype_data(s, pfont, options & ~WRITE_TRUETYPE_CID, &genum, subset_glyphs != 0, alt_font_name); } /* Write a stripped TrueType font. */ int psf_write_truetype_stripped(stream *s, gs_font_type42 *pfont) { psf_glyph_enum_t genum; byte no_subset = 0; psf_enumerate_bits_begin(&genum, (gs_font *)pfont, &no_subset, 0, GLYPH_SPACE_INDEX); return psf_write_truetype_data(s, pfont, WRITE_TRUETYPE_STRIPPED, &genum, true, NULL); } /* Write a CIDFontType 2 font. */ int psf_write_cid2_font(stream *s, gs_font_cid2 *pfont, int options, const byte *subset_bits, uint subset_size, const gs_const_string *alt_font_name) { gs_font *const font = (gs_font *)pfont; psf_glyph_enum_t genum; psf_enumerate_bits_begin(&genum, font, subset_bits, (subset_bits ? subset_size : 0), GLYPH_SPACE_INDEX); return psf_write_truetype_data(s, (gs_font_type42 *)font, options | WRITE_TRUETYPE_CID, &genum, subset_bits != 0, alt_font_name); } /* Write a stripped CIDFontType 2 font. */ int psf_write_cid2_stripped(stream *s, gs_font_cid2 *pfont) { gs_font *const font = (gs_font *)pfont; psf_glyph_enum_t genum; byte no_subset = 0; psf_enumerate_bits_begin(&genum, font, &no_subset, 0, GLYPH_SPACE_INDEX); return psf_write_truetype_data(s, (gs_font_type42 *)font, WRITE_TRUETYPE_STRIPPED | WRITE_TRUETYPE_CID, &genum, true, NULL); }