/* Font backend for the Microsoft W32 Uniscribe API. Windows-specific parts of the HarfBuzz font backend. Copyright (C) 2008-2023 Free Software Foundation, Inc. This file is part of GNU Emacs. GNU Emacs is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. GNU Emacs 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 General Public License for more details. You should have received a copy of the GNU General Public License along with GNU Emacs. If not, see . */ #include /* Override API version - Uniscribe is only available as standard since Windows 2000, though most users of older systems will have it since it installs with Internet Explorer 5.0 and other software. Also, MinGW64 w32api headers by default define OPENTYPE_TAG typedef only if _WIN32_WINNT >= 0x0600. We only use the affected APIs if they are available, so there is no chance of calling non-existent functions. */ #undef _WIN32_WINNT #define _WIN32_WINNT 0x0600 #include #include #ifdef HAVE_HARFBUZZ # include # include /* for hb_ot_font_set_funcs */ # if GNUC_PREREQ (4, 3, 0) # define bswap_32(v) __builtin_bswap32(v) # else # include # endif #endif #include "lisp.h" #include "w32term.h" #include "frame.h" #include "composite.h" #include "font.h" #include "w32font.h" #include "pdumper.h" #include "w32common.h" /* Extension of w32font_info used by Uniscribe and HarfBuzz backends. */ struct uniscribe_font_info { struct w32font_info w32_font; /* This is used by the Uniscribe backend as a pointer to the script cache, and by the HarfBuzz backend as a pointer to a hb_font_t object. */ void *cache; /* This is used by the HarfBuzz backend to store the font scale. */ double scale; }; int uniscribe_available = 0; /* EnumFontFamiliesEx callback. */ static int CALLBACK ALIGN_STACK add_opentype_font_name_to_list (ENUMLOGFONTEX *, NEWTEXTMETRICEX *, DWORD, LPARAM); #ifdef HAVE_HARFBUZZ struct font_driver harfbuzz_font_driver; int harfbuzz_available = 0; /* Typedefs for HarfBuzz functions which we call through function pointers initialized after we load the HarfBuzz DLL. */ DEF_DLL_FN (hb_blob_t *, hb_blob_create, (const char *, unsigned int, hb_memory_mode_t, void *, hb_destroy_func_t)); DEF_DLL_FN (hb_face_t *, hb_face_create_for_tables, (hb_reference_table_func_t, void *, hb_destroy_func_t)); DEF_DLL_FN (unsigned, hb_face_get_glyph_count, (const hb_face_t *)); DEF_DLL_FN (hb_font_t *, hb_font_create, (hb_face_t *)); DEF_DLL_FN (void, hb_font_destroy, (hb_font_t *)); DEF_DLL_FN (void, hb_face_destroy, (hb_face_t *)); DEF_DLL_FN (unsigned int, hb_face_get_upem, (hb_face_t *)); DEF_DLL_FN (hb_bool_t, hb_font_get_nominal_glyph, (hb_font_t *, hb_codepoint_t, hb_codepoint_t *)); DEF_DLL_FN (hb_bool_t, hb_font_get_variation_glyph, (hb_font_t *, hb_codepoint_t, hb_codepoint_t, hb_codepoint_t *)); DEF_DLL_FN (void, hb_ot_font_set_funcs, (hb_font_t *)); #define hb_blob_create fn_hb_blob_create #define hb_face_create_for_tables fn_hb_face_create_for_tables #define hb_face_get_glyph_count fn_hb_face_get_glyph_count #define hb_font_create fn_hb_font_create #define hb_font_destroy fn_hb_font_destroy #define hb_face_destroy fn_hb_face_destroy #define hb_face_get_upem fn_hb_face_get_upem #define hb_font_get_nominal_glyph fn_hb_font_get_nominal_glyph #define hb_font_get_variation_glyph fn_hb_font_get_variation_glyph #define hb_ot_font_set_funcs fn_hb_ot_font_set_funcs #endif /* Used by uniscribe_otf_capability. */ static Lisp_Object otf_features (HDC context, const char *table); static int memq_no_quit (Lisp_Object elt, Lisp_Object list) { while (CONSP (list) && ! EQ (XCAR (list), elt)) list = XCDR (list); return (CONSP (list)); } /* Font backend interface implementation. */ static Lisp_Object uniscribe_list (struct frame *f, Lisp_Object font_spec) { Lisp_Object fonts = w32font_list_internal (f, font_spec, true); FONT_ADD_LOG ("uniscribe-list", font_spec, fonts); return fonts; } static Lisp_Object uniscribe_match (struct frame *f, Lisp_Object font_spec) { Lisp_Object entity = w32font_match_internal (f, font_spec, true); FONT_ADD_LOG ("uniscribe-match", font_spec, entity); return entity; } static Lisp_Object uniscribe_list_family (struct frame *f) { Lisp_Object list = Qnil; LOGFONT font_match_pattern; HDC dc; memset (&font_match_pattern, 0, sizeof (font_match_pattern)); /* Limit enumerated fonts to outline fonts to save time. */ font_match_pattern.lfOutPrecision = OUT_OUTLINE_PRECIS; /* Prevent quitting while EnumFontFamiliesEx runs and conses the list it will return. That's because get_frame_dc acquires the critical section, so we cannot quit before we release it in release_frame_dc. */ Lisp_Object prev_quit = Vinhibit_quit; Vinhibit_quit = Qt; dc = get_frame_dc (f); EnumFontFamiliesEx (dc, &font_match_pattern, (FONTENUMPROC) add_opentype_font_name_to_list, (LPARAM) &list, 0); release_frame_dc (f, dc); Vinhibit_quit = prev_quit; return list; } static Lisp_Object uniscribe_open (struct frame *f, Lisp_Object font_entity, int pixel_size) { Lisp_Object font_object = font_make_object (VECSIZE (struct uniscribe_font_info), font_entity, pixel_size); struct uniscribe_font_info *uniscribe_font = (struct uniscribe_font_info *) XFONT_OBJECT (font_object); if (!NILP (AREF (font_entity, FONT_TYPE_INDEX))) ASET (font_object, FONT_TYPE_INDEX, AREF (font_entity, FONT_TYPE_INDEX)); else /* paranoia: this should never happen */ ASET (font_object, FONT_TYPE_INDEX, Quniscribe); if (!w32font_open_internal (f, font_entity, pixel_size, font_object)) { return Qnil; } /* Initialize the cache for this font. */ uniscribe_font->cache = NULL; /* Uniscribe and HarfBuzz backends use glyph indices. */ uniscribe_font->w32_font.glyph_idx = ETO_GLYPH_INDEX; #ifdef HAVE_HARFBUZZ if (EQ (AREF (font_object, FONT_TYPE_INDEX), Qharfbuzz)) uniscribe_font->w32_font.font.driver = &harfbuzz_font_driver; else #endif /* HAVE_HARFBUZZ */ uniscribe_font->w32_font.font.driver = &uniscribe_font_driver; return font_object; } static void uniscribe_close (struct font *font) { struct uniscribe_font_info *uniscribe_font = (struct uniscribe_font_info *) font; #ifdef HAVE_HARFBUZZ if (uniscribe_font->w32_font.font.driver == &harfbuzz_font_driver && uniscribe_font->cache) hb_font_destroy ((hb_font_t *) uniscribe_font->cache); else #endif if (uniscribe_font->cache) ScriptFreeCache ((SCRIPT_CACHE) &(uniscribe_font->cache)); uniscribe_font->cache = NULL; w32font_close (font); } /* Return a list describing which scripts/languages FONT supports by which GSUB/GPOS features of OpenType tables. Implementation note: otf_features called by this function uses GetFontData to access the font tables directly, instead of using ScriptGetFontScriptTags etc. APIs even if those are available. The reason is that font-get, which uses the result of this function, expects a cons cell (GSUB . GPOS) where the features are reported separately for these 2 OTF tables, while the Uniscribe APIs report the features as a single list. There doesn't seem to be a reason for returning the features in 2 separate parts, except for compatibility with libotf; the features are disjoint (each can appear only in one of the 2 slots), and no client of this data discerns between the two slots: the few that request this data all look in both slots. If use of the Uniscribe APIs ever becomes necessary here, and the 2 separate slots are still required, it should be possible to split the feature list the APIs return into 2 because each sub-list is alphabetically sorted, so the place where the sorting order breaks is where the GSUB features end and GPOS features begin. But for now, this is not necessary, so we leave the original code in place. */ static Lisp_Object uniscribe_otf_capability (struct font *font) { HDC context; HFONT old_font; struct frame *f; Lisp_Object capability = Fcons (Qnil, Qnil); Lisp_Object features; f = XFRAME (selected_frame); /* Prevent quitting while we cons the lists in otf_features. That's because get_frame_dc acquires the critical section, so we cannot quit before we release it in release_frame_dc. */ Lisp_Object prev_quit = Vinhibit_quit; Vinhibit_quit = Qt; context = get_frame_dc (f); old_font = SelectObject (context, FONT_HANDLE (font)); features = otf_features (context, "GSUB"); XSETCAR (capability, features); features = otf_features (context, "GPOS"); XSETCDR (capability, features); SelectObject (context, old_font); release_frame_dc (f, context); Vinhibit_quit = prev_quit; return capability; } /* Uniscribe implementation of shape for font backend. Shape text in LGSTRING. See the docstring of `composition-get-gstring' for the format of LGSTRING. If the (N+1)th element of LGSTRING is nil, input of shaping is from the 1st to (N)th elements. In each input glyph, FROM, TO, CHAR, and CODE are already set. DIRECTION is either L2R or R2L, or nil if unknown. During redisplay, this comes from applying the UBA, is passed from composition_reseat_it, and is used by the HarfBuzz shaper. This function updates all fields of the input glyphs. If the output glyphs (M) are more than the input glyphs (N), (N+1)th through (M)th elements of LGSTRING are updated possibly by making a new glyph object and storing it in LGSTRING. If (M) is greater than the length of LGSTRING, nil should be returned. In that case, this function is called again with a larger LGSTRING. */ static Lisp_Object uniscribe_shape (Lisp_Object lgstring, Lisp_Object direction) { struct font *font = CHECK_FONT_GET_OBJECT (LGSTRING_FONT (lgstring)); struct uniscribe_font_info *uniscribe_font = (struct uniscribe_font_info *) font; EMACS_UINT nchars; int nitems, max_items, i, max_glyphs, done_glyphs; wchar_t *chars; WORD *glyphs, *clusters; SCRIPT_ITEM *items; SCRIPT_VISATTR *attributes; int *advances; GOFFSET *offsets; ABC overall_metrics; HRESULT result; struct frame * f = NULL; HDC context = NULL; HFONT old_font = NULL; /* Get the chars from lgstring in a form we can use with uniscribe. */ max_glyphs = nchars = LGSTRING_GLYPH_LEN (lgstring); done_glyphs = 0; chars = (wchar_t *) alloca (nchars * sizeof (wchar_t)); /* FIXME: This loop assumes that characters in the input LGSTRING are all inside the BMP. Need to encode characters beyond the BMP as UTF-16. */ for (i = 0; i < nchars; i++) { /* lgstring can be bigger than the number of characters in it, in the case where more glyphs are required to display those characters. If that is the case, note the real number of characters. */ if (NILP (LGSTRING_GLYPH (lgstring, i))) nchars = i; else chars[i] = LGLYPH_CHAR (LGSTRING_GLYPH (lgstring, i)); } /* First we need to break up the glyph string into runs of glyphs that can be treated together. First try a single run. */ max_items = 2; items = xmalloc (sizeof (SCRIPT_ITEM) * max_items + 1); while ((result = ScriptItemize (chars, nchars, max_items, NULL, NULL, items, &nitems)) == E_OUTOFMEMORY) { /* If that wasn't enough, keep trying with one more run. */ max_items++; items = (SCRIPT_ITEM *) xrealloc (items, sizeof (SCRIPT_ITEM) * max_items + 1); } if (FAILED (result)) { xfree (items); return Qnil; } glyphs = alloca (max_glyphs * sizeof (WORD)); clusters = alloca (nchars * sizeof (WORD)); attributes = alloca (max_glyphs * sizeof (SCRIPT_VISATTR)); advances = alloca (max_glyphs * sizeof (int)); offsets = alloca (max_glyphs * sizeof (GOFFSET)); for (i = 0; i < nitems; i++) { int nglyphs, nchars_in_run; nchars_in_run = items[i+1].iCharPos - items[i].iCharPos; /* Force ScriptShape to generate glyphs in the same order as they are in the input LGSTRING, which is in the logical order. */ items[i].a.fLogicalOrder = 1; /* Context may be NULL here, in which case the cache should be used without needing to select the font. */ result = ScriptShape (context, (SCRIPT_CACHE) &(uniscribe_font->cache), chars + items[i].iCharPos, nchars_in_run, max_glyphs - done_glyphs, &(items[i].a), glyphs, clusters, attributes, &nglyphs); if (result == E_PENDING && !context) { /* This assumes the selected frame is on the same display as the one we are drawing. It would be better for the frame to be passed in. */ f = XFRAME (selected_frame); context = get_frame_dc (f); old_font = SelectObject (context, FONT_HANDLE (font)); result = ScriptShape (context, (SCRIPT_CACHE) &(uniscribe_font->cache), chars + items[i].iCharPos, nchars_in_run, max_glyphs - done_glyphs, &(items[i].a), glyphs, clusters, attributes, &nglyphs); } if (result == E_OUTOFMEMORY) { /* Need a bigger lgstring. */ lgstring = Qnil; break; } else if (FAILED (result)) { /* Can't shape this run - return results so far if any. */ break; } else if (items[i].a.fNoGlyphIndex) { /* Glyph indices not supported by this font (or OS), means we can't really do any meaningful shaping. */ break; } else { result = ScriptPlace (context, (SCRIPT_CACHE) &(uniscribe_font->cache), glyphs, nglyphs, attributes, &(items[i].a), advances, offsets, &overall_metrics); if (result == E_PENDING && !context) { /* Cache not complete... */ f = XFRAME (selected_frame); context = get_frame_dc (f); old_font = SelectObject (context, FONT_HANDLE (font)); result = ScriptPlace (context, (SCRIPT_CACHE) &(uniscribe_font->cache), glyphs, nglyphs, attributes, &(items[i].a), advances, offsets, &overall_metrics); } if (SUCCEEDED (result)) { int j, from, to, adj_offset = 0; int cluster_offset = 0; from = 0; to = from; for (j = 0; j < nglyphs; j++) { int lglyph_index = j + done_glyphs; Lisp_Object lglyph = LGSTRING_GLYPH (lgstring, lglyph_index); ABC char_metric; unsigned gl; if (NILP (lglyph)) { lglyph = LGLYPH_NEW (); LGSTRING_SET_GLYPH (lgstring, lglyph_index, lglyph); } /* Copy to a 32-bit data type to shut up the compiler warning in LGLYPH_SET_CODE about comparison being always false. */ gl = glyphs[j]; LGLYPH_SET_CODE (lglyph, gl); /* Detect clusters, for linking codes back to characters. */ if (attributes[j].fClusterStart) { while (from < nchars_in_run && clusters[from] < j) from++; if (from >= nchars_in_run) from = to = nchars_in_run - 1; else { int k; to = nchars_in_run - 1; for (k = from + 1; k < nchars_in_run; k++) { if (clusters[k] > j) { to = k - 1; break; } } } cluster_offset = 0; /* For RTL text, the Uniscribe shaper prepares the values in ADVANCES array for layout in reverse order, whereby "advance width" is applied to move the pen in reverse direction and _before_ drawing the glyph. Since we draw glyphs in their normal left-to-right order, we need to adjust the coordinates of each non-base glyph in a grapheme cluster via X-OFF component of the gstring's ADJUSTMENT sub-vector. This loop computes, for each grapheme cluster, the initial value of the adjustment for the base character, which is then updated for each successive glyph in the grapheme cluster. */ /* FIXME: Should we use DIRECTION here instead of what ScriptItemize guessed? */ if (items[i].a.fRTL) { int j1 = j; adj_offset = 0; while (j1 < nglyphs && !attributes[j1].fClusterStart) { adj_offset += advances[j1]; j1++; } } } int char_idx = items[i].iCharPos + from + cluster_offset; if (from + cluster_offset > to) char_idx = items[i].iCharPos + to; cluster_offset++; LGLYPH_SET_CHAR (lglyph, chars[char_idx]); LGLYPH_SET_FROM (lglyph, items[i].iCharPos + from); LGLYPH_SET_TO (lglyph, items[i].iCharPos + to); /* Metrics. */ LGLYPH_SET_WIDTH (lglyph, advances[j]); LGLYPH_SET_ASCENT (lglyph, font->ascent); LGLYPH_SET_DESCENT (lglyph, font->descent); result = ScriptGetGlyphABCWidth (context, (SCRIPT_CACHE) &(uniscribe_font->cache), glyphs[j], &char_metric); if (result == E_PENDING && !context) { /* Cache incomplete... */ f = XFRAME (selected_frame); context = get_frame_dc (f); old_font = SelectObject (context, FONT_HANDLE (font)); result = ScriptGetGlyphABCWidth (context, (SCRIPT_CACHE) &(uniscribe_font->cache), glyphs[j], &char_metric); } if (SUCCEEDED (result)) { int lbearing = char_metric.abcA; int rbearing = char_metric.abcA + char_metric.abcB; LGLYPH_SET_LBEARING (lglyph, lbearing); LGLYPH_SET_RBEARING (lglyph, rbearing); } else { LGLYPH_SET_LBEARING (lglyph, 0); LGLYPH_SET_RBEARING (lglyph, advances[j]); } if (offsets[j].du || offsets[j].dv /* For non-base glyphs of RTL grapheme clusters, adjust the X offset even if both DU and DV are zero. */ || (!attributes[j].fClusterStart && items[i].a.fRTL)) { Lisp_Object vec = make_uninit_vector (3); if (items[i].a.fRTL) { /* Empirically, it looks like Uniscribe interprets DU in reverse direction for RTL clusters. E.g., if we don't reverse the direction, the Hebrew point HOLAM is drawn above the right edge of the base consonant, instead of above the left edge. */ ASET (vec, 0, make_fixnum (-offsets[j].du + adj_offset)); /* Update the adjustment value for the width advance of the glyph we just emitted. */ adj_offset -= 2 * advances[j]; } else ASET (vec, 0, make_fixnum (offsets[j].du + adj_offset)); /* In the font definition coordinate system, the Y coordinate points up, while in our screen coordinates Y grows downwards. So we need to reverse the sign of Y-OFFSET here. */ ASET (vec, 1, make_fixnum (-offsets[j].dv)); /* Based on what ftfont.c does... */ ASET (vec, 2, make_fixnum (advances[j])); LGLYPH_SET_ADJUSTMENT (lglyph, vec); } else { LGLYPH_SET_ADJUSTMENT (lglyph, Qnil); /* Update the adjustment value to compensate for the width of the base character. */ if (items[i].a.fRTL) adj_offset -= advances[j]; } } } } done_glyphs += nglyphs; } xfree (items); if (context) { SelectObject (context, old_font); release_frame_dc (f, context); } if (NILP (lgstring)) return Qnil; else return make_fixnum (done_glyphs); } /* Uniscribe implementation of encode_char for font backend. Return a glyph code of FONT for character C (Unicode code point). If FONT doesn't have such a glyph, return FONT_INVALID_CODE. */ static unsigned uniscribe_encode_char (struct font *font, int c) { HDC context = NULL; struct frame *f = NULL; HFONT old_font = NULL; unsigned code = FONT_INVALID_CODE; wchar_t ch[2]; int len; SCRIPT_ITEM* items; int nitems; struct uniscribe_font_info *uniscribe_font = (struct uniscribe_font_info *)font; if (c < 0x10000) { ch[0] = (wchar_t) c; len = 1; } else { DWORD surrogate = c - 0x10000; /* High surrogate: U+D800 - U+DBFF. */ ch[0] = 0xD800 + ((surrogate >> 10) & 0x03FF); /* Low surrogate: U+DC00 - U+DFFF. */ ch[1] = 0xDC00 + (surrogate & 0x03FF); len = 2; } /* Non BMP characters must be handled by the uniscribe shaping engine as GDI functions (except blindly displaying lines of Unicode text) and the promising looking ScriptGetCMap do not convert surrogate pairs to glyph indexes correctly. */ { items = (SCRIPT_ITEM *) alloca (sizeof (SCRIPT_ITEM) * 2 + 1); if (SUCCEEDED (ScriptItemize (ch, len, 2, NULL, NULL, items, &nitems))) { HRESULT result; /* Surrogates seem to need 2 here, even though only one glyph is returned. Indic characters can also produce 2 or more glyphs for a single code point, but they need to use uniscribe_shape above for correct display. */ WORD glyphs[2], clusters[2]; SCRIPT_VISATTR attrs[2]; int nglyphs; /* Force ScriptShape to generate glyphs in the logical order. */ items[0].a.fLogicalOrder = 1; result = ScriptShape (context, (SCRIPT_CACHE) &(uniscribe_font->cache), ch, len, 2, &(items[0].a), glyphs, clusters, attrs, &nglyphs); if (result == E_PENDING) { /* Use selected frame until API is updated to pass the frame. */ f = XFRAME (selected_frame); context = get_frame_dc (f); old_font = SelectObject (context, FONT_HANDLE (font)); result = ScriptShape (context, (SCRIPT_CACHE) &(uniscribe_font->cache), ch, len, 2, &(items[0].a), glyphs, clusters, attrs, &nglyphs); } if (SUCCEEDED (result) && nglyphs == 1) { /* Some fonts return .notdef glyphs instead of failing. (TrueType spec reserves glyph code 0 for .notdef) */ if (glyphs[0]) code = glyphs[0]; } else if (SUCCEEDED (result) || result == E_OUTOFMEMORY) { /* This character produces zero or more than one glyph when shaped. But we still need the return from here to be valid for the shaping engine to be invoked later. */ result = ScriptGetCMap (context, (SCRIPT_CACHE) &(uniscribe_font->cache), ch, len, 0, glyphs); if (SUCCEEDED (result) && glyphs[0]) code = glyphs[0]; } } } if (context) { SelectObject (context, old_font); release_frame_dc (f, context); } return code; } /* Shared with w32font: Lisp_Object uniscribe_get_cache (Lisp_Object frame); void uniscribe_free_entity (Lisp_Object font_entity); int uniscribe_has_char (Lisp_Object entity, int c); void uniscribe_text_extents (struct font *font, unsigned *code, int nglyphs, struct font_metrics *metrics); int uniscribe_draw (struct glyph_string *s, int from, int to, int x, int y, int with_background); Unused: int uniscribe_prepare_face (struct frame *f, struct face *face); void uniscribe_done_face (struct frame *f, struct face *face); int uniscribe_get_bitmap (struct font *font, unsigned code, struct font_bitmap *bitmap, int bits_per_pixel); void uniscribe_free_bitmap (struct font *font, struct font_bitmap *bitmap); int uniscribe_anchor_point (struct font *font, unsigned code, int index, int *x, int *y); int uniscribe_start_for_frame (struct frame *f); int uniscribe_end_for_frame (struct frame *f); */ /* Callback function for EnumFontFamiliesEx. Adds the name of opentype fonts to a Lisp list (passed in as the lParam arg). */ static int CALLBACK ALIGN_STACK add_opentype_font_name_to_list (ENUMLOGFONTEX *logical_font, NEWTEXTMETRICEX *physical_font, DWORD font_type, LPARAM list_object) { Lisp_Object* list = (Lisp_Object *) list_object; Lisp_Object family; /* Skip vertical fonts (intended only for printing) */ if (logical_font->elfLogFont.lfFaceName[0] == '@') return 1; /* Skip non opentype fonts. Count old truetype fonts as opentype, as some of them do contain GPOS and GSUB data that Uniscribe can make use of. */ if (!(physical_font->ntmTm.ntmFlags & NTMFLAGS_OPENTYPE) && font_type != TRUETYPE_FONTTYPE) return 1; /* Skip fonts that have no Unicode coverage. */ if (!physical_font->ntmFontSig.fsUsb[3] && !physical_font->ntmFontSig.fsUsb[2] && !physical_font->ntmFontSig.fsUsb[1] && !(physical_font->ntmFontSig.fsUsb[0] & 0x3fffffff)) return 1; family = intern_font_name (logical_font->elfLogFont.lfFaceName); if (! memq_no_quit (family, *list)) *list = Fcons (family, *list); return 1; } /* :otf property handling. Since the necessary Uniscribe APIs for getting font tag information are only available in Vista, we may need to parse the font data directly according to the OpenType Specification. */ /* Push into DWORD backwards to cope with endianness. */ #define OTF_TAG(STR) \ ((STR[3] << 24) | (STR[2] << 16) | (STR[1] << 8) | STR[0]) #define OTF_INT16_VAL(TABLE, OFFSET, PTR) \ do { \ BYTE temp, data[2]; \ if (GetFontData (context, TABLE, OFFSET, data, 2) != 2) \ goto font_table_error; \ temp = data[0], data[0] = data[1], data[1] = temp; \ memcpy (PTR, data, 2); \ } while (0) /* Do not reverse the bytes, because we will compare with a OTF_TAG value that has them reversed already. */ #define OTF_DWORDTAG_VAL(TABLE, OFFSET, PTR) \ do { \ if (GetFontData (context, TABLE, OFFSET, PTR, 4) != 4) \ goto font_table_error; \ } while (0) #define OTF_TAG_VAL(TABLE, OFFSET, STR) \ do { \ if (GetFontData (context, TABLE, OFFSET, STR, 4) != 4) \ goto font_table_error; \ STR[4] = '\0'; \ } while (0) #define SNAME(VAL) SSDATA (SYMBOL_NAME (VAL)) /* Uniscribe APIs available only since Windows Vista. */ typedef HRESULT (WINAPI *ScriptGetFontScriptTags_Proc) (HDC, SCRIPT_CACHE *, SCRIPT_ANALYSIS *, int, OPENTYPE_TAG *, int *); typedef HRESULT (WINAPI *ScriptGetFontLanguageTags_Proc) (HDC, SCRIPT_CACHE *, SCRIPT_ANALYSIS *, OPENTYPE_TAG, int, OPENTYPE_TAG *, int *); typedef HRESULT (WINAPI *ScriptGetFontFeatureTags_Proc) (HDC, SCRIPT_CACHE *, SCRIPT_ANALYSIS *, OPENTYPE_TAG, OPENTYPE_TAG, int, OPENTYPE_TAG *, int *); ScriptGetFontScriptTags_Proc script_get_font_scripts_fn; ScriptGetFontLanguageTags_Proc script_get_font_languages_fn; ScriptGetFontFeatureTags_Proc script_get_font_features_fn; static bool uniscribe_new_apis; /* Verify that all the required features in FEATURES, each of whose elements is a list or nil, can be found among the N feature tags in FTAGS. Return 'true' if the required features are supported, 'false' if not. Each list in FEATURES can include an element of nil, which means all the elements after it must not be in FTAGS. */ static bool uniscribe_check_features (Lisp_Object features[2], OPENTYPE_TAG *ftags, int n) { int j; for (j = 0; j < 2; j++) { bool negative = false; Lisp_Object rest; for (rest = features[j]; CONSP (rest); rest = XCDR (rest)) { Lisp_Object feature = XCAR (rest); /* The font must NOT have any of the features after nil. See the doc string of 'font-spec', under ':otf'. */ if (NILP (feature)) negative = true; else { OPENTYPE_TAG feature_tag = OTF_TAG (SNAME (feature)); int i; for (i = 0; i < n; i++) { if (ftags[i] == feature_tag) { /* Test fails if we find a feature that the font must NOT have. */ if (negative) return false; break; } } /* Test fails if we do NOT find a feature that the font should have. */ if (i >= n && !negative) return false; } } } return true; } /* Check if font supports the required OTF script/language/features using the Unsicribe APIs available since Windows Vista. We prefer these APIs as a kind of future-proofing Emacs: they seem to retrieve script tags that the old code (and also libotf) doesn't seem to be able to get, e.g., some fonts that claim support for "dev2" script don't show "deva", but the new APIs do report it. */ static int uniscribe_check_otf_1 (HDC context, Lisp_Object script, Lisp_Object lang, Lisp_Object features[2], int *retval) { SCRIPT_CACHE cache = NULL; OPENTYPE_TAG tags[32], script_tag, lang_tag; int max_tags = ARRAYELTS (tags); int ntags, i, ret = 0; HRESULT rslt; *retval = 0; rslt = script_get_font_scripts_fn (context, &cache, NULL, max_tags, tags, &ntags); if (FAILED (rslt)) { DebPrint (("ScriptGetFontScriptTags failed with 0x%x\n", rslt)); ret = -1; goto no_support; } if (NILP (script)) script_tag = OTF_TAG ("DFLT"); else script_tag = OTF_TAG (SNAME (script)); for (i = 0; i < ntags; i++) if (tags[i] == script_tag) break; if (i >= ntags) goto no_support; if (NILP (lang)) lang_tag = OTF_TAG ("dflt"); else { rslt = script_get_font_languages_fn (context, &cache, NULL, script_tag, max_tags, tags, &ntags); if (FAILED (rslt)) { DebPrint (("ScriptGetFontLanguageTags failed with 0x%x\n", rslt)); ret = -1; goto no_support; } if (ntags == 0) lang_tag = OTF_TAG ("dflt"); else { lang_tag = OTF_TAG (SNAME (lang)); for (i = 0; i < ntags; i++) if (tags[i] == lang_tag) break; if (i >= ntags) goto no_support; } } if (!NILP (features[0])) { /* Are the 2 feature lists valid? */ if (!CONSP (features[0]) || (!NILP (features[1]) && !CONSP (features[1]))) goto no_support; rslt = script_get_font_features_fn (context, &cache, NULL, script_tag, lang_tag, max_tags, tags, &ntags); if (FAILED (rslt)) { DebPrint (("ScriptGetFontFeatureTags failed with 0x%x\n", rslt)); ret = -1; goto no_support; } /* ScriptGetFontFeatureTags doesn't let us query features separately for GSUB and GPOS, so we check them all together. It doesn't really matter, since the features in GSUB and GPOS are disjoint, i.e. no feature can appear in both tables. */ if (!uniscribe_check_features (features, tags, ntags)) goto no_support; } ret = 1; *retval = 1; no_support: if (cache) ScriptFreeCache (&cache); return ret; } /* Check if font supports the otf script/language/features specified. OTF_SPEC is in the format (script lang [(gsub_feature ...)|nil] [(gpos_feature ...)]?) */ int uniscribe_check_otf (LOGFONT *font, Lisp_Object otf_spec) { Lisp_Object script, lang, rest; Lisp_Object features[2]; DWORD feature_tables[2]; DWORD script_tag, default_script, lang_tag = 0; struct frame * f; HDC context; HFONT check_font, old_font; int i, retval = 0; /* Check the spec is in the right format. */ if (!CONSP (otf_spec) || XFIXNUM (Flength (otf_spec)) < 3) return 0; /* Break otf_spec into its components. */ script = XCAR (otf_spec); rest = XCDR (otf_spec); lang = XCAR (rest); rest = XCDR (rest); features[0] = XCAR (rest); rest = XCDR (rest); if (NILP (rest)) features[1] = Qnil; else features[1] = XCAR (rest); /* Set up graphics context so we can use the font. */ f = XFRAME (selected_frame); context = get_frame_dc (f); check_font = CreateFontIndirect (font); old_font = SelectObject (context, check_font); /* If we are on Vista or later, use the new APIs. */ if (uniscribe_new_apis && !w32_disable_new_uniscribe_apis && uniscribe_check_otf_1 (context, script, lang, features, &retval) != -1) goto done; /* Set up tags we will use in the search. */ feature_tables[0] = OTF_TAG ("GSUB"); feature_tables[1] = OTF_TAG ("GPOS"); default_script = OTF_TAG ("DFLT"); if (NILP (script)) script_tag = default_script; else script_tag = OTF_TAG (SNAME (script)); if (!NILP (lang)) lang_tag = OTF_TAG (SNAME (lang)); /* Scan GSUB and GPOS tables. */ for (i = 0; i < 2; i++) { int j, n_match_features; unsigned short scriptlist_table, feature_table, n_scripts; unsigned short script_table, langsys_table, n_langs; unsigned short feature_index, n_features; DWORD tbl = feature_tables[i]; DWORD feature_id, *ftags; Lisp_Object farray[2]; /* Skip if no features requested from this table. */ if (NILP (features[i])) continue; /* If features is not a cons, this font spec is messed up. */ if (!CONSP (features[i])) goto no_support; /* Read GPOS/GSUB header. */ OTF_INT16_VAL (tbl, 4, &scriptlist_table); OTF_INT16_VAL (tbl, 6, &feature_table); OTF_INT16_VAL (tbl, scriptlist_table, &n_scripts); /* Find the appropriate script table. */ script_table = 0; for (j = 0; j < n_scripts; j++) { DWORD script_id; OTF_DWORDTAG_VAL (tbl, scriptlist_table + 2 + j * 6, &script_id); if (script_id == script_tag) { OTF_INT16_VAL (tbl, scriptlist_table + 6 + j * 6, &script_table); break; } #if 0 /* Causes false positives. */ /* If there is a DFLT script defined in the font, use it if the specified script is not found. */ else if (script_id == default_script) OTF_INT16_VAL (tbl, scriptlist_table + 6 + j * 6, &script_table); #endif } /* If no specific or default script table was found, then this font does not support the script. */ if (!script_table) goto no_support; /* Offset is from beginning of scriptlist_table. */ script_table += scriptlist_table; /* Get default langsys table. */ OTF_INT16_VAL (tbl, script_table, &langsys_table); /* If lang was specified, see if font contains a specific entry. */ if (!NILP (lang)) { OTF_INT16_VAL (tbl, script_table + 2, &n_langs); for (j = 0; j < n_langs; j++) { DWORD lang_id; OTF_DWORDTAG_VAL (tbl, script_table + 4 + j * 6, &lang_id); if (lang_id == lang_tag) { OTF_INT16_VAL (tbl, script_table + 8 + j * 6, &langsys_table); break; } } } if (!langsys_table) goto no_support; /* Offset is from beginning of script table. */ langsys_table += script_table; /* If there are no features to check, skip checking. */ if (NILP (features[i])) continue; if (!CONSP (features[i])) goto no_support; n_match_features = 0; /* First get required feature (if any). */ OTF_INT16_VAL (tbl, langsys_table + 2, &feature_index); if (feature_index != 0xFFFF) n_match_features = 1; OTF_INT16_VAL (tbl, langsys_table + 4, &n_features); n_match_features += n_features; USE_SAFE_ALLOCA; SAFE_NALLOCA (ftags, 1, n_match_features); int k = 0; if (feature_index != 0xFFFF) { OTF_DWORDTAG_VAL (tbl, feature_table + 2 + feature_index * 6, &feature_id); ftags[k++] = feature_id; } /* Now get all the other features. */ for (j = 0; j < n_features; j++) { OTF_INT16_VAL (tbl, langsys_table + 6 + j * 2, &feature_index); OTF_DWORDTAG_VAL (tbl, feature_table + 2 + feature_index * 6, &feature_id); ftags[k++] = feature_id; } /* Check the features for this table. */ farray[0] = features[i]; farray[1] = Qnil; if (!uniscribe_check_features (farray, ftags, n_match_features)) goto no_support; SAFE_FREE (); } retval = 1; done: no_support: font_table_error: /* restore graphics context. */ SelectObject (context, old_font); DeleteObject (check_font); release_frame_dc (f, context); return retval; } static Lisp_Object otf_features (HDC context, const char *table) { Lisp_Object script_list = Qnil; unsigned short scriptlist_table, n_scripts, feature_table; DWORD tbl = OTF_TAG (table); int i, j, k; /* Look for scripts in the table. */ OTF_INT16_VAL (tbl, 4, &scriptlist_table); OTF_INT16_VAL (tbl, 6, &feature_table); OTF_INT16_VAL (tbl, scriptlist_table, &n_scripts); for (i = n_scripts - 1; i >= 0; i--) { char script[5], lang[5]; unsigned short script_table, lang_count, langsys_table, feature_count; Lisp_Object script_tag, langsys_list, langsys_tag, feature_list; unsigned short record_offset = scriptlist_table + 2 + i * 6; OTF_TAG_VAL (tbl, record_offset, script); OTF_INT16_VAL (tbl, record_offset + 4, &script_table); /* Offset is from beginning of script table. */ script_table += scriptlist_table; script_tag = intern (script); langsys_list = Qnil; /* Optional default lang. */ OTF_INT16_VAL (tbl, script_table, &langsys_table); if (langsys_table) { /* Offset is from beginning of script table. */ langsys_table += script_table; langsys_tag = Qnil; feature_list = Qnil; OTF_INT16_VAL (tbl, langsys_table + 4, &feature_count); for (k = feature_count - 1; k >= 0; k--) { char feature[5]; unsigned short index; OTF_INT16_VAL (tbl, langsys_table + 6 + k * 2, &index); OTF_TAG_VAL (tbl, feature_table + 2 + index * 6, feature); feature_list = Fcons (intern (feature), feature_list); } langsys_list = Fcons (Fcons (langsys_tag, feature_list), langsys_list); } /* List of supported languages. */ OTF_INT16_VAL (tbl, script_table + 2, &lang_count); for (j = lang_count - 1; j >= 0; j--) { record_offset = script_table + 4 + j * 6; OTF_TAG_VAL (tbl, record_offset, lang); OTF_INT16_VAL (tbl, record_offset + 4, &langsys_table); /* Offset is from beginning of script table. */ langsys_table += script_table; langsys_tag = intern (lang); feature_list = Qnil; OTF_INT16_VAL (tbl, langsys_table + 4, &feature_count); for (k = feature_count - 1; k >= 0; k--) { char feature[5]; unsigned short index; OTF_INT16_VAL (tbl, langsys_table + 6 + k * 2, &index); OTF_TAG_VAL (tbl, feature_table + 2 + index * 6, feature); feature_list = Fcons (intern (feature), feature_list); } langsys_list = Fcons (Fcons (langsys_tag, feature_list), langsys_list); } script_list = Fcons (Fcons (script_tag, langsys_list), script_list); } return script_list; font_table_error: return Qnil; } #ifdef HAVE_HARFBUZZ /* W32 implementation of the 'list' method for HarfBuzz backend. */ static Lisp_Object w32hb_list (struct frame *f, Lisp_Object font_spec) { Lisp_Object fonts = w32font_list_internal (f, font_spec, true); FONT_ADD_LOG ("harfbuzz-list", font_spec, fonts); for (Lisp_Object tail = fonts; CONSP (tail); tail = XCDR (tail)) ASET (XCAR (tail), FONT_TYPE_INDEX, Qharfbuzz); return fonts; } /* W32 implementation of the 'match' method for HarfBuzz backend. */ static Lisp_Object w32hb_match (struct frame *f, Lisp_Object font_spec) { Lisp_Object entity = w32font_match_internal (f, font_spec, true); FONT_ADD_LOG ("harfbuzz-match", font_spec, entity); if (! NILP (entity)) ASET (entity, FONT_TYPE_INDEX, Qharfbuzz); return entity; } /* Callback function to free memory. We need this so we could pass it to HarfBuzz as the function to call to destroy objects for which we allocated data by calling our 'malloc' (as opposed to 'malloc' from the MS CRT, against which HarfBuzz was linked). */ static void free_cb (void *ptr) { free (ptr); } /* A function used as reference_table_func for HarfBuzz. It returns the data of a specified table of a font as a blob. */ static hb_blob_t * w32hb_get_font_table (hb_face_t *face, hb_tag_t tag, void *data) { struct frame *f = XFRAME (selected_frame); HDC context = get_frame_dc (f); HFONT old_font = SelectObject (context, (HFONT) data); char *font_data = NULL; DWORD font_data_size = 0, val; DWORD table = bswap_32 (tag); hb_blob_t *blob = NULL; val = GetFontData (context, table, 0, font_data, font_data_size); if (val != GDI_ERROR) { font_data_size = val; /* Don't call xmalloc, because it can signal an error, while we are inside a critical section established by get_frame_dc. */ font_data = malloc (font_data_size); if (font_data) { val = GetFontData (context, table, 0, font_data, font_data_size); if (val != GDI_ERROR) blob = hb_blob_create (font_data, font_data_size, HB_MEMORY_MODE_READONLY, font_data, free_cb); } } /* Restore graphics context. */ SelectObject (context, old_font); release_frame_dc (f, context); return blob; } /* Helper function used by the HarfBuzz implementations of the encode_char, has_char, and begin_hb_font methods. It creates an hb_font_t object for a given Emacs font. */ static hb_font_t * w32hb_get_font (struct font *font, double *scale) { hb_font_t *hb_font = NULL; HFONT font_handle = FONT_HANDLE (font); hb_face_t *hb_face = hb_face_create_for_tables (w32hb_get_font_table, font_handle, NULL); if (hb_face_get_glyph_count (hb_face) > 0) { hb_font = hb_font_create (hb_face); /* This is needed for HarfBuzz before 2.0.0; it is the default in later versions. */ hb_ot_font_set_funcs (hb_font); } struct uniscribe_font_info *uniscribe_font = (struct uniscribe_font_info *) font; unsigned upem = hb_face_get_upem (hb_face); eassert (upem > 0); /* https://support.microsoft.com/en-sg/help/74299/info-calculating-the-logical-height-and-point-size-of-a-font. */ LONG font_point_size = uniscribe_font->w32_font.metrics.tmHeight - uniscribe_font->w32_font.metrics.tmInternalLeading; /* https://docs.microsoft.com/en-us/typography/opentype/spec/ttch01, under "Converting FUnits to pixels". */ *scale = font_point_size * 1.0 / upem; hb_face_destroy (hb_face); /* FIXME: Can hb_font be non-NULL and yet invalid? Compare to get_empty? */ return hb_font; } /* W32 implementation of encode_char method for HarfBuzz backend. */ static unsigned w32hb_encode_char (struct font *font, int c) { struct uniscribe_font_info *uniscribe_font = (struct uniscribe_font_info *) font; eassert (uniscribe_font->w32_font.font.driver == &harfbuzz_font_driver); hb_font_t *hb_font = uniscribe_font->cache; /* First time we use this font with HarfBuzz, create the hb_font_t object and cache it. */ if (!hb_font) { double scale; hb_font = w32hb_get_font (font, &scale); if (!hb_font) return FONT_INVALID_CODE; uniscribe_font->cache = hb_font; eassert (scale > 0.0); uniscribe_font->scale = scale; } hb_codepoint_t glyph; if (hb_font_get_nominal_glyph (hb_font, c, &glyph)) return glyph; return FONT_INVALID_CODE; } /* W32 implementation of HarfBuzz begin_hb_font and end_hb_font methods. */ /* Return a HarfBuzz font object for FONT and store in POSITION_UNIT the scale factor to convert a hb_position_t value to the number of pixels. Return NULL if HarfBuzz font object is not available for FONT. */ static hb_font_t * w32hb_begin_font (struct font *font, double *position_unit) { struct uniscribe_font_info *uniscribe_font = (struct uniscribe_font_info *) font; eassert (uniscribe_font->w32_font.font.driver == &harfbuzz_font_driver); /* First time we use this font with HarfBuzz, create the hb_font_t object and cache it. */ if (!uniscribe_font->cache) { double scale; uniscribe_font->cache = w32hb_get_font (font, &scale); eassert (scale > 0.0); uniscribe_font->scale = scale; } *position_unit = uniscribe_font->scale; return (hb_font_t *) uniscribe_font->cache; } /* W32 implementation of get_variation_glyphs method for HarfBuzz. Return the number of variation glyphs of character C supported by FONT. VARIATIONS is an array of 256 elements. If the variation selector N (1..256) defines a glyph, that glyph code is stored in the (N-1)th element of VARIATIONS. */ static int w32hb_get_variation_glyphs (struct font *font, int c, unsigned variations[256]) { struct uniscribe_font_info *uniscribe_font = (struct uniscribe_font_info *) font; eassert (uniscribe_font->w32_font.font.driver == &harfbuzz_font_driver); /* First time we use this font with HarfBuzz, create the hb_font_t object and cache it. */ if (!uniscribe_font->cache) { double scale; uniscribe_font->cache = w32hb_get_font (font, &scale); eassert (scale > 0.0); uniscribe_font->scale = scale; } int i, n = 0; hb_font_t *hb_font = uniscribe_font->cache; for (i = 0; i < 16; i++) { if (hb_font_get_variation_glyph (hb_font, c, 0xFE00 + i, &variations[i])) n++; else variations[i] = 0; } for ( ; i < 256; i++) { if (hb_font_get_variation_glyph (hb_font, c, 0xE0100 + (i - 16), &variations[i])) n++; else variations[i] = 0; } return n; } #endif /* HAVE_HARFBUZZ */ #undef OTF_INT16_VAL #undef OTF_TAG_VAL #undef OTF_TAG struct font_driver uniscribe_font_driver = { LISPSYM_INITIALLY (Quniscribe), 0, /* case insensitive */ w32font_get_cache, uniscribe_list, uniscribe_match, uniscribe_list_family, NULL, /* free_entity */ uniscribe_open, uniscribe_close, NULL, /* prepare_face */ NULL, /* done_face */ w32font_has_char, uniscribe_encode_char, w32font_text_extents, w32font_draw, NULL, /* get_bitmap */ NULL, /* free_bitmap */ NULL, /* anchor_point */ uniscribe_otf_capability, /* Defined so (font-get FONTOBJ :otf) works. */ NULL, /* otf_drive - use shape instead. */ NULL, /* start_for_frame */ NULL, /* end_for_frame */ uniscribe_shape, NULL, /* check */ NULL, /* get_variation_glyphs */ NULL, /* filter_properties */ NULL, /* cached_font_ok */ }; /* Note that this should be called at every startup, not just when dumping, as it needs to test for the existence of the Uniscribe library. */ void syms_of_w32uniscribe (void); static void syms_of_w32uniscribe_for_pdumper (void); void syms_of_w32uniscribe (void) { pdumper_do_now_and_after_load (syms_of_w32uniscribe_for_pdumper); } #ifdef HAVE_HARFBUZZ static bool load_harfbuzz_funcs (HMODULE library) { LOAD_DLL_FN (library, hb_blob_create); LOAD_DLL_FN (library, hb_face_create_for_tables); LOAD_DLL_FN (library, hb_face_get_glyph_count); LOAD_DLL_FN (library, hb_font_create); LOAD_DLL_FN (library, hb_font_destroy); LOAD_DLL_FN (library, hb_face_get_upem); LOAD_DLL_FN (library, hb_face_destroy); LOAD_DLL_FN (library, hb_font_get_nominal_glyph); LOAD_DLL_FN (library, hb_font_get_variation_glyph); LOAD_DLL_FN (library, hb_ot_font_set_funcs); return hbfont_init_w32_funcs (library); } #endif /* HAVE_HARFBUZZ */ static void syms_of_w32uniscribe_for_pdumper (void) { /* Don't init Uniscribe and HarfBuzz when dumping */ if (!initialized) return; /* Don't register if Uniscribe is not available. */ HMODULE uniscribe = GetModuleHandle ("usp10"); if (!uniscribe) return; uniscribe_available = 1; register_font_driver (&uniscribe_font_driver, NULL); script_get_font_scripts_fn = (ScriptGetFontScriptTags_Proc) get_proc_addr (uniscribe, "ScriptGetFontScriptTags"); script_get_font_languages_fn = (ScriptGetFontLanguageTags_Proc) get_proc_addr (uniscribe, "ScriptGetFontLanguageTags"); script_get_font_features_fn = (ScriptGetFontFeatureTags_Proc) get_proc_addr (uniscribe, "ScriptGetFontFeatureTags"); if (script_get_font_scripts_fn && script_get_font_languages_fn && script_get_font_features_fn) uniscribe_new_apis = true; else uniscribe_new_apis = false; #ifdef HAVE_HARFBUZZ /* Currently, HarfBuzz DLLs are always named libharfbuzz-0.dll, as the project keeps the ABI backward-compatible. So we can hard-code the name of the library here, for now. If they ever break ABI compatibility, we may need to load the DLL that corresponds to the HarfBuzz version for which Emacs was built. */ HMODULE harfbuzz = LoadLibrary ("libharfbuzz-0.dll"); /* Don't register if HarfBuzz is not available. */ if (!harfbuzz) return; if (!load_harfbuzz_funcs (harfbuzz)) return; Fput (Quniscribe, Qfont_driver_superseded_by, Qharfbuzz); harfbuzz_available = 1; harfbuzz_font_driver = uniscribe_font_driver; harfbuzz_font_driver.type = Qharfbuzz; harfbuzz_font_driver.list = w32hb_list; harfbuzz_font_driver.match = w32hb_match; harfbuzz_font_driver.encode_char = w32hb_encode_char; harfbuzz_font_driver.otf_capability = hbfont_otf_capability; harfbuzz_font_driver.shape = hbfont_shape; harfbuzz_font_driver.get_variation_glyphs = w32hb_get_variation_glyphs; harfbuzz_font_driver.combining_capability = hbfont_combining_capability; harfbuzz_font_driver.begin_hb_font = w32hb_begin_font; register_font_driver (&harfbuzz_font_driver, NULL); #endif /* HAVE_HARFBUZZ */ }