fontselect, coretext: match whole extended family on fallback

Currently implemented only for coretext,
but other providers may/should add this later.

This improves fallback font choice by considering the whole extended family.

This also fixes remaining cases of Core Text fallback failure caused by
differences among the names recognized by Core Text, the name we choose
to return from get_fallback, and the names we seek in find_font.

This partly reverts commit 152d0484e98f118d01987138695cf40579c9e297.
This fixes https://github.com/libass/libass/issues/512.

4 files changed, 71 insertions, 51 deletions

diff --git a/libass/ass_coretext.c b/libass/ass_coretext.c
index f331639..4041046 100644
--- a/libass/ass_coretext.c
+++ b/libass/ass_coretext.c
@@ -129,21 +129,30 @@ static char *get_name(CTFontDescriptorRef fontd, CFStringRef attr)
 static bool fill_family_name(CTFontDescriptorRef fontd,
                              ASS_FontProviderMetaData *info)
 {
-    if (info->n_family)
-        return true;
-
     char *family_name = get_name(fontd, kCTFontFamilyNameAttribute);
     if (!family_name)
         return false;
 
-    info->families = malloc(sizeof(char *));
-    if (!info->families) {
-        free(family_name);
-        return false;
+    info->extended_family = family_name;
+
+    if (!info->n_family) {
+        family_name = strdup(family_name);
+        if (!family_name) {
+            free(info->extended_family);
+            return false;
+        }
+
+        info->families = malloc(sizeof(char *));
+        if (!info->families) {
+            free(family_name);
+            free(info->extended_family);
+            return false;
+        }
+
+        info->families[0] = family_name;
+        info->n_family++;
     }
 
-    info->families[0] = family_name;
-    info->n_family++;
     return true;
 }
 
@@ -196,6 +205,7 @@ static void process_descriptors(ASS_Library *lib, FT_Library ftlib,
         free(meta.fullnames);
 
         free(meta.postscript_name);
+        free(meta.extended_family);
 
         free(path);
     }
@@ -280,8 +290,6 @@ cleanup:
 static char *get_fallback(void *priv, ASS_Library *lib,
                           const char *family, uint32_t codepoint)
 {
-    FT_Library ftlib = priv;
-
     CFStringRef name = CFStringCreateWithBytes(
         0, (UInt8 *)family, strlen(family), kCFStringEncodingUTF8, false);
     if (!name)
@@ -308,33 +316,15 @@ static char *get_fallback(void *priv, ASS_Library *lib,
     if (!fb)
         return NULL;
 
-    CTFontDescriptorRef fontd = CTFontCopyFontDescriptor(fb);
+    CFStringRef cffamily = CTFontCopyFamilyName(fb);
     CFRelease(fb);
-    if (!fontd)
+    if (!cffamily)
         return NULL;
 
-    char *res_name = NULL;
-    char *path = NULL;
-    ASS_FontProviderMetaData meta = {0};
-    if (get_font_info_ct(lib, ftlib, fontd, &path, &meta))
-        res_name = meta.families[0];
-
-    CFRelease(fontd);
-
-    for (int i = 1 /* skip res_name */; i < meta.n_family; i++)
-        free(meta.families[i]);
-
-    for (int i = 0; i < meta.n_fullname; i++)
-        free(meta.fullnames[i]);
-
-    free(meta.families);
-    free(meta.fullnames);
-
-    free(meta.postscript_name);
-
-    free(path);
+    char *res_family = cfstr2buf(cffamily);
+    CFRelease(cffamily);
 
-    return res_name;
+    return res_family;
 }
 
 static void get_substitutions(void *priv, const char *name,
diff --git a/libass/ass_fontconfig.c b/libass/ass_fontconfig.c
index a4ffb42..e5065b5 100644
--- a/libass/ass_fontconfig.c
+++ b/libass/ass_fontconfig.c
@@ -87,7 +87,7 @@ static void scan_fonts(FcConfig *config, ASS_FontProvider *provider)
 {
     int i;
     FcFontSet *fonts;
-    ASS_FontProviderMetaData meta;
+    ASS_FontProviderMetaData meta = {0};
 
     // get list of fonts
     fonts = FcConfigGetFonts(config, FcSetSystem);
diff --git a/libass/ass_fontselect.c b/libass/ass_fontselect.c
index 3013f4a..ef79066 100644
--- a/libass/ass_fontselect.c
+++ b/libass/ass_fontselect.c
@@ -71,6 +71,8 @@ struct font_info {
     char *postscript_name; // can be used as an alternative to index to
                            // identify a font inside a collection
 
+    char *extended_family;
+
     // font source
     ASS_FontProvider *provider;
 
@@ -235,6 +237,8 @@ static void ass_font_provider_free_fontinfo(ASS_FontInfo *info)
     if (info->postscript_name)
         free(info->postscript_name);
 
+    if (info->extended_family)
+        free(info->extended_family);
 }
 
 /**
@@ -334,6 +338,12 @@ ass_font_provider_add_font(ASS_FontProvider *provider,
             goto error;
     }
 
+    if (meta->extended_family) {
+        info->extended_family = strdup(meta->extended_family);
+        if (info->extended_family == NULL)
+            goto error;
+    }
+
     if (path) {
         info->path = strdup(path);
         if (info->path == NULL)
@@ -423,12 +433,17 @@ static bool check_postscript(ASS_FontInfo *fi)
 /**
  * \brief Return whether the given font is in the given family.
  */
-static bool matches_family_name(ASS_FontInfo *f, const char *family)
+static bool matches_family_name(ASS_FontInfo *f, const char *family,
+                                bool match_extended_family)
 {
     for (int i = 0; i < f->n_family; i++) {
         if (ass_strcasecmp(f->families[i], family) == 0)
             return true;
     }
+    if (match_extended_family && f->extended_family) {
+        if (ass_strcasecmp(f->extended_family, family) == 0)
+            return true;
+    }
     return false;
 }
 
@@ -517,7 +532,8 @@ static bool check_glyph(ASS_FontInfo *fi, uint32_t code)
 
 static char *
 find_font(ASS_FontSelector *priv, ASS_Library *library,
-          ASS_FontProviderMetaData meta, unsigned bold, unsigned italic,
+          ASS_FontProviderMetaData meta, bool match_extended_family,
+          unsigned bold, unsigned italic,
           int *index, char **postscript_name, int *uid, ASS_FontStream *stream,
           uint32_t code, bool *name_match)
 {
@@ -542,7 +558,7 @@ find_font(ASS_FontSelector *priv, ASS_Library *library,
             ASS_FontInfo *font = &priv->font_infos[x];
             unsigned score = UINT_MAX;
 
-            if (matches_family_name(font, fullname)) {
+            if (matches_family_name(font, fullname, match_extended_family)) {
                 // If there's a family match, compare font attributes
                 // to determine best match in that particular family
                 score = font_attributes_similarity(font, &req);
@@ -620,7 +636,8 @@ find_font(ASS_FontSelector *priv, ASS_Library *library,
 }
 
 static char *select_font(ASS_FontSelector *priv, ASS_Library *library,
-                         const char *family, unsigned bold, unsigned italic,
+                         const char *family, bool match_extended_family,
+                         unsigned bold, unsigned italic,
                          int *index, char **postscript_name, int *uid,
                          ASS_FontStream *stream, uint32_t code)
 {
@@ -647,8 +664,9 @@ static char *select_font(ASS_FontSelector *priv, ASS_Library *library,
         meta = default_meta;
     }
 
-    result = find_font(priv, library, meta, bold, italic, index,
-                       postscript_name, uid, stream, code, &name_match);
+    result = find_font(priv, library, meta, match_extended_family,
+                       bold, italic, index, postscript_name, uid,
+                       stream, code, &name_match);
 
     // If no matching font was found, it might not exist in the font list
     // yet. Call the match_fonts callback to fill in the missing fonts
@@ -662,8 +680,9 @@ static char *select_font(ASS_FontSelector *priv, ASS_Library *library,
                                                 library, default_provider,
                                                 meta.fullnames[i]);
         }
-        result = find_font(priv, library, meta, bold, italic, index,
-                           postscript_name, uid, stream, code, &name_match);
+        result = find_font(priv, library, meta, match_extended_family,
+                           bold, italic, index, postscript_name, uid,
+                           stream, code, &name_match);
     }
 
     // cleanup
@@ -699,11 +718,11 @@ char *ass_font_select(ASS_FontSelector *priv, ASS_Library *library,
     ASS_FontProvider *default_provider = priv->default_provider;
 
     if (family && *family)
-        res = select_font(priv, library, family, bold, italic, index,
+        res = select_font(priv, library, family, false, bold, italic, index,
                 postscript_name, uid, data, code);
 
     if (!res && priv->family_default) {
-        res = select_font(priv, library, priv->family_default, bold,
+        res = select_font(priv, library, priv->family_default, false, bold,
                 italic, index, postscript_name, uid, data, code);
         if (res)
             ass_msg(library, MSGL_WARN, "fontselect: Using default "
@@ -720,7 +739,7 @@ char *ass_font_select(ASS_FontSelector *priv, ASS_Library *library,
                 default_provider->priv, library, search_family, code);
 
         if (fallback_family) {
-            res = select_font(priv, library, fallback_family, bold, italic,
+            res = select_font(priv, library, fallback_family, true, bold, italic,
                     index, postscript_name, uid, data, code);
             free(fallback_family);
         }
diff --git a/libass/ass_fontselect.h b/libass/ass_fontselect.h
index 942ea00..ce71b90 100644
--- a/libass/ass_fontselect.h
+++ b/libass/ass_fontselect.h
@@ -132,7 +132,7 @@ typedef void    (*SubstituteFontFunc)(void *priv, const char *name,
                                       ASS_FontProviderMetaData *meta);
 
 /**
- * Get an appropriate fallback font for a given codepoint.
+ * Get an appropriate fallback extended font family for a given codepoint.
  *
  * This is called by fontselect whenever a glyph is not found in the
  * physical font list of a logical font. fontselect will try to add the
@@ -147,7 +147,7 @@ typedef void    (*SubstituteFontFunc)(void *priv, const char *name,
  * \param lib ASS_Library instance
  * \param family original font family name (try matching a similar font) (never NULL)
  * \param codepoint Unicode codepoint (UTF-32)
- * \return output font family, allocated with malloc(), must be freed
+ * \return output extended font family, allocated with malloc(), must be freed
  *         by caller.
  */
 typedef char   *(*GetFallbackFunc)(void *priv,
@@ -173,21 +173,32 @@ typedef struct font_provider_funcs {
  */
 struct ass_font_provider_meta_data {
     /**
-     * List of localized font family names, e.g. "Arial".
+     * List of localized font family names,
+     * e.g. "Arial", "Arial Narrow" or "Arial Black".
      */
     char **families;
 
     /**
-     * List of localized full names, e.g. "Arial Bold".
+     * List of localized full names, e.g. "Arial Bold",
+     * "Arial Narrow Bold", "Arial Black" or "Arial Black Normal".
      * The English name should be listed first to speed up typical matching.
      */
     char **fullnames;
 
     /**
-     * The PostScript name, e.g. "Arial-BoldMT".
+     * The PostScript name, e.g. "Arial-BoldMT",
+     * "ArialNarrow-Bold" or "Arial-Black".
      */
     char *postscript_name;
 
+    /**
+     * Any name that identifies an extended font family, e.g. "Arial".
+     * This could be the full designer-named typographic family or (perhaps
+     * even better) a (sub)family limited to weight/width/slant variations.
+     * Names returned by get_fallback are matched against this field.
+     */
+    char *extended_family;
+
     int n_family;       // Number of localized family names
     int n_fullname;     // Number of localized full names