[subset] Don't assume FeatureList is sorted

Though the spec said FeatureRecords are sorted alphabetically by feature
tag, there're font files with unsorted FeatureList. And harfbuzz is not
able to subset these files correctly because we use binary search in
finding featureRecords when collecting lookups. Also
find_duplicate_features needs to be updated to handle this.

src/hb-ot-layout-gsubgpos.hh

@@ -3658,24 +3658,33 @@ struct GSUBGPOS
                                 const hb_set_t *feature_indices,
                                 hb_map_t *duplicate_feature_map /* OUT */) const
   {
-    hb_set_t unique_features;
-    hb_tag_t prev_t = get_feature_tag (feature_indices->get_min ());
+    if (feature_indices->is_empty ()) return;
+    hb_hashmap_t<hb_tag_t, hb_set_t *, (unsigned)-1, nullptr> unique_features;
     //find out duplicate features after subset
     for (unsigned i : feature_indices->iter ())
     {
       hb_tag_t t = get_feature_tag (i);
-      if (t != prev_t)
+      if (!unique_features.has (t))
       {
-        prev_t = t;
-        unique_features.clear ();
-        unique_features.add (i);
+        hb_set_t* indices = hb_set_create ();
+        if (unlikely (indices == hb_set_get_empty () ||
+                      !unique_features.set (t, indices)))
+        {
+          hb_set_destroy (indices);
+          for (auto _ : unique_features.iter ())
+            hb_set_destroy (_.second);
+          return;
+        }
+        if (unique_features.get (t))
+          unique_features.get (t)->add (i);
         duplicate_feature_map->set (i, i);
         continue;
       }
 
       bool found = false;
 
-      for (unsigned other_f_index : unique_features.iter ())
+      hb_set_t* same_tag_features = unique_features.get (t);
+      for (unsigned other_f_index : same_tag_features->iter ())
       {
         const Feature& f = get_feature (i);
         const Feature& other_f = get_feature (other_f_index);
@@ -3707,10 +3716,13 @@ struct GSUBGPOS
       
       if (found == false)
       {
-        unique_features.add (i);
+        same_tag_features->add (i);
         duplicate_feature_map->set (i, i);
       }
     }
+
+    for (auto _ : unique_features.iter ())
+      hb_set_destroy (_.second);
   }
 
   void prune_features (const hb_map_t *lookup_indices, /* IN */

src/hb-ot-layout.cc

@@ -1013,24 +1013,15 @@ struct hb_collect_features_context_t
     }
 
     has_feature_filter = true;
+    hb_set_t features_set;
     for (; *features; features++)
+      features_set.add (*features);
+    
+    for (unsigned i = 0; i < g.get_feature_count (); i++)
     {
-      hb_tag_t tag = *features;
-      unsigned index;
-      g.find_feature_index (tag, &index);
-      if (index == OT::Index::NOT_FOUND_INDEX) continue;
-
-      feature_indices_filter.add(index);
-      for (int i = (int) index - 1; i >= 0; i--)
-      {
-        if (g.get_feature_tag (i) != tag) break;
+      hb_tag_t tag = g.get_feature_tag (i);
+      if (features_set.has (tag))
         feature_indices_filter.add(i);
-      }
-      for (unsigned i = index + 1; i < g.get_feature_count (); i++)
-      {
-        if (g.get_feature_tag (i) != tag) break;
-        feature_indices_filter.add(i);
-      }
     }
   }
 

test/subset/data/Makefile.am

@@ -36,6 +36,7 @@ EXTRA_DIST += \
 	expected/layout.notonastaliqurdu \
 	expected/layout.tinos \
 	expected/layout.duplicate_features \
+	expected/layout.unsorted_featurelist \
 	expected/cmap \
 	expected/cmap14 \
 	expected/sbix \

test/subset/data/Makefile.sources

@@ -34,6 +34,7 @@ TESTS = \
 	tests/layout.tests \
 	tests/layout.tinos.tests \
 	tests/layout.duplicate_features.tests \
+	tests/layout.unsorted_featurelist.tests \
 	tests/sbix.tests \
 	tests/variable.tests \
 	tests/glyph_names.tests \

test/subset/data/tests/layout.unsorted_featurelist.tests

@@ -0,0 +1,11 @@
+FONTS:
+NotoIKEAHebrewLatin-Regular.ttf
+
+PROFILES:
+layout-test.txt
+default.txt
+retain-gids.txt
+
+SUBSETS:
+U+392,U+3a7,U+3b2,U+3c7
+*

test/subset/meson.build

@@ -28,6 +28,7 @@ tests = [
   'layout.notonastaliqurdu',
   'layout.tinos',
   'layout.duplicate_features',
+  'layout.unsorted_featurelist',
   'cmap',
   'cmap14',
   'sbix',