parent 777ba47b50

author ckitagawa <ckitagawa@chromium.org> 1579631743 -0500
committer ckitagawa <ckitagawa@chromium.org> 1580506176 -0500

[subset] Add CBLC support

src/Makefile.sources

@@ -63,6 +63,7 @@ HB_BASE_sources = \
 	hb-ot-cff2-table.cc \
 	hb-ot-cff2-table.hh \
 	hb-ot-cmap-table.hh \
+	hb-ot-color-cbdt-table.cc \
 	hb-ot-color-cbdt-table.hh \
 	hb-ot-color-colr-table.hh \
 	hb-ot-color-cpal-table.hh \
@@ -253,6 +254,7 @@ HB_SUBSET_sources = \
 	hb-number.hh \
 	hb-ot-cff1-table.cc \
 	hb-ot-cff2-table.cc \
+	hb-ot-color-cbdt-table.cc \
 	hb-static.cc \
 	hb-subset-cff-common.cc \
 	hb-subset-cff-common.hh \

src/harfbuzz.cc

@@ -12,6 +12,7 @@
 #include "hb-number.cc"
 #include "hb-ot-cff1-table.cc"
 #include "hb-ot-cff2-table.cc"
+#include "hb-ot-color-cbdt-table.cc"
 #include "hb-ot-color.cc"
 #include "hb-ot-face.cc"
 #include "hb-ot-font.cc"

src/hb-ot-cmap-table.hh

@@ -1198,10 +1198,9 @@ struct cmap
 		})
     ;
 
-
     if (unlikely (!encodingrec_iter.len ())) return_trace (false);
 
-    const EncodingRecord *unicode_bmp= nullptr, *unicode_ucs4 = nullptr, *ms_bmp = nullptr, *ms_ucs4 = nullptr;
+    const EncodingRecord *unicode_bmp= nullptr, *unicode_ucs4 = nullptr, *unicode_uvs = nullptr, *ms_bmp = nullptr, *ms_ucs4 = nullptr;
     bool has_format12 = false;
 
     for (const EncodingRecord& _ : encodingrec_iter)
@@ -1212,11 +1211,12 @@ struct cmap
       const EncodingRecord *table = hb_addressof (_);
       if      (_.platformID == 0 && _.encodingID ==  3) unicode_bmp = table;
       else if (_.platformID == 0 && _.encodingID ==  4) unicode_ucs4 = table;
+      else if (_.platformID == 0 && _.encodingID ==  5) unicode_uvs = table;
       else if (_.platformID == 3 && _.encodingID ==  1) ms_bmp = table;
       else if (_.platformID == 3 && _.encodingID == 10) ms_ucs4 = table;
     }
 
-    if (unlikely (!unicode_bmp && !ms_bmp)) return_trace (false);
+    if (unlikely (!unicode_bmp && !ms_bmp && !unicode_uvs)) return_trace (false);
     if (unlikely (has_format12 && (!unicode_ucs4 && !ms_ucs4))) return_trace (false);
 
     auto it =

src/hb-ot-color-cbdt-table.cc

@@ -0,0 +1,75 @@
+/*
+ * Copyright © 2020  Google, Inc.
+ *
+ *  This is part of HarfBuzz, a text shaping library.
+ *
+ * Permission is hereby granted, without written agreement and without
+ * license or royalty fees, to use, copy, modify, and distribute this
+ * software and its documentation for any purpose, provided that the
+ * above copyright notice and the following two paragraphs appear in
+ * all copies of this software.
+ *
+ * IN NO EVENT SHALL THE COPYRIGHT HOLDER BE LIABLE TO ANY PARTY FOR
+ * DIRECT, INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES
+ * ARISING OUT OF THE USE OF THIS SOFTWARE AND ITS DOCUMENTATION, EVEN
+ * IF THE COPYRIGHT HOLDER HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH
+ * DAMAGE.
+ *
+ * THE COPYRIGHT HOLDER SPECIFICALLY DISCLAIMS ANY WARRANTIES, INCLUDING,
+ * BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND
+ * FITNESS FOR A PARTICULAR PURPOSE.  THE SOFTWARE PROVIDED HEREUNDER IS
+ * ON AN "AS IS" BASIS, AND THE COPYRIGHT HOLDER HAS NO OBLIGATION TO
+ * PROVIDE MAINTENANCE, SUPPORT, UPDATES, ENHANCEMENTS, OR MODIFICATIONS.
+ *
+ * Google Author(s): Calder Kitagawa
+ */
+
+#include "hb-open-type.hh"
+
+#include "hb-ot-color-cbdt-table.hh"
+
+namespace OT {
+
+namespace CBDT_internal {
+
+bool copy_data_to_cbdt (hb_vector_t<char> *cbdt_prime,
+                      const void *data,
+                      unsigned int length)
+{
+  unsigned int new_len = cbdt_prime->length + length;
+  if (unlikely (!cbdt_prime->alloc(new_len))) return false;
+  memcpy (cbdt_prime->arrayZ + cbdt_prime->length, data, length);
+  cbdt_prime->length = new_len;
+  return true;
+}
+
+}
+
+bool CBLC::subset (hb_subset_context_t *c) const
+{
+  TRACE_SUBSET (this);
+
+  auto* cblc_prime = c->serializer->start_embed<CBLC> ();
+
+  // Use a vector as a secondary buffer as the tables need to be built in parallel.
+  hb_vector_t<char> cbdt_prime;
+
+  if (unlikely (!cblc_prime)) return_trace (false);
+  if (unlikely (!c->serializer->extend_min(cblc_prime))) return_trace (false);
+  cblc_prime->version = version;
+
+  hb_blob_t* cbdt_blob = hb_sanitize_context_t ().reference_table<CBDT> (c->plan->source);
+  unsigned int cbdt_length;
+  CBDT* cbdt = (CBDT *) hb_blob_get_data (cbdt_blob, &cbdt_length);
+  if (unlikely (cbdt_length < CBDT::min_size)) return_trace (false);
+  CBDT_internal::copy_data_to_cbdt (&cbdt_prime, cbdt, CBDT::min_size);
+
+  for (const BitmapSizeTable& table : + sizeTables.iter ())
+    subset_size_table (c, table, (const char *) cbdt, cblc_prime, &cbdt_prime);
+
+  hb_blob_destroy (cbdt_blob);
+
+  return_trace (CBLC::sink_cbdt (c, &cbdt_prime));
+}
+
+} /* namespace OT */

src/hb-ot-color-cbdt-table.hh

@@ -21,7 +21,7 @@
  * ON AN "AS IS" BASIS, AND THE COPYRIGHT HOLDER HAS NO OBLIGATION TO
  * PROVIDE MAINTENANCE, SUPPORT, UPDATES, ENHANCEMENTS, OR MODIFICATIONS.
  *
- * Google Author(s): Seigo Nonaka
+ * Google Author(s): Seigo Nonaka, Calder Kitagawa
  */
 
 #ifndef HB_OT_COLOR_CBDT_TABLE_HH
@@ -43,6 +43,31 @@
 
 namespace OT {
 
+namespace CBDT_internal {
+
+// Helper function internal to CBDT.
+HB_INTERNAL bool copy_data_to_cbdt (hb_vector_t<char> *cbdt_prime,
+                                    const void *data,
+                                    unsigned int length);
+
+}
+
+struct cblc_bitmap_size_subset_context_t
+{
+  const char *cbdt;
+  hb_vector_t<char> *cbdt_prime;
+  unsigned int size;          /* INOUT
+                               *  Input: old size of IndexSubtable
+                               *  Output: new size of IndexSubtable
+                               */
+  unsigned int num_tables;    /* INOUT
+                               *  Input: old number of subtables.
+                               *  Output: new number of subtables.
+                               */
+  hb_codepoint_t start_glyph; /* OUT */
+  hb_codepoint_t end_glyph;   /* OUT */
+};
+
 struct SmallGlyphMetrics
 {
   bool sanitize (hb_sanitize_context_t *c) const
@@ -143,6 +168,18 @@ struct IndexSubtableFormat1Or3
     return true;
   }
 
+  bool add_offset (hb_serialize_context_t *c,
+                   unsigned int offset,
+                   unsigned int *size /* OUT (accumulated) */)
+  {
+    TRACE_SERIALIZE (this);
+    Offset<OffsetType> embedded_offset;
+    embedded_offset = offset;
+    *size += sizeof(OffsetType);
+    auto* o = c->embed (embedded_offset);
+    return_trace ((bool) o);
+  }
+
   IndexSubtableHeader	header;
   UnsizedArrayOf<Offset<OffsetType>>
 			offsetArrayZ;
@@ -166,6 +203,99 @@ struct IndexSubtable
     }
   }
 
+  bool finish_subtable (hb_serialize_context_t *c,
+                        unsigned int cbdt_prime_len,
+                        unsigned int num_glyphs,
+                        unsigned int *size /* OUT (accumulated) */)
+  {
+    TRACE_SERIALIZE (this);
+
+    unsigned int local_offset = cbdt_prime_len - u.header.imageDataOffset;
+    switch (u.header.indexFormat) {
+    case 1: return_trace (u.format1.add_offset (c, local_offset, size));
+    case 3: {
+      if (!u.format3.add_offset (c, local_offset, size))
+        return_trace (false);
+      if (!(num_glyphs & 0x01))  // Pad to 32-bit alignment if needed.
+        return_trace (u.format3.add_offset (c, 0, size));
+      return_trace(true);
+    }
+    // TODO: implement 2, 4, 5.
+    case 2: case 4:  // No-op.
+    case 5:  // Pad to 32-bit aligned.
+    default: return_trace (false);
+    }
+  }
+
+  bool fill_missing_glyphs (hb_serialize_context_t *c,
+                            unsigned int cbdt_prime_len,
+                            unsigned int num_missing,
+                            unsigned int *size /* OUT (accumulated) */,
+                            unsigned int *num_glyphs /* OUT (accumulated) */)
+  {
+    TRACE_SERIALIZE (this);
+
+    unsigned int local_offset = cbdt_prime_len - u.header.imageDataOffset;
+    switch (u.header.indexFormat) {
+    case 1: {
+      for (unsigned int i = 0; i < num_missing; i++)
+      {
+        if (unlikely (!u.format1.add_offset (c, local_offset, size)))
+          return_trace (false);
+        *num_glyphs += 1;
+      }
+      return_trace (true);
+    }
+    case 3: {
+      for (unsigned int i = 0; i < num_missing; i++)
+      {
+        if (unlikely (!u.format3.add_offset (c, local_offset, size)))
+          return_trace (false);
+        *num_glyphs += 1;
+      }
+      return_trace (true);
+    }
+    // TODO: implement 2, 4, 5.
+    case 2:  // Add empty space in cbdt_prime?.
+    case 4: case 5:  // No-op as sparse is supported.
+    default: return_trace (false);
+    }
+  }
+
+  bool copy_glyph_at_idx (hb_serialize_context_t *c,
+                          unsigned int idx,
+                          const char *cbdt,
+                          hb_vector_t<char> *cbdt_prime /* INOUT */,
+                          IndexSubtable *subtable_prime /* INOUT */,
+                          unsigned int *size /* OUT (accumulated) */) const
+  {
+    TRACE_SERIALIZE (this);
+
+    unsigned int offset, length, format;
+    if (unlikely (!get_image_data (idx, &offset, &length, &format))) return_trace (false);
+
+    auto* header_prime = subtable_prime->get_header();
+    unsigned int new_local_offset = cbdt_prime->length - (unsigned int) header_prime->imageDataOffset;
+    if (unlikely (!CBDT_internal::copy_data_to_cbdt (cbdt_prime, cbdt + offset, length))) return_trace (false);
+
+    return_trace (subtable_prime->add_offset (c, new_local_offset, size));
+  }
+
+  bool add_offset (hb_serialize_context_t *c,
+                   unsigned int local_offset,
+                   unsigned int *size /* OUT (accumulated) */)
+  {
+    TRACE_SERIALIZE (this);
+    switch (u.header.indexFormat) {
+    case 1: return_trace (u.format1.add_offset (c, local_offset, size));
+    case 3: return_trace (u.format3.add_offset (c, local_offset, size));
+    // TODO: Implement tables 2, 4, 5
+    case 2:  // Should be a no-op.
+    case 4: case 5:  // Handle sparse cases.
+    default: return_trace (false);
+    }
+  }
+
   bool get_extents (hb_glyph_extents_t *extents HB_UNUSED) const
   {
     switch (u.header.indexFormat) {
@@ -188,6 +318,25 @@ struct IndexSubtable
     }
   }
 
+  const IndexSubtableHeader* get_header() const
+  {
+    return &u.header;
+  }
+
+  void populate_header (unsigned index_format,
+                        unsigned image_format,
+                        unsigned int image_data_offset,
+                        unsigned int *size)
+  {
+    u.header.indexFormat = index_format;
+    u.header.imageFormat = image_format;
+    u.header.imageDataOffset = image_data_offset;
+    switch (u.header.indexFormat) {
+    case 1: *size += IndexSubtableFormat1::min_size; break;
+    case 3: *size += IndexSubtableFormat3::min_size; break;
+    }
+  }
+
   protected:
   union {
   IndexSubtableHeader	header;
@@ -209,6 +358,127 @@ struct IndexSubtableRecord
 		  offsetToSubtable.sanitize (c, base, lastGlyphIndex - firstGlyphIndex + 1));
   }
 
+  const IndexSubtable* get_subtable (const void *base) const
+  {
+    return &(base+offsetToSubtable);
+  }
+
+  bool add_new_subtable (hb_subset_context_t* c,
+                         cblc_bitmap_size_subset_context_t *bitmap_size_context,
+                         IndexSubtableRecord *record,
+                         const hb_vector_t<hb_pair_t<hb_codepoint_t, const IndexSubtableRecord*>> *lookup, /* IN */
+                         const void *base,
+                         unsigned int *start /* INOUT */) const
+  {
+    TRACE_SERIALIZE (this);
+
+    auto* subtable = c->serializer->start_embed<IndexSubtable> ();
+    if (unlikely (!subtable)) return_trace (false);
+    if (unlikely (!c->serializer->extend_min (subtable))) return_trace (false);
+
+    auto* old_subtable = get_subtable (base);
+    auto* old_header = old_subtable->get_header ();
+
+    subtable->populate_header (old_header->indexFormat,
+                               old_header->imageFormat,
+                               bitmap_size_context->cbdt_prime->length,
+                               &bitmap_size_context->size);
+
+    unsigned int num_glyphs = 0;
+    bool early_exit = false;
+    for (unsigned int i = *start; i < lookup->length; i++)
+    {
+      hb_codepoint_t new_gid = (*lookup)[i].first;
+      const IndexSubtableRecord *next_record = (*lookup)[i].second;
+      const IndexSubtable *next_subtable = next_record->get_subtable (base);
+      auto* next_header = next_subtable->get_header ();
+      if (next_header != old_header) {
+        *start = i;
+        early_exit = true;
+        break;
+      }
+      unsigned int num_missing = record->add_glyph_for_subset (new_gid);
+      if (unlikely (!subtable->fill_missing_glyphs (c->serializer,
+                                                    bitmap_size_context->cbdt_prime->length,
+                                                    num_missing,
+                                                    &bitmap_size_context->size,
+                                                    &num_glyphs)))
+        return_trace (false);
+
+      hb_codepoint_t old_gid = 0;
+      c->plan->old_gid_for_new_gid (new_gid, &old_gid);
+      if (old_gid < next_record->firstGlyphIndex)
+        return_trace (false);
+
+      unsigned int old_idx = (unsigned int) old_gid - next_record->firstGlyphIndex;
+      if (unlikely (!next_subtable->copy_glyph_at_idx (c->serializer,
+                                                       old_idx,
+                                                       bitmap_size_context->cbdt,
+                                                       bitmap_size_context->cbdt_prime,
+                                                       subtable,
+                                                       &bitmap_size_context->size)))
+        return_trace (false);
+      num_glyphs += 1;
+    }
+    if (!early_exit)
+      *start = lookup->length;
+    if (unlikely (!subtable->finish_subtable (c->serializer,
+                                              bitmap_size_context->cbdt_prime->length,
+                                              num_glyphs,
+                                              &bitmap_size_context->size)))
+      return_trace (false);
+    return_trace (true);
+  }
+
+  bool add_new_record (hb_subset_context_t *c,
+                       cblc_bitmap_size_subset_context_t *bitmap_size_context,
+                       const hb_vector_t<hb_pair_t<hb_codepoint_t, const IndexSubtableRecord*>> *lookup, /* IN */
+                       const void *base,
+                       unsigned int *start, /* INOUT */
+                       hb_vector_t<IndexSubtableRecord>* records /* INOUT */) const
+  {
+    TRACE_SERIALIZE (this);
+    auto snap = c->serializer->snapshot ();
+    unsigned int old_size = bitmap_size_context->size;
+    unsigned int old_cbdt_prime_length = bitmap_size_context->cbdt_prime->length;
+
+    // Set to invalid state to indicate filling glyphs is not yet started.
+    records->resize (records->length + 1);
+    (*records)[records->length - 1].firstGlyphIndex = 1;
+    (*records)[records->length - 1].lastGlyphIndex = 0;
+    bitmap_size_context->size += IndexSubtableRecord::min_size;
+
+    c->serializer->push ();
+
+    if (unlikely (!add_new_subtable (c, bitmap_size_context, &((*records)[records->length - 1]), lookup, base, start)))
+    {
+      c->serializer->pop_discard ();
+      c->serializer->revert (snap);
+      bitmap_size_context->cbdt_prime->shrink (old_cbdt_prime_length);
+      bitmap_size_context->size = old_size;
+      records->resize (records->length - 1);
+      return_trace (false);
+    }
+
+    bitmap_size_context->num_tables += 1;
+    return_trace (true);
+  }
+
+  unsigned int add_glyph_for_subset (hb_codepoint_t gid) {
+    if (firstGlyphIndex > lastGlyphIndex)
+    {
+      firstGlyphIndex = gid;
+      lastGlyphIndex = gid;
+      return 0;
+    }
+    // TODO maybe assert? this shouldn't occur.
+    if (lastGlyphIndex > gid)
+      return 0;
+    unsigned int num_missing = (unsigned int) (gid - lastGlyphIndex - 1);
+    lastGlyphIndex = gid;
+    return num_missing;
+  }
+
   bool get_extents (hb_glyph_extents_t *extents,
 		    const void *base) const
   {
@@ -243,6 +513,70 @@ struct IndexSubtableArray
     return_trace (indexSubtablesZ.sanitize (c, count, this));
   }
 
+  void build_lookup (hb_subset_context_t *c,
+                     cblc_bitmap_size_subset_context_t *bitmap_size_context,
+                     hb_vector_t<hb_pair_t<hb_codepoint_t, const IndexSubtableRecord*>> *lookup /* OUT */) const
+  {
+    bool start_glyph_is_set = false;
+    for (hb_codepoint_t new_gid = 0; new_gid < c->plan->num_output_glyphs (); new_gid++)
+    {
+      hb_codepoint_t old_gid;
+      if (unlikely (!c->plan->old_gid_for_new_gid (new_gid, &old_gid))) continue;
+
+      const IndexSubtableRecord* record = find_table(old_gid, bitmap_size_context->num_tables);
+      if (unlikely (!record)) continue;
+
+      // Don't add gaps to the lookup. The best way to determine if a glyph is a
+      // gap is that it has no image data.
+      unsigned int offset, length, format;
+      if (unlikely (!record->get_image_data (old_gid, this, &offset, &length, &format))) continue;
+
+      lookup->push (hb_pair_t<hb_codepoint_t, const IndexSubtableRecord*> (new_gid, record));
+
+      if (!start_glyph_is_set)
+      {
+        bitmap_size_context->start_glyph = new_gid;
+        start_glyph_is_set = true;
+      }
+
+      bitmap_size_context->end_glyph = new_gid;
+    }
+  }
+
+  bool subset (hb_subset_context_t *c,
+               cblc_bitmap_size_subset_context_t *bitmap_size_context) const
+  {
+    TRACE_SUBSET (this);
+
+    auto* dst = c->serializer->start_embed<IndexSubtableArray> ();
+    if (unlikely (!dst)) return_trace (false);
+
+    hb_vector_t<hb_pair_t<hb_codepoint_t, const IndexSubtableRecord*>> lookup;
+    build_lookup (c, bitmap_size_context, &lookup);
+
+    bitmap_size_context->size = 0;
+    bitmap_size_context->num_tables = 0;
+    hb_vector_t<IndexSubtableRecord> records;
+    for (unsigned int start = 0; start < lookup.length;)
+    {
+      if (unlikely (!lookup[start].second->add_new_record (c, bitmap_size_context, &lookup, this, &start, &records)))
+        return_trace (false);
+    }
+
+    // Workaround to ensure offset ordering is from least to greatest when
+    // resolving links.
+    hb_vector_t<hb_serialize_context_t::objidx_t> objidxs;
+    for (unsigned int i = 0; i < records.length; i++)
+      objidxs.push (c->serializer->pop_pack());
+    for (unsigned int i = 0; i < records.length; i++)
+    {
+      IndexSubtableRecord* record = c->serializer->embed (records[i]);
+      if (unlikely (!record)) return_trace (false);
+      c->serializer->add_link (record->offsetToSubtable, objidxs[records.length - 1 - i], dst);
+    }
+    return_trace (true);
+  }
+
   public:
   const IndexSubtableRecord* find_table (hb_codepoint_t glyph, unsigned int numTables) const
   {
@@ -282,6 +616,42 @@ struct BitmapSizeTable
     return (base+indexSubtableArrayOffset).find_table (glyph, numberOfIndexSubtables);
   }
 
+  bool subset(hb_subset_context_t *c,
+              const void *src_base,
+              const void *dst_base,
+              const char *cbdt,
+              hb_vector_t<char> *cbdt_prime /* INOUT */) const
+  {
+    TRACE_SUBSET (this);
+    auto* out_table = c->serializer->embed (this);
+    if (unlikely (!out_table)) return_trace (false);
+
+    cblc_bitmap_size_subset_context_t bitmap_size_context;
+    bitmap_size_context.cbdt = cbdt;
+    bitmap_size_context.cbdt_prime = cbdt_prime;
+    bitmap_size_context.size = indexTablesSize;
+    bitmap_size_context.num_tables = numberOfIndexSubtables;
+    bitmap_size_context.start_glyph = 1;
+    bitmap_size_context.end_glyph = 0;
+
+    if (!out_table->indexSubtableArrayOffset.serialize_subset(c,
+                                                              indexSubtableArrayOffset,
+                                                              src_base,
+                                                              dst_base,
+                                                              &bitmap_size_context))
+      return_trace (false);
+    if (!bitmap_size_context.size ||
+        !bitmap_size_context.num_tables ||
+        bitmap_size_context.start_glyph > bitmap_size_context.end_glyph)
+      return_trace (false);
+
+    out_table->indexTablesSize = bitmap_size_context.size;
+    out_table->numberOfIndexSubtables = bitmap_size_context.num_tables;
+    out_table->startGlyphIndex = bitmap_size_context.start_glyph;
+    out_table->endGlyphIndex = bitmap_size_context.end_glyph;
+    return_trace (true);
+  }
+
   protected:
   LNNOffsetTo<IndexSubtableArray>
 			indexSubtableArrayOffset;
@@ -342,6 +712,45 @@ struct CBLC
 		  sizeTables.sanitize (c, this));
   }
 
+  static bool sink_cbdt(hb_subset_context_t *c,
+                        hb_vector_t<char>* cbdt_prime)
+  {
+    hb_blob_t *cbdt_prime_blob = hb_blob_create (cbdt_prime->arrayZ,
+                                                 cbdt_prime->length,
+                                                 HB_MEMORY_MODE_WRITABLE,
+                                                 cbdt_prime->arrayZ,
+                                                 free);
+    cbdt_prime->init ();  // Leak arrayZ to the blob.
+    bool ret = c->plan->add_table (HB_OT_TAG_CBDT, cbdt_prime_blob);
+    hb_blob_destroy (cbdt_prime_blob);
+    return ret;
+  }
+
+  bool subset_size_table (hb_subset_context_t *c,
+                          const BitmapSizeTable& table,
+                          const char *cbdt /* IN */,
+                          CBLC *cblc_prime /* INOUT */,
+                          hb_vector_t<char> *cbdt_prime /* INOUT */) const
+  {
+    TRACE_SUBSET (this);
+    cblc_prime->sizeTables.len++;
+
+    auto snap = c->serializer->snapshot ();
+    auto cbdt_prime_len = cbdt_prime->length;
+
+    if (!table.subset (c, this, cblc_prime, cbdt, cbdt_prime))
+    {
+      cblc_prime->sizeTables.len--;
+      c->serializer->revert (snap);
+      cbdt_prime->shrink (cbdt_prime_len);
+      return_trace (false);
+    }
+    return_trace (true);
+  }
+
+  // Implemented in cc file as it depends on definition of CBDT.
+  HB_INTERNAL bool subset (hb_subset_context_t *c) const;
+
   protected:
   const BitmapSizeTable &choose_strike (hb_font_t *font) const
   {

src/hb-ot-head-table.hh

@@ -54,6 +54,18 @@ struct head
     return 16 <= upem && upem <= 16384 ? upem : 1000;
   }
 
+  bool serialize (hb_serialize_context_t *c) const
+  {
+    TRACE_SERIALIZE (this);
+    return_trace ((bool) c->embed(this));
+  }
+
+  bool subset (hb_subset_context_t *c) const
+  {
+    TRACE_SUBSET (this);
+    return_trace (serialize(c->serializer));
+  }
+
   enum mac_style_flag_t {
     BOLD	= 1u<<0,
     ITALIC	= 1u<<1,

src/hb-subset.cc

@@ -45,6 +45,7 @@
 #include "hb-ot-cff2-table.hh"
 #include "hb-ot-vorg-table.hh"
 #include "hb-ot-name-table.hh"
+#include "hb-ot-color-cbdt-table.hh"
 #include "hb-ot-layout-gsub-table.hh"
 #include "hb-ot-layout-gpos-table.hh"
 #include "hb-ot-var-gvar-table.hh"
@@ -153,6 +154,53 @@ _subset (hb_subset_plan_t *plan)
 }
 
 
+static bool
+_is_table_present (hb_face_t *source,
+                   hb_tag_t tag)
+{
+  unsigned tables = hb_face_get_table_tags (source, 0, nullptr, nullptr);
+  hb_vector_t<uint32_t> tags;
+  tags.resize (tables);
+  hb_face_get_table_tags (source, 0, &tables, tags.arrayZ);
+  for (unsigned int i = 0; i < tables; i++) {
+    if (tags[i] == tag)
+      return true;
+  }
+  return false;
+}
+
+static bool
+_should_drop_table (hb_subset_plan_t *plan, hb_tag_t tag)
+{
+  if (plan->drop_tables->has (tag))
+    return true;
+
+  switch (tag) {
+    case HB_TAG ('c', 'v', 'a', 'r'): /* hint table, fallthrough */
+    case HB_TAG ('c', 'v', 't', ' '): /* hint table, fallthrough */
+    case HB_TAG ('f', 'p', 'g', 'm'): /* hint table, fallthrough */
+    case HB_TAG ('p', 'r', 'e', 'p'): /* hint table, fallthrough */
+    case HB_TAG ('h', 'd', 'm', 'x'): /* hint table, fallthrough */
+    case HB_TAG ('V', 'D', 'M', 'X'): /* hint table, fallthrough */
+      return plan->drop_hints;
+
+#ifdef HB_NO_SUBSET_LAYOUT
+    // Drop Layout Tables if requested.
+    case HB_OT_TAG_GDEF:
+    case HB_OT_TAG_GPOS:
+    case HB_OT_TAG_GSUB:
+    case HB_TAG ('m', 'o', 'r', 'x'):
+    case HB_TAG ('m', 'o', 'r', 't'):
+    case HB_TAG ('k', 'e', 'r', 'x'):
+    case HB_TAG ('k', 'e', 'r', 'n'):
+      return true;
+#endif
+
+    default:
+      return false;
+  }
+}
+
 static bool
 _subset_table (hb_subset_plan_t *plan,
 	       hb_tag_t          tag)
@@ -169,11 +217,15 @@ _subset_table (hb_subset_plan_t *plan,
     case HB_OT_TAG_name:
       result = _subset2<const OT::name> (plan);
       break;
-    case HB_OT_TAG_head:
+    case HB_OT_TAG_head: {
-      // TODO that won't work well if there is no glyf
+      if (_is_table_present (plan->source, HB_OT_TAG_glyf) && !_should_drop_table (plan, HB_OT_TAG_glyf))
+      {
         DEBUG_MSG(SUBSET, nullptr, "skip head, handled by glyf");
-      result = true;
+        return true;
+      }
+      result = _subset2<const OT::head> (plan);
       break;
+    }
     case HB_OT_TAG_hhea:
       DEBUG_MSG(SUBSET, nullptr, "skip hhea handled by hmtx");
       return true;
@@ -207,6 +259,12 @@ _subset_table (hb_subset_plan_t *plan,
     case HB_OT_TAG_COLR:
       result = _subset2<const OT::COLR> (plan);
       break;
+    case HB_OT_TAG_CBLC:
+      result = _subset2<const OT::CBLC> (plan);
+      break;
+    case HB_OT_TAG_CBDT:
+      DEBUG_MSG(SUBSET, nullptr, "skip CBDT handled by CBLC");
+      return true;
 
 #ifndef HB_NO_SUBSET_CFF
     case HB_OT_TAG_cff1:
@@ -254,38 +312,6 @@ _subset_table (hb_subset_plan_t *plan,
   return result;
 }
 
-static bool
-_should_drop_table (hb_subset_plan_t *plan, hb_tag_t tag)
-{
-  if (plan->drop_tables->has (tag))
-    return true;
-
-  switch (tag) {
-    case HB_TAG ('c', 'v', 'a', 'r'): /* hint table, fallthrough */
-    case HB_TAG ('c', 'v', 't', ' '): /* hint table, fallthrough */
-    case HB_TAG ('f', 'p', 'g', 'm'): /* hint table, fallthrough */
-    case HB_TAG ('p', 'r', 'e', 'p'): /* hint table, fallthrough */
-    case HB_TAG ('h', 'd', 'm', 'x'): /* hint table, fallthrough */
-    case HB_TAG ('V', 'D', 'M', 'X'): /* hint table, fallthrough */
-      return plan->drop_hints;
-
-#ifdef HB_NO_SUBSET_LAYOUT
-    // Drop Layout Tables if requested.
-    case HB_OT_TAG_GDEF:
-    case HB_OT_TAG_GPOS:
-    case HB_OT_TAG_GSUB:
-    case HB_TAG ('m', 'o', 'r', 'x'):
-    case HB_TAG ('m', 'o', 'r', 't'):
-    case HB_TAG ('k', 'e', 'r', 'x'):
-    case HB_TAG ('k', 'e', 'r', 'n'):
-      return true;
-#endif
-
-    default:
-      return false;
-  }
-}
-
 /**
  * hb_subset:
  * @source: font face data to be subset.

test/api/Makefile.am

@@ -59,6 +59,7 @@ TEST_PROGS = \
 	test-subset-sbix \
 	test-subset-gpos \
 	test-subset-colr \
+	test-subset-cbdt \
 	test-unicode \
 	test-version \
 	test-subset-nameids \
@@ -79,6 +80,7 @@ test_subset_gvar_LDADD = $(LDADD) $(top_builddir)/src/libharfbuzz-subset.la
 test_subset_hvar_LDADD = $(LDADD) $(top_builddir)/src/libharfbuzz-subset.la
 test_subset_vvar_LDADD = $(LDADD) $(top_builddir)/src/libharfbuzz-subset.la
 test_subset_sbix_LDADD = $(LDADD) $(top_builddir)/src/libharfbuzz-subset.la
+test_subset_cbdt_LDADD = $(LDADD) $(top_builddir)/src/libharfbuzz-subset.la
 test_subset_nameids_LDADD = $(LDADD) $(top_builddir)/src/libharfbuzz-subset.la
 test_subset_gpos_LDADD = $(LDADD) $(top_builddir)/src/libharfbuzz-subset.la
 test_subset_colr_LDADD = $(LDADD) $(top_builddir)/src/libharfbuzz-subset.la

test/api/test-subset-cbdt.c

@@ -0,0 +1,158 @@
+/*
+ * Copyright © 2020  Google, Inc.
+ *
+ *  This is part of HarfBuzz, a text shaping library.
+ *
+ * Permission is hereby granted, without written agreement and without
+ * license or royalty fees, to use, copy, modify, and distribute this
+ * software and its documentation for any purpose, provided that the
+ * above copyright notice and the following two paragraphs appear in
+ * all copies of this software.
+ *
+ * IN NO EVENT SHALL THE COPYRIGHT HOLDER BE LIABLE TO ANY PARTY FOR
+ * DIRECT, INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES
+ * ARISING OUT OF THE USE OF THIS SOFTWARE AND ITS DOCUMENTATION, EVEN
+ * IF THE COPYRIGHT HOLDER HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH
+ * DAMAGE.
+ *
+ * THE COPYRIGHT HOLDER SPECIFICALLY DISCLAIMS ANY WARRANTIES, INCLUDING,
+ * BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND
+ * FITNESS FOR A PARTICULAR PURPOSE.  THE SOFTWARE PROVIDED HEREUNDER IS
+ * ON AN "AS IS" BASIS, AND THE COPYRIGHT HOLDER HAS NO OBLIGATION TO
+ * PROVIDE MAINTENANCE, SUPPORT, UPDATES, ENHANCEMENTS, OR MODIFICATIONS.
+ *
+ * Google Author(s): Calder Kitagawa
+ */
+
+#include "hb-test.h"
+#include "hb-subset-test.h"
+
+/* Unit tests for CBDT/CBLC subsetting */
+
+static void
+test_subset_cbdt_noop (void)
+{
+  hb_face_t *face = hb_test_open_font_file ("fonts/NotoColorEmoji.subset.ttf");
+
+  hb_set_t *codepoints = hb_set_create ();
+  hb_face_t *face_subset;
+  hb_set_add (codepoints, 0x38);
+  hb_set_add (codepoints, 0x39);
+  hb_set_add (codepoints, 0xAE);
+  hb_set_add (codepoints, 0x2049);
+  hb_set_add (codepoints, 0x20E3);
+  face_subset = hb_subset_test_create_subset (face, hb_subset_test_create_input (codepoints));
+  hb_set_destroy (codepoints);
+
+  hb_subset_test_check (face, face_subset, HB_TAG ('C','B','L','C'));
+  hb_subset_test_check (face, face_subset, HB_TAG ('C','B','D','T'));
+
+  hb_face_destroy (face_subset);
+  hb_face_destroy (face);
+}
+
+static void
+test_subset_cbdt_keep_one (void)
+{
+  hb_face_t *face = hb_test_open_font_file ("fonts/NotoColorEmoji.subset.ttf");
+  hb_face_t *face_expected = hb_test_open_font_file ("fonts/NotoColorEmoji.subset.default.39.ttf");
+
+  hb_set_t *codepoints = hb_set_create ();
+  hb_face_t *face_subset;
+  hb_set_add (codepoints, 0x39);
+  face_subset = hb_subset_test_create_subset (face, hb_subset_test_create_input (codepoints));
+  hb_set_destroy (codepoints);
+
+  hb_subset_test_check (face_expected, face_subset, HB_TAG ('C','B','L','C'));
+  hb_subset_test_check (face_expected, face_subset, HB_TAG ('C','B','D','T'));
+
+  hb_face_destroy (face_subset);
+  hb_face_destroy (face_expected);
+  hb_face_destroy (face);
+}
+
+static void
+test_subset_cbdt_keep_one_last_subtable (void)
+{
+  hb_face_t *face = hb_test_open_font_file ("fonts/NotoColorEmoji.subset.ttf");
+  hb_face_t *face_expected = hb_test_open_font_file ("fonts/NotoColorEmoji.subset.default.2049.ttf");
+
+  hb_set_t *codepoints = hb_set_create ();
+  hb_face_t *face_subset;
+  hb_set_add (codepoints, 0x2049);
+  face_subset = hb_subset_test_create_subset (face, hb_subset_test_create_input (codepoints));
+  hb_set_destroy (codepoints);
+
+  hb_subset_test_check (face_expected, face_subset, HB_TAG ('C','B','L','C'));
+  hb_subset_test_check (face_expected, face_subset, HB_TAG ('C','B','D','T'));
+
+  hb_face_destroy (face_subset);
+  hb_face_destroy (face_expected);
+  hb_face_destroy (face);
+}
+
+static void
+test_subset_cbdt_keep_multiple_subtables (void)
+{
+  hb_face_t *face = hb_test_open_font_file ("fonts/NotoColorEmoji.subset.multiple_size_tables.ttf");
+  hb_face_t *face_expected = hb_test_open_font_file ("fonts/NotoColorEmoji.subset.multiple_size_tables.default.38,AE,2049.ttf");
+
+  hb_set_t *codepoints = hb_set_create ();
+  hb_face_t *face_subset;
+  hb_set_add (codepoints, 0x38);
+  hb_set_add (codepoints, 0xAE);
+  hb_set_add (codepoints, 0x2049);
+  face_subset = hb_subset_test_create_subset (face, hb_subset_test_create_input (codepoints));
+  hb_set_destroy (codepoints);
+
+  hb_subset_test_check (face_expected, face_subset, HB_TAG ('C','B','L','C'));
+  hb_subset_test_check (face_expected, face_subset, HB_TAG ('C','B','D','T'));
+
+  hb_face_destroy (face_subset);
+  hb_face_destroy (face_expected);
+  hb_face_destroy (face);
+}
+
+static void
+test_subset_cbdt_index_format_3 (void)
+{
+  hb_face_t *face = hb_test_open_font_file ("fonts/NotoColorEmoji.subset.index_format3.ttf");
+  hb_face_t *face_expected = hb_test_open_font_file ("fonts/NotoColorEmoji.subset.index_format3.default.38,AE,2049.ttf");
+
+  hb_set_t *codepoints = hb_set_create ();
+  hb_face_t *face_subset;
+  hb_set_add (codepoints, 0x38);
+  hb_set_add (codepoints, 0xAE);
+  hb_set_add (codepoints, 0x2049);
+  face_subset = hb_subset_test_create_subset (face, hb_subset_test_create_input (codepoints));
+  hb_set_destroy (codepoints);
+
+  hb_subset_test_check (face_expected, face_subset, HB_TAG ('C','B','L','C'));
+  hb_subset_test_check (face_expected, face_subset, HB_TAG ('C','B','D','T'));
+
+  hb_face_destroy (face_subset);
+  hb_face_destroy (face_expected);
+  hb_face_destroy (face);
+}
+
+// TODO: add support/tests for index formats 2,4,5 (image formats are treated as
+// opaque blobs when subsetting so don't need to be tested separately).
+// TODO: add a test that keeps no codepoints.
+
+int
+main (int argc, char **argv)
+{
+  hb_test_init (&argc, &argv);
+
+  hb_test_add (test_subset_cbdt_noop);
+  hb_test_add (test_subset_cbdt_keep_one);
+  hb_test_add (test_subset_cbdt_keep_one_last_subtable);
+  // The following use manually crafted expectation files as they are not
+  // binary compatible with FontTools.
+  hb_test_add (test_subset_cbdt_keep_multiple_subtables);
+  // Can use FontTools after https://github.com/fonttools/fonttools/issues/1817
+  // is resolved.
+  hb_test_add (test_subset_cbdt_index_format_3);
+
+  return hb_test_run();
+}

test/subset/data/Makefile.am

@@ -20,6 +20,7 @@ EXTRA_DIST += \
 	expected/cmap14 \
 	expected/sbix \
 	expected/colr \
+	expected/cbdt \
 	fonts \
 	profiles \
 	$(NULL)

test/subset/data/Makefile.sources

@@ -12,6 +12,7 @@ TESTS = \
 	tests/cmap14.tests \
 	tests/sbix.tests \
 	tests/colr.tests \
+	tests/cbdt.tests \
 	$(NULL)
 
 XFAIL_TESTS = \