KF8: Improved handling of startoffset

src/calibre/ebooks/mobi/reader/mobi8.py

@@ -109,7 +109,7 @@ class Mobi8Reader(object):
             table, cncx = read_index(self.kf8_sections, self.header.othidx,
                     self.header.codec)
             Item = namedtuple('Item',
-                'type title div_frag_num')
+                'type title pos_fid')
 
             for i, ref_type in enumerate(table.iterkeys()):
                 tag_map = table[ref_type]
@@ -119,7 +119,7 @@ class Mobi8Reader(object):
                 if 3 in tag_map.keys():
                     fileno  = tag_map[3][0]
                 if 6 in tag_map.keys():
-                    fileno = tag_map[6][0]
+                    fileno = tag_map[6]
                 self.guide.append(Item(ref_type.decode(self.header.codec),
                     title, fileno))
 
@@ -287,23 +287,24 @@ class Mobi8Reader(object):
 
     def create_guide(self):
         guide = Guide()
-        for ref_type, ref_title, fileno in self.guide:
+        has_start = False
+        for ref_type, ref_title, pos_fid in self.guide:
             try:
-                elem = self.elems[fileno]
-            except IndexError:
-                # Happens for thumbnailstandard in Amazon book samples
-                continue
-            fi = self.get_file_info(elem.insert_pos)
-            idtext = self.get_id_tag(elem.insert_pos).decode(self.header.codec)
-            linktgt = fi.filename
+                if len(pos_fid) != 2:
+                    continue
+            except TypeError:
+                continue # thumbnailstandard record, ignore it
+            linktgt, idtext = self.get_id_tag_by_pos_fid(*pos_fid)
             if idtext:
                 linktgt += b'#' + idtext
-            g = Guide.Reference('%s/%s'%(fi.type, linktgt), os.getcwdu())
+            g = Guide.Reference(linktgt, os.getcwdu())
             g.title, g.type = ref_title, ref_type
+            if g.title == 'start' or g.type == 'text':
+                has_start = True
             guide.append(g)
 
         so = self.header.exth.start_offset
-        if so not in {None, NULL_INDEX}:
+        if so not in {None, NULL_INDEX} and not has_start:
             fi = self.get_file_info(so)
             if fi.filename is not None:
                 idtext = self.get_id_tag(so).decode(self.header.codec)

src/calibre/ebooks/mobi/writer8/exth.py

@@ -153,9 +153,14 @@ def build_exth(metadata, prefer_author_sort=False, is_periodical=False,
         nrecs += 1
 
     if start_offset is not None:
-        exth.write(pack(b'>III', EXTH_CODES['startreading'], 12,
-            start_offset))
-        nrecs += 1
+        try:
+            len(start_offset)
+        except TypeError:
+            start_offset = [start_offset]
+        for so in start_offset:
+            exth.write(pack(b'>III', EXTH_CODES['startreading'], 12,
+                so))
+            nrecs += 1
 
     if num_of_resources is not None:
         exth.write(pack(b'>III', EXTH_CODES['num_of_resources'], 12,

src/calibre/ebooks/mobi/writer8/main.py

@@ -368,11 +368,11 @@ class KF8Writer(object):
             if aid is None:
                 continue
             pos, fid = self.aid_offset_map[aid]
-            if is_guide_ref_start(ref) and fid == 0:
-                # If fid != 0 then we cannot represent the start position as a
-                # single number in the EXTH header, so we do not write it to
-                # EXTH
-                self.start_offset = pos
+            if is_guide_ref_start(ref):
+                chunk = self.chunk_table[pos]
+                skel = [s for s in self.skel_table if s.file_number ==
+                        chunk.file_number][0]
+                self.start_offset = skel.start_pos + skel.length + chunk.start_pos + fid
             self.guide_table.append(GuideRef(ref.title or
                 _('Unknown'), ref.type, (pos, fid)))