lib/hexapdf/parser.rb in hexapdf-0.14.2 vs lib/hexapdf/parser.rb in hexapdf-0.14.3

@@ -70,22 +70,28 @@
     # HexaPDF::XRefSection::Entry.
     def load_object(xref_entry)
       obj, oid, gen, stream =
         case xref_entry.type
         when :in_use
-          parse_indirect_object(xref_entry.pos)
+          if xref_entry.pos == 0 && xref_entry.oid != 0
+            # Handle seen-in-the-wild objects with invalid offset 0
+            maybe_raise("Indirect object (#{xref_entry.oid},#{xref_entry.gen}) has offset 0", pos: 0)
+            [nil, xref_entry.oid, xref_entry.gen, nil]
+          else
+            parse_indirect_object(xref_entry.pos)
+          end
         when :free
           [nil, xref_entry.oid, xref_entry.gen, nil]
         when :compressed
           load_compressed_object(xref_entry)
         else
           raise_malformed("Invalid cross-reference type '#{xref_entry.type}' encountered")
         end
 
       if xref_entry.oid != 0 && (oid != xref_entry.oid || gen != xref_entry.gen)
         raise_malformed("The oid,gen (#{oid},#{gen}) values of the indirect object don't match " \
-          "the values (#{xref_entry.oid},#{xref_entry.gen}) from the xref")
+                        "the values (#{xref_entry.oid},#{xref_entry.gen}) from the xref")
       end
 
       @document.wrap(obj, oid: oid, gen: gen, stream: stream)
     rescue HexaPDF::MalformedPDFError
       reconstructed_revision.object(xref_entry)
@@ -131,11 +137,13 @@
           raise_malformed("A stream needs a dictionary, not a(n) #{object.class}", pos: offset)
         end
         tok1 = @tokenizer.next_byte
         tok2 = @tokenizer.next_byte if tok1 == 13 # 13=CR, 10=LF
         if tok1 != 10 && tok1 != 13
-          raise_malformed("Keyword stream must be followed by LF or CR/LF", pos: @tokenizer.pos)
+          tok2 = @tokenizer.next_byte
+          maybe_raise("Keyword stream must be followed by LF or CR/LF", pos: @tokenizer.pos,
+                      force: tok1 != 32 || (tok2 != 10 && tok2 != 13)) # 32=space
         elsif tok1 == 13 && tok2 != 10
           maybe_raise("Keyword stream must be followed by LF or CR/LF, not CR alone",
                       pos: @tokenizer.pos)
           @tokenizer.pos -= 1
         end
@@ -388,17 +396,17 @@
       xref = XRefSection.new
       @tokenizer.pos = 0
       while true
         @tokenizer.skip_whitespace
         pos = @tokenizer.pos
-        @tokenizer.scan_until(/(\n|\r\n?)+/)
+        @tokenizer.scan_until(/(\n|\r\n?)+|\z/)
         next_new_line_pos = @tokenizer.pos
         @tokenizer.pos = pos
 
-        token = @tokenizer.next_token rescue nil
+        token = @tokenizer.next_integer_or_keyword rescue nil
         if token.kind_of?(Integer)
-          gen = @tokenizer.next_token rescue nil
-          tok = @tokenizer.next_token rescue nil
+          gen = @tokenizer.next_integer_or_keyword rescue nil
+          tok = @tokenizer.next_integer_or_keyword rescue nil
           if @tokenizer.pos > next_new_line_pos
             @tokenizer.pos = next_new_line_pos
           elsif gen.kind_of?(Integer) && tok.kind_of?(Tokenizer::Token) && tok == 'obj'
             xref.add_in_use_entry(token, gen, pos)
             @tokenizer.scan_until(/(?:\n|\r\n?)endobj\b/)