Fix parsing error that occurs due to an invalid stream length

Currently, an error is raised when parsing a file that contains a stream
whose length value is invalid and where the supposed end is not followed
by the keyword 'endstream' but by an invalid PDF object.

We can infer from the invalid PDF object that the length value is not
valid and thus initiate the stream recovery algorithm.

CHANGELOG.md

@@ -5,6 +5,10 @@
 * [HexaPDF::TestUtils] module that contains helper methods useful for testing
  various parts of HexaPDF
 
+### Fixed
+
+* Parsing of streams with an invalid length value that led to a parsing error
+
 
 ## 0.24.0 - 2022-08-01
 

lib/hexapdf/parser.rb

@@ -172,7 +172,7 @@ def parse_indirect_object(offset = nil)
  end
  @tokenizer.pos = pos + length rescue pos
 
- tok = @tokenizer.next_token
+ tok = @tokenizer.next_token rescue nil
  unless tok.kind_of?(Tokenizer::Token) && tok == 'endstream'
  maybe_raise("Invalid stream length, keyword endstream not found", pos: @tokenizer.pos)
  @tokenizer.pos = pos

test/hexapdf/test_parser.rb

@@ -120,6 +120,13 @@ def create_parser(str)
  assert_equal('12', collector(stream.fiber))
  end
 
+ it "recovers from an incorrect stream length value which leads to a parsing error" do
+ create_parser("1 0 obj<</Length 2>> stream\n12(ab\nendstream endobj")
+ obj, _, _, stream = @parser.parse_indirect_object
+ assert_equal(5, obj[:Length])
+ assert_equal('12(ab', collector(stream.fiber))
+ end
+
  it "recovers from an invalid stream length value" do
  create_parser("1 0 obj<</Length 2 0 R>> stream\n12endstream endobj")
  @document.add([5], oid: 2)