Failure to run tests in JRuby due to number parsing error

[headius]

This may be a bug in JRuby's strscan or number parsing, but given the peculiar error and string input, I thought I'd file it here to see if you have any thoughts.

$ rake test
ArgumentError: invalid value for Integer(): "xref\n0 8\n0000000000 65535 f \n0000000015 00000 n \n0000000064 00000 n \n0000000149 00000 n \n0000000229 00000 n \n0000000329 00000 n \n0000000396 00000 n \n0000000476 00000 n \ntrailer << /Root 1 0 R /Info 7 0 R /Size 8 /ID [<6790ffa610024e78369114311fc0df96><6790ffa610024e78369114311fc0df96>] >>\nstartxref\n523\n%%EOF\n"
             scan_base10_integer at org/jruby/ext/strscan/RubyStringScanner.java:602
                    scan_integer at /Users/headius/work/jruby/lib/ruby/gems/shared/gems/strscan-3.1.3-java/lib/strscan/strscan.rb:18
                    parse_number at /Users/headius/work/hexapdf/lib/hexapdf/tokenizer.rb:290
                      next_token at /Users/headius/work/hexapdf/lib/hexapdf/tokenizer.rb:124
  parse_xref_section_and_trailer at /Users/headius/work/hexapdf/lib/hexapdf/parser.rb:287
                   load_revision at /Users/headius/work/hexapdf/lib/hexapdf/parser.rb:243
                         from_io at /Users/headius/work/hexapdf/lib/hexapdf/revisions.rb:81
                      initialize at /Users/headius/work/hexapdf/lib/hexapdf/document.rb:195
                             new at org/jruby/RubyClass.java:1026
                          <main> at /Users/headius/work/hexapdf/test/hexapdf/encryption/test_standard_security_handler.rb:70
                     module_eval at org/jruby/RubyModule.java:4174
                     module_eval at org/jruby/RubyModule.java:4205
                        describe at /Users/headius/work/jruby/lib/ruby/gems/shared/gems/minitest-5.25.5/lib/minitest/spec.rb:95
                          <main> at /Users/headius/work/hexapdf/test/hexapdf/encryption/test_standard_security_handler.rb:65
                         require at org/jruby/RubyKernel.java:1183
                         require at /Users/headius/work/jruby/lib/ruby/stdlib/rubygems/core_ext/kernel_require.rb:136
                          <main> at /Users/headius/work/jruby/lib/ruby/gems/shared/gems/rake-13.2.1/lib/rake/rake_test_loader.rb:21
                          select at org/jruby/RubyArray.java:2954
                          <main> at /Users/headius/work/jruby/lib/ruby/gems/shared/gems/rake-13.2.1/lib/rake/rake_test_loader.rb:6

I am assuming that this string is not what was intended to be parsed as an integer, but perhaps you can confirm that. Assuming I'm right, it may be a bug somewhere that is supposed to be splitting output, since there's clearly numbers there.

I'll look into this on my end for now, assuming it is a JRuby issue, but if you have any thoughts let me know.

[gettalong]

Ah, sorry, I just opened jruby/jruby#8823 over at JRuby's issue tracker as the notification e-mail for this issue came to late.

At lib/hexapdf/parser.rb:287 the "xref" part should already be scanned (done at line 281) and the call to #next_token should actually return 0, the first integer after "xref".

Another observation:

irb(main):001> require 'strscan'
=> true
irb(main):002> ss = StringScanner.new("xref\n0")
=> #<StringScanner 0/6 @ "xref\n...">
irb(main):003> ss.scan_integer
=> nil

This returns nil and doesn't raise an error, as expected.

[headius]

I'm guessing that it's something in the strscan impl for JRuby not advancing a position properly. Shouldn't be difficult to fix. Once I confirm it's in our code I'll close this.

[headius]

Yep, it was a silly mistake in the JRuby extension for strscan, apparently not tested by any of the suites. 🙄

Will be fixed by ruby/strscan#150 once we have a strscan release and resolved for JRuby in jruby/jruby#8823 once we've updated the stdlib gem.