Environment

$ jruby -v
jruby 9.1.15.0 (2.3.3) 2017-12-07 929fde8 Java HotSpot(TM) 64-Bit Server VM 25.40-b25 on 1.8.0_40-b27 +jit [darwin-x86_64]
$ uname -a
Darwin Jankos-MacBook-Pro-2.local 17.4.0 Darwin Kernel Version 17.4.0: Sun Dec 17 09:19:54 PST 2017; root:xnu-4570.41.2~1/RELEASE_X86_64 x86_64

Expected Behavior

I'm using a trick in HTTP.rb where I utilize IO.copy_stream to yield chunks of data into a block. However, the tests for this behaviour are failing on JRuby.

I managed to isolate the problem in this script:

require "stringio"

class ProcIO
  def initialize(&block)
    @block = block
  end

  def write(data)
    @block.call(data)
    data.bytesize
  end
end

chunks = []

stringio = StringIO.new("a" * 16*1024 + "b" * 10*1024)
procio   = ProcIO.new { |data| chunks << data.dup }

IO.copy_stream(stringio, procio)

puts chunks

On MRI this prints 16*1024 characters a and 10*1024 characters b.

Actual Behavior

On JRuby this script prints 3 x 8*1024 characters b and 1 x 2*1024 characters b. This totals to the correct number of characters, but they're all b. Since JRuby seems to use chunk size of 8KB (unlike MRI which uses 16KB), I expected the output to be 2 x 8*1024 characters a, 1 x 8*1024 characters b and 1 x 2*1024 characters b.

I tried printing each chunk as it's yielded:

procio   = ProcIO.new { |data| chunks << p(data.dup) }

And the chunks were in fact correct. So it seems that, even though I duplicated the string, JRuby still somehow overrode it when reading subsequent chunks. And in a strange way too, I would have expected to get 4x the last chunk (2*1024 characters b), but instead JRuby kept the string length and still somehow replaced the content.