Travis: jruby-9.1.15.0

[olleolleolle]

This PR updates the CI matrix to use latest JRuby.

http://jruby.org/2017/12/07/jruby-9-1-15-0.html

[olleolleolle]

In order to see if it was a regression, I moved down to jruby-9.1.14.0, and found the same failure on my macOS machine:

$ rspec --bisect
Bisect started using options: ""
Running suite to find failures... (11.5 seconds)
Starting bisect with 1 failing example and 101 non-failing examples.
Checking that failure(s) are order-dependent... failure(s) do not require any non-failures to run first

Bisect complete! Reduced necessary non-failing examples from 101 to 0 in 8.93 seconds.

The minimal reproduction command is:
  rspec './spec/nio/selectables/udp_socket_spec.rb[1:1:3]'

[tarcieri]

Huh, strange. It's persisting across a retry on Travis too. May have to add it to the tests we skip on JRuby

[tarcieri]

@headius @enebo don't suppose you have any idea why we'd start seeing IOError with UDP sockets, do you?

[ioquatix]

Does JRuby support UDP sockets?

[ioquatix]

https://github.com/jruby/jruby/pull/4684 was one issue I had.

[tarcieri]

Does JRuby support UDP sockets?

Yes, and these tests were previously passing on JRuby

[headius]

@tarcieri There have been some changes over the summer to "improve" socket errors. Perhaps one of these was in error. We should investigate.

[headius]

This may have never passed. It appears that the failing spec was just added earlier this year: f8e468cc34ba90c62a978892143592b44fb0094f

This is likely an easy fix; we're just not handling a connection error by raising the appropriate Errno exception.

[headius]

I have a fix. The JDK's UDP subsystem was raising java.net.PortUnreachableException, which appears to be the UDP equivalent of a connection refused. With JRuby fixed to raise ECONNREFUSED, we get the apparently-intermittent skip behavior in the spec.

[headius]

Should be good now.

[tarcieri]

@headius sweet, thank you!