Tests broken on jRuby 19 about split

[buddhamagnet]

I know we need to look into it.

[m0wfo]

Just tried running on my machine with JRuby 1.6.8 and 1.7 and everything passes :/ I'm using bundler v1.2.1, otherwise all gem versions are the same. Is there anything I'm missing about Travis's environment I wonder?

[andrew]

Maybe @rkh can provide some insight?

[m0wfo]

My JVM is the Oracle one (I suspect Travis uses a distro-supplied build) and it targets amd64 rather than i386 but Mr Google didn't turn up any architecture-related bugs. Both are at 1.7.0_07.

[buddhamagnet]

Oddly enough I made an innocuous commit to resque and the build
failed under jruby 1.9 also. A redis thing?

[buddhamagnet]

OK so it just happened again and looks like it fails any spec where an experiment is saved. Code for that method:

def save
  if new_record?
    Split.redis.sadd(:experiments, name)
    Split.redis.hset(:experiment_start_times, @name, Time.now)
    @alternatives.reverse.each {|a| Split.redis.lpush(name, a.name) }
  else
     Split.redis.del(name)
     @alternatives.reverse.each {|a| Split.redis.lpush(name, a.name) }
   end
 end

[buddhamagnet]

Just added a couple of specs to test atomic Redis operations (sadd and hset) under jruby...

[andrew]

Looks like mock_redis was completely borked on jruby, especially in 19 mode, it's been removed for now and the tests are passing again.