Lock file oscilates on platform specific gem dependencies since 1.12.x

[dekellum]

I have a project that I frequently test locally with both cruby and jruby. The gemspec based dependencies include nokogiri which has specific ruby and java platform gems. When upgrading from bundler 1.11.2 to 1.12.5, I find that the lock file (which I check in to git) changes to include nokogiri (1.6.8-java) every time I jrake test (jruby) and is removed every time I rake test (cruby). These lock file changes do not occur with bundler 1.11.2, so I wonder what might have changed?

The lock file changes would just be nuisance except that my project also includes safeguards in tasks like rake tag and rake push to fail early if there are any dirty files in the git tree. Thus as it stands I can't use these tasks with bundler 1.12.5 unless I stick with one ruby platform.

Gemfile:

# -*- ruby -*-

source 'https://rubygems.org'

gemspec :name => 'syncwrap'

Gemfile.lock (after jruby is used):

PATH
  remote: .
  specs:
    syncwrap (2.8.1)
      aws-sdk (~> 1.46)
      json (>= 1.7.1, < 1.9)
      nokogiri (>= 1.5.9, < 1.7)
      term-ansicolor (~> 1.2.2)
      tins (~> 0.13.2)

GEM
  remote: https://rubygems.org/
  specs:
    aws-sdk (1.66.0)
      aws-sdk-v1 (= 1.66.0)
    aws-sdk-v1 (1.66.0)
      json (~> 1.4)
      nokogiri (>= 1.4.4)
    json (1.8.3)
    json (1.8.3-java)
    mini_portile2 (2.1.0)
    minitest (5.8.4)
    nokogiri (1.6.8)
      mini_portile2 (~> 2.1.0)
      pkg-config (~> 1.1.7)
    nokogiri (1.6.8-java)
    pkg-config (1.1.7)
    rake (11.2.2)
    rdoc (4.0.1)
      json (~> 1.4)
    rjack-maven (3.0.5.0-java)
    rjack-tarpit (2.1.1)
      rake (>= 0.9.2.2, < 12)
      rdoc (>= 3.12, < 5)
    rjack-tarpit (2.1.1-java)
      rake (>= 0.9.2.2, < 12)
      rdoc (>= 3.12, < 5)
      rjack-maven (~> 3.0.4)
    term-ansicolor (1.2.2)
      tins (~> 0.8)
    tins (0.13.2)

PLATFORMS
  java
  ruby

DEPENDENCIES
  minitest (~> 5.8.4)
  rdoc (~> 4.0.1)
  rjack-tarpit (~> 2.1)
  syncwrap!

BUNDLED WITH
   1.12.5

[indirect]

@segiddins is this an intentional change in 1.12? I don't remember anything that would explicitly cause this...

[segiddins]

No, not intentional at all, at least on my part. I want to look into this when I have a chance.

[segiddins]

Reproduced the bug, I'll try and look into it later. Thanks for the issue @dekellum !