Simplecov does not cover sub-process entry point

[shakedlokits]

I'm attempting to perform integration coverage for Ruby services on a multi-layer product, for this mean I've initialized coverage at the bottom of /usr/share/rubygems/rubygems.rb using:

require 'simplecov'
SimpleCov.start do
  filters.clear
  root Dir.pwd
  command_name 'subprocess coverage'
  coverage_dir '/coverage'
end

Then I've create 2 files:

• /application/a.rb

puts "at #{__FILE__}"
require_relative 'b.rb'

• /application/b.rb

puts "at #{__FILE__}"

Now when I run ruby /application/a.rb, the generated /coverage/.resultset.json file is:

{
  "subprocess coverage": {
    "coverage": {
      "/application/a.rb": [
        null,
        null
      ],
      "/application/b.rb": [
        1
      ]
    },
    "timestamp": 1500283332
  }
}

As you can see, the entry point file is not being covered(Verified with a larger requirement chain).

System Info

# uname -a
Linux 786eab7389d4 4.6.7-200.fc23.x86_64  SMP Wed Aug 17 14:24:53 UTC 2016 x86_64 x86_64 x86_64 GNU/Linux
# gem which simplecov
/usr/local/share/gems/gems/simplecov-0.14.1/lib/simplecov.rb
# ruby -e "puts RUBY_DESCRIPTION"
ruby 2.2.5p319 (2016-04-26 revision 54774) [x86_64-linux]

[PragTob]

👋

huh, I'm not sure about entry level coverage actually - that's interesting. Woudn't a worthwhile, workaround be to have top_level.rb and then require_relative "a" ?

I think the stdlib coverage mechanism hooks mostly into require to know what to do but I'd have to investigate.

[shakedlokits]

@PragTob I am under the assumption I cannot augment the code. This assumption rests because the products which are tested are large-scale(>10m LOC) and maintained by several independent teams.

Just got back to pushing this issue, any ideas where could we start this investigation?

[PragTob]

👋

Imo there's not much we can do as this is how the stdlib coverage that we rely on behaves. Even if it is maintained by lots of people, just creating a file a_for_simplecov.rb that then just does require "a" and running this one instead should be possible imo :)

[PragTob]

👋

Hi sicne all the time has passed I'm not sure if you're still dealing with this issue. If you did, an example repo would be cool :D

If not running all the tests separately and merging their results (with hopefully forthcoming better support) might be a good idea.