Default code path doesn't appear to handle fork correctly?

[ioquatix]

I had to do this.

https://gist.github.com/clicube/5017378

This removed what appeared to be superfluous output:

Coverage report generated for RSpec to /home/samuel/Documents/Programming/rubydns/coverage. 0 / 599 LOC (0.0%) covered.

multiple times per run.

Perhaps this should be the default behaviour?

[bf4]

pid = Process.pid
SimpleCov.at_exit do
  SimpleCov.result.format! if Process.pid == pid
end

I like that.

[bf4]

are you using a merge timeout?

[ioquatix]

What do you mean?

Sent from my phone.

On 18/02/2015, at 2:28 pm, Benjamin Fleischer notifications@github.com wrote:

are you using a merge timeout?

�
Reply to this email directly or view it on GitHub.

[bf4]

Would you paste your simplecov config? (How you require, configure, start it)

[coderanger]

My solution for this was to add SimpleCov.at_exit { SimpleCov.instance_variable_set(:@result, nil) } to the post-fork code. Might be nice if the main at_exit implementation checked pids and only ran formatters or the minimum coverage check if the current pid was the one you ran start in.

Code that forks. SimpleCov start.

[bf4]

I'm up for a PR

[coderanger]

Mostly trying to think if that would break anything. The Coverage module already doesn't actually work across a fork, so I think it should be safe in all cases?

[colszowka]

Yes, I think that would be a good change - although maybe we can wrap the top level pid under SimpleCov.pid so we don't get weird interferences