Threads run in the context of pry/byebug hang in >= 2.0.0

[travisp]

On pry-byebug 1.3.3:

[1] pry(#<RSpec::Core::ExampleGroup::Nested_1>)> x = Thread.new { sleep 0.1; print "x"; print "y"; print "z" } => #<Thread:0x007fbfbb0337c0@(pry):1 sleep> [2] pry(#<RSpec::Core::ExampleGroup::Nested_1>)> x.join xyz=> #<Thread:0x007fbfbb0337c0@(pry):1 dead>

On pry-byebug 2.0.0:
[1] pry(#<RSpec::Core::ExampleGroup::Nested_1>)> x = Thread.new { sleep 0.1; print "x"; print "y"; print "z" } => #<Thread:0x007f807dd83600@(pry):1 sleep> [2] pry(#<RSpec::Core::ExampleGroup::Nested_1>)> x.join
^^ hangs indefinitely

We noticed this while trying to access a connection in capybara-webkit, where threads are used in the connection code: thoughtbot/capybara-webkit#718

[travisp]

More specifically, the above threads work on pry-byebug commit 5c72fb5, but not 2 commits later b0eba72 (I can't test the commit in the middle because I get exceptions). It appears to be that it's simply of a matter of whether Byebug.start has been called or not (in the earlier version, Byebug.start was not called initially). Even on the earlier version, once the user has entered a command such as next, any threads hang.

[deivid-rodriguez]

Could you try to reproduce this using latest pry-byebug and byebug (from master). I think it should be fixed in byebug.

[travisp]

@deivid-rodriguez Unfortunately it seems that in pry-byebug master 917c365 and byebug master 99bd3269deb790 the problem still remains.

[deivid-rodriguez]

You means the problem remains separately in byebug´s master or using pry-byebug (master) + byebug (master).

[olivierlacan]

@deivid-rodriguez Running both byebug master and pry-byebug master I'm still seeing hangs inside of capybara-webkit threads as well. Is there anything that could help you with this issue?

Basic setup I have is rspec with Capybara and capybara-webkit.
In spec_helper.rb:

Capybara.javascript_driver = :webkit

RSpec.configure do |config|
#...
  config.around(:each, js: true) do |example|
    Capybara.using_driver(:webkit, &example)
  end
end

In the spec itself:

require 'spec_helper'

describe 'Something', js: true do
  it 'does something' do
    before do
      binding.pry
    end
  end
end

Running the spec, when the Pry session starts, any call to find('#selector') hangs indefinitely.

[deivid-rodriguez]

Are you able to reproduce using byebug master only?

[olivierlacan]

@deivid-rodriguez Yep. That's what I did originally. I had the last release of pry-byebug I think and byebug master.

[deivid-rodriguez]

@olivierlacan Did you uninstall pry-byebug, installed byebug's master branch and then reproduced this using byebug instead of binding.pry?

If that's the case, open an issue in byebug's repo?

[travisp]

@deivid-rodriguez I tested again with the latest (after 091cff) in byebug. With byebug alone the problem does not seem to exist. With pry-byebug the problem remains for me and thoughtbot/capybara-webkit#718 seems to still be an issue for me too.

[deivid-rodriguez]

@travisp I guess this is because pry-byebug is not using latest byebug's master. This should be fixed in next release.

@olivierlacan Sorry if my last sentence sounded rude, it's missing "could you".. :)

[deivid-rodriguez]

Closing for lack of feedback. Feel free to reopen if still happening.

[travisp]

@deivid-rodriguez on pry-byebug 3.1.0 and byebug 4.0, as well as pry-byebug master this still remains an issue -- any threads run in this context seem to hang.

[justin808]

I just tried this and also ran into this issue. So still seems like an issue.

[deivid-rodriguez]

This seems like a duplicate of #69. Let's comment there.