Build fails on JRuby

[route]

Sometimes it passes without issues like here https://travis-ci.org/route/ferrum/builds/587362556
but more often it fails sporadically at any given time https://travis-ci.org/route/ferrum/jobs/585764841 with timeout.

We use concurrent-ruby primitives here like Hash, Ivar, Mvar and some Ruby threads for websocket/driver nothing unusual but I cannot spot the issue...

/cc @headius

[headius]

Hello, thanks for letting me know about this issue!

The failure is a bit confusing... the spec that fails is "Ferrum::Browser lists the open windows" but it appears that the failure happens downstream from a browser.reset call. The only such call I can see in browser_spec.rb is in a different spec, "Ferrum::Browser clears local storage after reset".

I think the failure is happening in the "before" section here: https://github.com/route/ferrum/blob/master/spec/spec_helper.rb#L55-L59

In that code, I notice a timeout of 5 seconds. If this timeout involves starting a subprocess JRuby instance, there's a chance that slower environments may not start up fast enough. If that's the problem here, then bumping to a larger timeout like 10 seconds should make this more reliable.

[route]

Thank you Charles for taking a look I'll try to investigate it further

[route]

Related to #75

[route]

I don't think I have expertise and time to support JRuby

[headius]

Hey there, this fell off my radar years ago after I commented. I'd like to help!

[headius]

I did a local run and it appears to do well for a while, then something times out and everything fails after that. It looks like the Puma server is in "reset" method, so perhaps something tries to restart the server but fails, and from then on it's unavailable?

I'm happy to help get things green if you can explain how the tests work and we can collab to figure out why it starts failing. It seems to do well until it doesn't!

[headius]

There's lots of this warning too, which makes the test output pretty noisy:

/Users/headius/work/jruby/lib/ruby/gems/shared/gems/sinatra-3.2.0/lib/sinatra/base.rb:1140: warning: URI::RFC3986_PARSER.unescape is obsolete. Use URI::RFC2396_PARSER.unescape explicitly.

[headius]

It does slow down a bit when it starts failing, but it does not appear to time out completely! Here's the results I got on my local run:

Finished in 12 minutes 39 seconds (files took 5.48 seconds to load)
505 examples, 108 failures, 42 pending

Passing 80% of examples is good, right? 😀

Nearly all of the failures are this:

 81) Ferrum::Page::Screenshot#screenshot when encoding is base64 with jpeg supports screenshotting just the selected element
      Failure/Error: raise NoSuchTargetError unless resolved
      
      Ferrum::NoSuchTargetError:
        Ferrum::NoSuchTargetError
      Shared Example Group: "screenshot screen" called from ./spec/page/screenshot_spec.rb:321
      # ./lib/ferrum/context.rb:54:in 'create_target'
      # ./lib/ferrum/context.rb:20:in 'default_target'
      # ./lib/ferrum/context.rb:24:in 'page'
      # ./spec/page/screenshot_spec.rb:38:in 'block in <main>'

It seems like something fails and is not recovered, and then it starts failing all of these specs because of that. Any thoughts on where to look?