Most common segfault about mini_racer HOT 8 CLOSED

Can you confirm this happens as well on latest (2.3.1?)

from mini_racer.

Oh! You are referring to ruby right?

from mini_racer.

I assume this is true. It's going to take me a bit of time to get environments up to 2.3.1.

from mini_racer.

Due to multiple factors, I cannot upgrade to 2.3.1 at this time. My current version is 2.2.4 -
I'll attempt to create a repro in mini_racer based off the linked article. If I get it to work, then I'll try for a test run against 2.3.1

from mini_racer.

Managed to get the load testing setup working and got 31 instances of this in 25 minutes. So, if we do get a 'fix', it'll be pretty easy to test it. :)

from mini_racer.

Ok. I have repro for the segfault. Can't tell what it is, but it seems running full context many times in a loop can cause it sometimes. Potentially it's the GC freeing something it shouldn't be.
Originally I had the loop running 50 times, and right before I was going to give up and change the code-- I got the repro, so I pumped it up to 5000 -- obv. not the most efficient test case, and not even consistent, but man, it hits a segfault pretty consistently in a similar way.

  def test_segfault
    5000.times do
      GC.start

      context = MiniRacer::Context.new(timeout: 5)
      context.attach("echo", proc{|msg| msg.to_sym.to_s})
      assert_raises(MiniRacer::EvalError) do
        context.eval("while(true) echo('foo');")
      end
    end
  end

from mini_racer.

Ok and looks like if you just run the above code in a loop with or without GC.start, you basically get every class of c-level exception I have seen in my logs.
Probably should have provided that test case a while back.

from mini_racer.

I have confirmed that these segfaults are related to during script-run GC issues (most likely moving ruby/c-memory around when it's still in use by ruby or c++)
Please see #41

from mini_racer.