Ractors benchmark memory seg fault

[PedroAugustoRamalhoDuarte]

Hello, thanks for this gem. I am trying to setup a memory benchmark with new ruby ractors and i its returns a segfault, maybe is a problem with ruby its self or memory-profiler

Ruby versions tested: 3.1 and 3.2.2

# frozen_string_literal: true
require 'benchmark'
require 'benchmark/memory' # gem install benchmark-memory

def factorial(n)
  n == 0 ? 1 : n * factorial(n - 1)
end

Benchmark.memory do |x|
  x.report('ractors:') do
    ractors = []
    4.times do
      ractors << Ractor.new do
        1000.times { factorial(1000) }
      end
    end
    # take response from ractor, so it will actually execute
    ractors.each(&:take)
  end
end

Calculating -------------------------------------
main.rb:104: warning: Ractor is experimental, and the behavior may change in future versions of Ruby! Also there are many implementation issues.
[BUG] Segmentation fault at 0x0000000000000000
ruby 3.2.2 (2023-03-30 revision e51014f9c0) [x86_64-linux]

-- Control frame information -----------------------------------------------

[michaelherold]

This is due to memory_profiler not supporting Ractors. I'm not sure what the source of the segfault is, but when I change your example to the following, I experience the same segfault:

# frozen_string_literal: true

require "memory_profiler"

def factorial(n)
  n == 0 ? 1 : n * factorial(n - 1)
end

MemoryProfiler.start

ractor = Ractor.new do
  1000.times { factorial(1000) }
end

# take response from ractor, so it will actually execute
ractor.take

MemoryProfiler.stop.pretty_print

I will file an issue upstream.

[michaelherold]

This is due to an issue in the VM of MRI that was reported and fixed for the upcoming 3.3 release. I verified this on the latest commit of Ruby's trunk.

The fix was back ported for 3.2 as well, but would be out in a 3.2.3 release, which has not yet been released.

There is a backport for 3.1 as well, which I assume is in the same state.

All that to say that this should not be a problem in Rubies newer than the ones out today.

I don't think the effort of catching a case where someone uses a Ractor in a benchmark is worth the effort so I will go ahead and close this for posterity.

Thanks for reporting!