Ruby 3.2 will have native IO#wait_* methods, don't require io/wait

[MSP-Greg]

Description

In Ruby 3.1 and earlier versions, IO#wait_* methods were contained in the standard library gem io/wait. Ruby 3.2 has moved the methods to the main IO code. See 'Move wait* methods to io.c' ruby/ruby#6036.

So, io/wait does not need to be required for Ruby 3.2 and later. Adjust code to only require it when needed.

Your checklist for this pull request

• I have reviewed the guidelines for contributing to this repository.
• I have added (or updated) appropriate tests if this PR fixes a bug or adds a feature.
• My pull request is 100 lines added/removed or less so that it can be easily reviewed.
• If this PR doesn't need tests (docs change), I added [ci skip] to the title of the PR.
• If this closes any issues, I have added "Closes #issue" to the PR description or my commit messages.
• I have updated the documentation accordingly.
• All new and existing tests passed, including Rubocop.

[silva96]

I'm running ruby 3.1 and ::IO.public_instance_methods(false).include? :wait_readable returns true. This causes the io/wait library not being loaded I suppose. Then my puma wont start:

 Early termination of worker
[1979180] + Gemfile in context: /home/ubuntu/www/releases/20221108181543/Gemfile
[1979176] ! Unable to start worker
[1979176] /home/ubuntu/.asdf/installs/ruby/3.1.2/lib/ruby/gems/3.1.0/gems/bundler-2.3.12/lib/bundler/runtime.rb:309:in `check_for_activated_spec!'

I upgraded from rails (7.0.2.4) to rails (7.0.4)

[MSP-Greg]

I'm running ruby 3.1 and ::IO.public_instance_methods(false).include? :wait_readable returns true

Something is loading io/wait, if I run that via the below, it returns false.

ruby -e "puts ::IO.public_instance_methods(false).include? :wait_readable"

Any changes re Ruby default gems in your Gemfile.lock when you upgraded?

[silva96]

This is the gist of the upgrade:

https://gist.github.com/silva96/f69a42f9ef3510e7761a8331a13edd3e

[MSP-Greg]

Can you try a patch? Might be something particular with your setup. There are three files that have the conditional:

unless Puma::HAS_NATIVE_IO_WAIT

change that to:

unless ::IO.public_instance_methods(false).include?(:wait_readable)

[MSP-Greg]

Finally looked at Bundler code to see what was happening...

This causes the io/wait library not being loaded I suppose

Nothing in the log you've shown indicates that. I suspect you need to add io-wait to your Gemfile, as the error logged probably implies that Bundler can't locate it.

EDIT: Not sure what is going on, as I just checked locally. io-wait is not in my Gemfile.lock. Puma loads correctly. Using Ruby 3.1.2 with Bundler 2.3.12 installed, as listed above.

[silva96]

Yes is not in mine either, it never was. But I read somewhere that rails removed it from their dependencies in the latests versions. I came from this issue: #2843

[MSP-Greg]

I assume you're having the problem with restarts? Regardless, can you patch Puma, or work from a GitHub branch, etc?

The Puma code that's logging what you listed is:

puma/lib/puma/cluster/worker.rb

Lines 61 to 66 in f323d12

	rescue Exception => e
	log "! Unable to start worker"
	log e
	log e.backtrace.join("\n ")
	exit 1
	end

So, the line log e is showing pretty much nothing. Not sure if we can get more info, but maybe e.class and/or e.message may have more info. Not.Sure.

EDIT: Given what's raising the error, maybe the following:

rescue Exception => e
  log "! Unable to start worker"
  log "class: #{e.class}"
  log "  msg: #{e.message}"
  if e.respond_to? :requirement
    log "gem name: #{e.name}"
    log "     req: #{e.requirement}"
  end
  log e.backtrace.join("\n    ")
  exit 1
end

[silva96]

Thanks I'll give it a try maybe on friday and let you know the output.

[MSP-Greg]

Thanks.

FYI, the above isn't fixing anything, but it's an attempt to get more info about what went wrong.

As in #2843, it's difficult to repo locally, and since you've got the problem, the above patch should (hopefully) show enough info to identify the problem...

[silva96]

Just if it helps, this is my puma.rb config in production:

#!/usr/bin/env puma

directory '/home/ubuntu/www/current'
rackup "/home/ubuntu/www/current/config.ru"
environment 'production'

tag ''

pidfile "/home/ubuntu/www/shared/tmp/pids/puma.pid"
state_path "/home/ubuntu/www/shared/tmp/pids/puma.state"
stdout_redirect '/home/ubuntu/www/shared/log/puma_access.log', '/home/ubuntu/www/shared/log/puma_error.log', true

threads 0,16

bind 'unix:///home/ubuntu/www/shared/tmp/sockets/puma.sock'

workers 2

prune_bundler

on_restart do
  puts 'Refreshing Gemfile'
  ENV["BUNDLE_GEMFILE"] = ""
end

[silva96]

Ok I found the problem,

Context: I'm deploying using capistrano, and after deploy I restart puma using systemd

In my puma.service

# I changed this
ExecStart=/home/ubuntu/.asdf/shims/puma -C /home/ubuntu/www/shared/puma.rb
# to this
ExecStart=/usr/bin/bash -lc '/home/ubuntu/.asdf/shims/bundle exec puma -C /home/ubuntu/www/shared/puma.rb'

The thing is after deploying, asdf puma shim was not being updating from 5.x to 6 so the Gemfile version differed from the puma that was running.

Using bundle exec would run the puma version recently installed.