call_as_daemon method don't stop child process

[sada]

Your environment

• ruby -v: ruby 3.2.2 (2023-03-30 revision e51014f9c0) [x86_64-linux]
• daemons (1.4.1)

Describe the bug

Calling Process.waitpid inside the at_exit method will prevent the child process from stopping.

To Reproduce

Script:

Gemfile

# frozen_string_literal: true

source "https://rubygems.org"

gem "daemons"

test_daemon.rb

#!/usr/bin/env ruby

require 'bundler/setup'
Bundler.setup
require 'daemons'

at_exit do
  begin
    Process.waitpid
  rescue Errno::ESRCH, Errno::ECHILD
  end
end

2.times do |i|
  Daemons.run_proc("test_daemon.#{i}") do
    loop do
    end
  end
end

Terminal:

$ test_daemon.rb start

Child process doesn't stop.

$ ps aux
--snip--
sada     14719  1.8  0.2 177232 37168 pts/6    Sl+  17:01   0:00 ruby ./test_daemon.rb start
sada     14735  0.0  0.2 177232 33832 ?        Ssl  17:01   0:00 ruby ./test_daemon.rb start
sada     14737 94.6  0.2 180308 37540 ?        Rl   17:01   0:08 test_daemon.0

The following fork operations are no longer possible.

3. Forks another child process and exits first child. This prevents the potential of acquiring a controlling terminal.

daemons/lib/daemons.rb

Lines 43 to 51 in a0e84bc

	# 1. Forks a child (and exits the parent process, if needed)
	# 2. Becomes a session leader (which detaches the program from
	# the controlling terminal).
	# 3. Forks another child process and exits first child. This prevents
	# the potential of acquiring a controlling terminal.
	# 4. Changes the current working directory to "/".
	# 5. Clears the file creation mask (sets +umask+ to 0000).
	# 6. Closes file descriptors (reopens +$stdout+ and +$stderr+ to point to a logfile if
	# possible).

By the way, I encountered this problem when using delayed_job and debug.

Expected behavior

Child processes can be stopped.
Since the first child process should be stopped, I thought it would be a good idea to change the call to the exit! method.

daemons/lib/daemons/daemonize.rb

Line 67 in a0e84bc

exit if pid = safefork

[thuehlinger]

Thank you for the report. I am not sure why you would want to use at_exit in such a scenario, but also I currently cannot follow on why the first child should not exit in this case? Do I understand correctly, that your proposed solution is to replace exit if pid == safefork by exit! if pid == safefork?

[sada]

exit! if pid == safefork

Yes.
I thought I should stop the first child process so that it is not affected by anything else.

I am not sure why you would want to use at_exit in such a scenario, but also I currently cannot follow on why the first child should not exit in this case?

This issue occurs with delayed_job and debug.
Increase the number of workers using delayed_job's number_of_workers option.

https://github.com/collectiveidea/delayed_job/blob/v4.1.11/lib/delayed/command.rb#L102-L105

Next, loaded debug.

https://github.com/ruby/debug/blob/v1.9.2/lib/debug/local.rb#L106-L107

delayed_job tries to increase the number of workers, but the child process is not stopped, just like the repro program.
Also, in Rails development, debug is loaded by default, When starting delayed_job, unexpected block is caused by this issue.

https://github.com/rails/rails/blob/v7.1.3.2/railties/lib/rails/generators/rails/app/templates/Gemfile.tt#L38

[hanhb]

@sada , I am too seeing your issue with delayed_job just hung when you start it in daemon mode (no pid file created). I was able to solve my issue by requiring debug gem differently.

gem 'debug', platforms: %i[ mri windows ], require: 'debug/prelude'

Hope that information helps you