SIGINT or Ctrl+c have different outcomes when stopping server being run via puma and rails s

[shayonj]

Steps to reproduce

1. Clone a simple/barebones rails app: https://github.com/shayonj/test-repro-app

2. rails s

3. ctrl + c

Expected behavior

rails s
=> Booting Puma
=> Rails 5.1.2 application starting in development on http://localhost:3000
=> Run `rails server -h` for more startup options
Puma starting in single mode...
* Version 3.9.1 (ruby 2.3.1-p112), codename: Private Caller
* Min threads: 5, max threads: 5
* Environment: development
* Listening on tcp://0.0.0.0:3000
Use Ctrl-C to stop
^C- Gracefully stopping, waiting for requests to finish
=== puma shutdown: 2017-07-10 20:45:22 -0700 ===
- Goodbye!

Notice, its gracefully stopping and logs shutdown timestamp.

Actual behavior

rails s
=> Booting Puma
=> Rails 5.1.2 application starting in development on http://localhost:3000
=> Run `rails server -h` for more startup options
Puma starting in single mode...
* Version 3.9.1 (ruby 2.3.1-p112), codename: Private Caller
* Min threads: 5, max threads: 5
* Environment: development
* Listening on tcp://0.0.0.0:3000
Use Ctrl-C to stop
^CExiting

Perhaps, no graceful stop/exit?

System configuration

Ruby version: ruby 2.3.1p112 (2016-04-26 revision 54768) [x86_64-darwin16]
Rails version: Rails 5.1.2

Further info

The graceful stop/exit, works fine from the cli version. For example:

1. Clone a simple/barebones rails app: https://github.com/shayonj/test-repro-app

2. puma

3. ctrl + c

puma
Puma starting in single mode...
* Version 3.9.1 (ruby 2.3.1-p112), codename: Private Caller
* Min threads: 5, max threads: 5
* Environment: development
* Listening on tcp://0.0.0.0:3000
Use Ctrl-C to stop
^C- Gracefully stopping, waiting for requests to finish
=== puma shutdown: 2017-07-10 20:47:45 -0700 ===
- Goodbye!

---

Proposal

This is something, I came across when working on #1183. After digging a little deeper, it appears Rails::Server inherits from Rack::Server, which traps SIGINT signal and exits when no shutdown function is found.

So, to better handle SIGINT / ctrl + c, does introducing a .shutdown function to Rack::Handler::Puma, which would restore this signal and then kill the process sound like a better workaround?

Ex:

      def self.shutdown
        Signal.trap("SIGINT", "DEFAULT")
        Process.kill("INT", Process.pid)
      end

I can see an argument being made that this implementation is very specific to support Rack's requirement of a server instance having a shutdown function. But, given this may live in Rack::Handler::Puma, it still makes sense to me. Please, let me know if I am missing something :).

[shayonj]

Even the proposed solution above could work, proposed trapping of SIGINT as part of setting up signals in Puma's Launcher class at: #1377

[nateberkopec]

Closed by #1377