Error on restart using control server on Windows

[emassip]

Windows 2003 Server, puma 1.6.3, MRI 1.9.3p125.

D:\redmine>puma --control tcp://127.0.0.1:3001 --control-token control
C:/Ruby193/lib/ruby/gems/1.9.1/gems/activesupport-3.2.8/lib/active_support/depen
dencies.rb:251:in `block in require': iconv will be deprecated in the future, us
e String#encode instead.
Puma 1.6.3 starting...
* Min threads: 0, max threads: 16
* Environment: development
* Listening on tcp://0.0.0.0:9292
* Starting status server on tcp://127.0.0.1:3001
*** Sorry signal SIGUSR2 not implemented, restart feature disabled!
Use Ctrl-C to stop

Getting http://127.0.0.1:3001/restart?token=control returns status = 'OK' but:

* Restarting...
C:/Ruby193/lib/ruby/gems/1.9.1/gems/activesupport-3.2.8/lib/active_support/depen
dencies.rb:251:in `block in require': iconv will be deprecated in the future, us
e String#encode instead.
Puma 1.6.3 starting...
* Min threads: 0, max threads: 16
* Environment: development
* Inherited tcp://0.0.0.0:9292
C:/Ruby193/lib/ruby/gems/1.9.1/gems/puma-1.6.3/lib/puma/server.rb:128:in `for_fd
': not a socket file descriptor (ArgumentError)
        from C:/Ruby193/lib/ruby/gems/1.9.1/gems/puma-1.6.3/lib/puma/server.rb:1
28:in `inherit_tcp_listener'
        from C:/Ruby193/lib/ruby/gems/1.9.1/gems/puma-1.6.3/lib/puma/cli.rb:320:
in `block in run'
        from C:/Ruby193/lib/ruby/gems/1.9.1/gems/puma-1.6.3/lib/puma/cli.rb:314:
in `each'
        from C:/Ruby193/lib/ruby/gems/1.9.1/gems/puma-1.6.3/lib/puma/cli.rb:314:
in `run'
        from C:/Ruby193/lib/ruby/gems/1.9.1/gems/puma-1.6.3/bin/puma:10:in `<top
 (required)>'
        from C:/Ruby193/bin/puma:19:in `load'
        from C:/Ruby193/bin/puma:19:in `<main>'

[emassip]

The exact same error happens using the -S option so this issue may be some duplicate of #165 (but with 1.6.3).

[evanphx]

Hm, perhaps TCPSocket.for_fd doesn't work properly on windows...

[evanphx]

So, it appears that to make the socket descriptors inherit properly into the child process by win32 you have to use DuplicateHandle which it appears MRI doesn't do. I have an idea for a fix, but I'll have to setup a win32 VM to test on. Could you tell me about the setup you're using?

[emassip]

Well, as I said, ESXed Windows 2003 server with MRI 1.9.3p125 installed using RubyInstaller but I think that any win32 setup would lead to the error.

About the software setup I don't think this will change anything either but I was running the Redmine trunk @ http://svn.redmine.org/redmine/trunk with additional Gemfile.local file containing a single 'gem "puma"' line in app root dir.