Bug #7648

GServer does not close cleanly from signal interrupt context

Added by Joe Leo over 1 year ago. Updated 8 months ago.

Assignee:Motohiro KOSAKI
Target version:next minor
ruby -v:ruby 2.0.0dev (2013-01-02 trunk 38676) [i686-linux] Backport:


When a signal interrupt is trapped, we can no longer call #close on GServer without it throwing a ThreadError.

1) Run the following code:

require 'gserver'

server = GServer.new 8080
trap("SIGINT") { server.stop }
gets # or any command that keeps the process running

2) Hit CTRL+C or whichever command will send the interrupt signal to this program.

WHAT I EXPECTED: In version 1.9.3, CTRL+C sends an interrupt signal and the program exits cleanly.

WHAT HAPPENED: When running the version from trunk the following stack trace is thrown.

C/home/joe/.rvm/rubies/ruby-head/lib/ruby/2.0.0/gserver.rb:116:in synchronize': can't be called from trap context (ThreadError)
from /home/joe/.rvm/rubies/ruby-head/lib/ruby/2.0.0/gserver.rb:116:in
from gserverbug.rb:5:in block in <main>'
from gserver_bug.rb:6:in
from gserver
bug.rb:6:in gets'
from gserver_bug.rb:6:in
from gserver_bug.rb:6:in `'

POSSIBLY RELEVANT: https://bugs.ruby-lang.org/issues/6416

NOTE: This was tried with AND without RVM with the same results.


#1 Updated by Joe Leo over 1 year ago

Confirming that this is still an issue on RC1:

ruby -v
ruby 2.0.0dev (2013-01-07 trunk 38733) [i686-linux]

/home/joe/lib/lib/ruby/2.0.0/gserver.rb:116:in synchronize': can't be called from trap context (ThreadError)
from /home/joe/lib/lib/ruby/2.0.0/gserver.rb:116:in
from /home/joe/dev/bane/lib/bane/launcher.rb:19:in block in stop'
from /home/joe/dev/bane/lib/bane/launcher.rb:19:in
from /home/joe/dev/bane/lib/bane/launcher.rb:19:in stop'
from ./bin/bane:22:in
block in '
from /home/joe/lib/lib/ruby/2.0.0/gserver.rb:140:in call'
from /home/joe/lib/lib/ruby/2.0.0/gserver.rb:140:in
from /home/joe/lib/lib/ruby/2.0.0/gserver.rb:140:in join'
from /home/joe/dev/bane/lib/bane/launcher.rb:15:in
block in join'
from /home/joe/dev/bane/lib/bane/launcher.rb:15:in each'
from /home/joe/dev/bane/lib/bane/launcher.rb:15:in
from ./bin/bane:23:in `'

#2 Updated by Koichi Sasada about 1 year ago

  • Category set to lib
  • Target version set to 2.0.0

Who can check it?

#3 Updated by Koichi Sasada about 1 year ago

  • Assignee set to Yusuke Endoh

No response here.

#4 Updated by Yusuke Endoh about 1 year ago

  • Target version changed from 2.0.0 to next minor

#5 Updated by Yusuke Endoh 11 months ago

  • Status changed from Open to Assigned
  • Assignee changed from Yusuke Endoh to Motohiro KOSAKI

kosaki-san, what do you think?

Yusuke Endoh mame@tsg.ne.jp

#6 Updated by Motohiro KOSAKI 8 months ago

  • Status changed from Assigned to Rejected

Holding mutex in trap is deadlockable. It is what ruby complained. The best workaround is to make new thread in trap handler and stop gserver asynchnorously, I think.

Also available in: Atom PDF