ReaderThreadError: wrong number of arguments (given 0, expected 1)

[bjeanes]

Your environment

• ruby -v: 3.3.6
• rdbg -v: 1.9.2
• Inside docker (docker image ruby:3.3.6-slim-bookworm@sha256:7a89be75c6768a888f70c90f4626366de90909c4e4d62729775ea3b64672547c to be exact)
• Process started like: bundle exec rdbg -O -n --session-name='web' -c -- bin/rails server -b 0.0.0.0 -p 3000
  • RUBY_DEBUG_SOCK_DIR=/app/tmp (Rails root being /app inside container, /app/tmp is a volume mounted into container so socket can be attached to from outside container)

Describe the bug

Quite often during the course of debugging the debugger process seems to becomes in a paused state but unable to display the current thread or breakpoint. Rails requests hang even if disconnecting from the remote debugger (VSCode connecting via a socket).

After that point the main process becomes unresponsive and the only way to continue is to restart the parent process (which in our case means restarting the docker container).

To Reproduce

We have seen this arise sporadically. The following is one way we've been able to reproduce it but we have seen it occur even when avoiding the below:

• When paused at a breakpoint, calling another method manually (or in a watch) which has its own breakpoints set will cause execution to run to and pause at the nested breakpoint, but continuing from there will break the debugger.

  class Foo
    def a
      puts "hello"    # BP 1 <= active
      1 + 1
    end
  
    def b
      42              # BP 2
    end
  end
  Foo.new.a

  Given the following file with breakpoints set above, when stopped at BP 1, calling b will re-break at BP 2, but continuing from there seems to break the debugger and leave some thread in a hung state.

Again, this is just one cause but we have been able to reproduce it at times without any manual execution of methods with BPs. However, we do not have a reproduction case for this.

Expected behavior

The debugger doesn't hang

Additional context

When this is encountered, I disconnected VSCode's debugger and manually attached to the same debugger instance via rdbg -A socket. When doing so the debugger output shows:

root@b032c5a1c83b:/app> rdbg -A tmp/rdbg-7-web
(unknown) "Content-Length: 154\r\n"
(unknown) "\r\n"
(unknown) "{\"type\":\"event\",\"event\":\"output\",\"body\":{\"category\":\"console\",\"output\":\"DEBUGGER (client): Connected. PID:7, $0:bin/rails, session_name:web\\n\"},\"seq\":434}"
root@b032c5a1c83b:/app> # ^ exits

The application logs show:

DEBUGGER: Connected.
DEBUGGER: ReaderThreadError: wrong number of arguments (given 0, expected 1)
["/usr/local/lib/ruby/gems/3.3.0/gems/debug-1.9.2/lib/debug/server_dap.rb:504:in `puts'",
 "/usr/local/lib/ruby/gems/3.3.0/gems/debug-1.9.2/lib/debug/server.rb:156:in `greeting'",
 "/usr/local/lib/ruby/gems/3.3.0/gems/debug-1.9.2/lib/debug/server.rb:56:in `block (3 levels) in activate'",
 "/usr/local/lib/ruby/gems/3.3.0/gems/debug-1.9.2/lib/debug/server.rb:54:in `synchronize'",
 "/usr/local/lib/ruby/gems/3.3.0/gems/debug-1.9.2/lib/debug/server.rb:54:in `block (2 levels) in activate'",
 "/usr/local/lib/ruby/gems/3.3.0/gems/debug-1.9.2/lib/debug/server.rb:506:in `block in accept'",
 "/usr/local/lib/ruby/3.3.0/socket.rb:815:in `block (2 levels) in accept_loop'",
 "/usr/local/lib/ruby/3.3.0/socket.rb:812:in `each'",
 "/usr/local/lib/ruby/3.3.0/socket.rb:812:in `block in accept_loop'",
 "<internal:kernel>:187:in `loop'",
 "/usr/local/lib/ruby/3.3.0/socket.rb:810:in `accept_loop'",
 "/usr/local/lib/ruby/3.3.0/socket.rb:1170:in `block in unix_server_loop'",
 "/usr/local/lib/ruby/3.3.0/socket.rb:1124:in `unix_server_socket'",
 "/usr/local/lib/ruby/3.3.0/socket.rb:1169:in `unix_server_loop'",
 "/usr/local/lib/ruby/gems/3.3.0/gems/debug-1.9.2/lib/debug/server.rb:502:in `accept'",
 "/usr/local/lib/ruby/gems/3.3.0/gems/debug-1.9.2/lib/debug/server.rb:49:in `block in activate'"]
DEBUGGER: Disconnected.

[ko1]

ReaderThreadError: wrong number of arguments (given 0, expected 1) will be solved with #1124, but the original issue (hanging randomly) is not fixed.

Sorry for inconvenient. I also got similar issue but no idea how to debug it...

[bjeanes]

Ah the other piece of context here is that the web app is using Puma. I saw #1027 (comment) mentions some rough interactions there, so we I am experimenting with setting min_threads, max_threads = 1, 1 in Puma to see if it decreases the likelihood of these kinds of hangs.