From 1ceeae13d2a84ea6f105d568529f3cd256ffb16d Mon Sep 17 00:00:00 2001
From: Koichi Sasada <ko1@atdot.net>
Date: Sat, 13 Dec 2025 03:22:04 +0900
Subject: [PATCH] call Thread.pass just after Thread.stop

The following `test_irb_fg_jobs_and_kill_commands` expects that after `fg 0`
Thread#1 should be `sleep_forever`.

```ruby
    def test_irb_fg_jobs_and_kill_commands
      out = run_ruby_file do
        type "irb"
        type "fg 0"
        type "jobs"
        type "kill 1"
        type "exit"
      end

      assert_match(/#0->irb on main \(#<Thread:0x.+ run>: running\)/, out)
      assert_match(/#1->irb#1 on main \(#<Thread:0x.+ sleep_forever>: stop\)/, out)
    end
  end
```

However, the following switching method expects that

```
    def switch(key)
      th, irb = search(key)
      ...
      th.run
      Thread.stop
```

(1) resume target thread
(2) Suspend current thread (-> sleep_forever)

`th.run` calls `Thread.pass` internally and maybe the context swtich to
resumed thread (#0 in this case).

Thread #0 should switch to #1 before `jobs` command.
Until Ruby 3.4, I/O operation makes context switching to another ready thread,
so that several I/O output can resume `#1`.

However, our planning new thread scheduler delays context switching on I/O
operations, so the thread #1 still running (because `Thread.stop` is not called yet).

This patch add `Thraed.pass` to make workaround for this test.
---
 lib/irb/ext/multi-irb.rb | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/lib/irb/ext/multi-irb.rb b/lib/irb/ext/multi-irb.rb
index 9f234f0cd..71a50ff29 100644
--- a/lib/irb/ext/multi-irb.rb
+++ b/lib/irb/ext/multi-irb.rb
@@ -65,6 +65,7 @@ def switch(key)
       @current_job = irb
       th.run
       Thread.stop
+      Thread.pass
       @current_job = irb(Thread.current)
     end
 
@@ -220,6 +221,7 @@ def IRB.irb(file = nil, *main) # :nodoc:
       end
     end
     Thread.stop
+    Thread.pass
     @JobManager.current_job = @JobManager.irb(Thread.current)
   end