From: normalperson@...
Date: 2014-11-24T21:08:31+00:00
Subject: [ruby-core:66442] [ruby-trunk - Bug #10540] Yielded fibers do not	execute ensure blocks

Issue #10540 has been updated by Eric Wong.


 This seems unfortunate.  I'm not sure if there's a good way to do
 this automatically with current APIs because Fibers require
 explicit scheduling.
 
 Perhaps storing Fibers in Thread.current and having a
 terminate_all_fibers method can be helpful:
 
 Thread.new do
   begin
     do_something_which_creates_fibers
   ensure
     # all thread-local fibers are resumed until they're dead
     terminate_all_fibers
   end
 end

----------------------------------------
Bug #10540: Yielded fibers do not execute ensure blocks
https://bugs.ruby-lang.org/issues/10540#change-50069

* Author: Yehuda Katz
* Status: Open
* Priority: Normal
* Assignee: 
* Category: 
* Target version: 
* ruby -v: ruby 2.1.2p95 (2014-05-08 revision 45877) [x86_64-darwin13.0]
* Backport: 2.0.0: UNKNOWN, 2.1: UNKNOWN
----------------------------------------
When a thread has paused fibers, if an exception occurs anywhere in the thread, `ensure` blocks in the paused fibers do not execute.

The effect of this is that block-scoped resources (like File.open, Mutex#synchronize, ActiveRecord transactions) can easily leak due to non-local effects. This may result in strange effects, like checking connections back into a connection pool that are in the wrong state, where the correct state was guarded with an `ensure` block.

Here is a very simple repro of this situation:

~~~
Thread.new {
  Fiber.new {
    begin
      puts "YIELD"
      Fiber.yield
    ensure
      puts "UNWIND"
    end
  }.resume
  raise
}
~~~

Expected result: `YIELD` is printed, followed by `UNWIND`
Actual result: `YIELD` is printed, but `UNWIND` is never printed



-- 
https://bugs.ruby-lang.org/