Bug #2371: [BUG] thread_free: locking _mutex must be NULL

• Status: Closed
• Assignee: -
• Target version: 2.0.0
• ruby -v: ruby 1.9.2dev (2009-11-10 trunk 25704) [x86_64-linux]

[orem (Chris Schlaeger)]

=begin
During the development of a project management application I encountered the following problem. Every 10th run or so, the ruby interpreter crashes with the error message in the subject. I've attached a full backtrace in the log file. I've tried to reproduce the problem with a smaller test case, but without success. Sorry for that!

Here is how to reproduce the problem:

1. Download the taskjuggler gem: http://www.taskjuggler.org/tj3/taskjuggler-0.0.3.gem and install it
2. change to the 'examples' directory that comes with the gem or download tutorial.tjp from http://www.taskjuggler.org/tj3/examples/tutorial.tjp
3. while /bin/true ; do tj3 -c 3 tutorial.tjp || break; done

The -c N option should be set slightly higher than the number of CPU cores in the system.
The program works fine with ruby 1.8.7 PL27.
=end

[mame (Yusuke Endoh)]

=begin
Hi --

2009/11/15 Chris Schlaeger [e-mail]:

Here is how to reproduce the problem:

1. Download the taskjuggler gem: http://www.taskjuggler.org/tj3/taskjuggler-0.0.3.gem and install it
2. change to the 'examples' directory that comes with the gem or download tutorial.tjp from http://www.taskjuggler.org/tj3/examples/tutorial.tjp
3. while /bin/true ; do tj3 -c 3 tutorial.tjp || break; done

The -c N option should be set slightly higher than the number of CPU cores in the system.
The program works fine with ruby 1.8.7 PL27.

Thank you for your report! Duplicated.

A smaller test case:

m = Mutex.new
Thread.new { m.lock; sleep }
sleep 0.2
Thread.new { sleep 0.2; fork { GC.start } }
m.lock

The following patch should fix the issue. Could you try?

diff --git a/thread.c b/thread.c
index ba717bd..5ea1982 100644
--- a/thread.c
+++ b/thread.c
@@ -380,6 +380,7 @@ thread_cleanup_func(void *th_ptr)
rb_mutex_unlock_all(th->keeping_mutexes, th);
th->keeping_mutexes = NULL;
}

• th->locking_mutex = Qfalse;
  thread_cleanup_func_before_exec(th_ptr);
  native_thread_destroy(th);
  }

--
Yusuke ENDOH [e-mail]

=end

[orem (Chris Schlaeger)]

=begin
Thanks for the very quick response!

The following patch should fix the issue.  Could you try?

Sure.

diff --git a/thread.c b/thread.c
index ba717bd..5ea1982 100644
--- a/thread.c
+++ b/thread.c
@@ -380,6 +380,7 @@ thread_cleanup_func(void *th_ptr)
       rb_mutex_unlock_all(th->keeping_mutexes, th);
       th->keeping_mutexes = NULL;
    }

•    th->locking_mutex = Qfalse;
      thread_cleanup_func_before_exec(th_ptr);
      native_thread_destroy(th);
   }

Yes, I can confirm that this patch fixes the problem. It also fixes
bug #2370 which seems to be just a different looking problem with the
same cause.

Chris

=end

[mame (Yusuke Endoh)]

• Status changed from Open to Closed
• % Done changed from 0 to 100

=begin
This issue was solved with changeset r25771.
Chris, thank you for reporting this issue.
Your contribution to Ruby is greatly appreciated.
May Ruby be with you.

=end