Segfault when forked Queue is being finalized by GC about matzruby

[mks-m]

any chance this will get fixed?

[brentr]

Sorry, I just saw this issue for the first time a couple days ago.
A fix has been pushed to github. Let me know how they work for you.
Note:
The problem is not really specific to ruby with the MBARI patch set.
See below:

The fastthread lib kills all threads waiting on a queue when that queue
finalized. If this happens in the child process of a fork, those threads
will likely have already been destroyed and may be slated for
finalization themselves. Trying to run them (even for kill) leads to
a exception out of in rb_thread_check()

This patch makes fastthread check whether the threads are intact
before trying to kill them.  Netshade provided the following test case:

ruby  -rthread -e '
q = Queue.new; Thread.new { q.pop }; pid = fork;
if pid.nil?; q = nil; GC.start; else; Process.wait(pid); end'

The test case fails more often when the MBARI patches are applied because
they make GC work better -- so, it is more likely that the threads
will be immediately slated for finalization by the first pass after fork.

This patch should also be applied to the MRI mainline.