threads, Hash#delete_if, and ruby segfaults.

[Neil Spring]

should the following code cause the ruby interpreter to segfault? If
so, resolv.rb should have a mutex guard @senders. If not, st_foreach
should avoid the race when independent threads delete adjacent elements
from the same linked list.

#!/usr/bin/ruby

desired_length = 10000

hash = Hash.new
threads = (0..4).map { |i|
Thread.new {
while true
hash.delete_if { |k,v| Kernel.rand > 0.5 }
while hash.length < desired_length
hash[Kernel.rand(desired_length * 10)] = 12
end
end
}
}

(I submitted a really clunky version of this as ruby bug 1326 with not
nearly so simple a test case, but can't figure out how to follow up to
the message).

thanks,
-neil

 

[Yukihiro Matsumoto]

Hi,

In message "threads, Hash#delete_if, and ruby segfaults."

|should the following code cause the ruby interpreter to segfault? If
|so, resolv.rb should have a mutex guard @senders. If not, st_foreach
|should avoid the race when independent threads delete adjacent elements
|from the same linked list.

It is fixed in the latest snapshot.

matz.