Justin Collins
8/22/2006 12:46:00 AM
ara.t.howard@noaa.gov wrote:
> On Tue, 22 Aug 2006, Justin Collins wrote:
>
>> Ara.T.Howard wrote:
>>>
>>> i'm unsure if this should be expected to cause deadlock or not?
>>>
>>> ruby -r thwait -e' ThWait.all_waits(*Array.new(42){
>>> Thread.new{Thread.stop} }) '
>>>
>>> for 1.8.1 - 1.8.4 it does.
>>>
>>> thoughts?
>>>
>>> -a
>>
>> Seems like it would. You create a bunch of threads, tell them not to
>> run, then tell the main thread to wait on all the stopped threads.
>>
>> There's no way for this program to complete or continue.
>
> it seems like it should just hang then. otherwise this program
> wouldn't be
> able to work
>
> harp:~ > cat a.rb
> require 'thwait'
>
> loop{
> stopped = Thread.new{ Thread.stop }
>
> notify = Thread.new{ stopped.wakeup }
>
> ThWait.all_waits stopped
> }
>
> this code __should__ be able to run - but it won't. the docs say:
It _does_ run. Try putting in an output statement in the loop. However,
what happens (I'm guessing) is something like this:
stopped = Thread.new { #=> context change
notify = Thread.new { stopped.wakeup } #=> okay, stopped can run
#
# Now, Thread.stop runs for stopped
#
ThWait.all_waits stopped #=> now it will never be able to complete
> a sleeping thread has not terminated, but whose to say it won't at
> some point
> in the future?
In some instances, like these examples, you can say that it won't
terminate. (I'm not sure if Ruby also detects deadlocks in more complex
examples incorrectly.)
>
> for instance. this is valid ruby
>
> ruby -e' Thread.new{ sleep }.join '
>
> it just hangs forever. similarly i feel that
>
> ThWait.all_waits Thread.new{ sleep }
>
> should also hang - not deadlock.
Well, I think what's happening is that Ruby can detect that, in the
second instance, there is no way for it not to deadlock. How could it
not? I'm not sure about the first instance, though.
-Justin