[Mono-dev] ObjectPool

[Greg Young]

We will be stress testing this. This seems to be the root cause of the
other issues we have been experiencing, literally hundreds of things
are broken because of this one little issue (anything with callbacks,
concurrentstack/queue, etc etc).


On Wed, Jul 25, 2012 at 7:43 AM, Jérémie Laval <jeremie.laval at gmail.com> wrote:
> There were indeed two bugs in there, a missing barrier and the issue you
> described.
>
> I also commented out the code if you want more details.
>
> --
> Jérémie Laval
> http://neteril.org
>
>
>
> On Wed, Jul 25, 2012 at 11:54 AM, Greg Young <gregoryyoung1 at gmail.com>
> wrote:
>>
>> We have had some moving forward in our mono stability issues. I
>> figured I would start up a chat here about one place that seems to be
>> causing many problems.
>>
>>
>>
>> https://github.com/mono/mono/blob/master/mcs/class/corlib/System.Collections.Concurrent/ObjectPool.cs
>>
>> I am not really sure the code here is actually thread safe... Maybe it
>> doesn't need to be for some reason?
>>
>> We have added volatiles for add/remove index (yuriy pointed out that
>> threads were often stuck at line 86) but even with that im still not
>> sure its actually threadsafe. There are some weird edge conditions
>> that seem to be here (like if CompareExchange fails 3 times it will
>> just never set the remove index and continue on).
>>
>> Somebody has obviously spent a lot of time thinking about this code
>> and optimizing it. I want to make sure we get the full "why" behind
>> things so we dont a) waste time b) introduce issues.
>>
>> Greg
>>
>> --
>> Le doute n'est pas une condition agréable, mais la certitude est absurde.
>> _______________________________________________
>> Mono-devel-list mailing list
>> Mono-devel-list at lists.ximian.com
>> http://lists.ximian.com/mailman/listinfo/mono-devel-list
>
>



-- 
Le doute n'est pas une condition agréable, mais la certitude est absurde.