[Mono-dev] TCP 3.0.1/2

[Rodrigo Kumpera]

The deadlock is explained on the commit message, just launch MD using the
tpoll backend and it will deadlock after a while.



On Thu, Nov 29, 2012 at 5:56 PM, NokNok Developer <developer at noknok.net>wrote:

>  Caqn you provide a test case/code which demonstrates the deadlock?  That
> would be beneficial, otherwise i do not see the BUG IN MONO TCP ASYNC
> process getting fixed any time soon, which is more then a complete bummer,
> its a complete mess.
>
> Shawn
>
>
> On 11/29/2012 5:50 PM, Rodrigo Kumpera wrote:
>
>
>
>
> On Thu, Nov 29, 2012 at 5:06 PM, james <james at mansionfamily.plus.com>wrote:
>
>> I don't buy it.  You knew who it was from.
>>
>
>  This was never questioned.
>
>
>> If there were issues with justification in the documentation with the
>> original patch set, surely that should be dealt with at the time it was
>> pulled in?
>>
>
>  No explanation on why the patch worked was given. "It improves
> stability" was the closest I got. Given the patch was vouched by the
> previous maintainer
> of that code base it was merged.
>
>  A deadlock on the code was found, so I reverted it. Greg is free to do a
> pull request on an patch that fixes the deadlock and has a proper
> explanation.
>
>
>
>
>> You appear to be saying that:
>>  - patches are applied without discussion, and without adequate
>> justification on the ticket (were they checked in against a ticket?)
>>
>
>  The patch was discussed ad-nausea on this mailing list and on the github
> pull request. You can look for both, both happened in the public.
>
>
>
>>  - patches were reverted without any discussion with the author (was THAT
>> against a ticket?)
>
>
>  Bad patches are reverted. There was no ticket, just a pull request on
> github. Greg's test could not be reproduced outside of his environment.
>
>
>
>> What is the process supposed to be? IS there a process?
>>
>
> Good code shaped in proper patches go in. Bad code goes out. It is that
> simple.
>
>
>  Isn't this concerning?  TCP/IP performance and async IO is rather
>> important to anyone doing server apps.
>
>
>
>  I have no idea on what you're talking anymore. That there are bugs, or
> that I broken patch was merged in?
>
>
>
>
>
> _______________________________________________
> Mono-devel-list mailing listMono-devel-list at lists.ximian.comhttp://lists.ximian.com/mailman/listinfo/mono-devel-list
>
>
>
> _______________________________________________
> Mono-devel-list mailing list
> Mono-devel-list at lists.ximian.com
> http://lists.ximian.com/mailman/listinfo/mono-devel-list
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ximian.com/pipermail/mono-devel-list/attachments/20121129/046219f8/attachment.html>