Hi,

At Tue, 23 Feb 2016 14:10:52 +0000,
Victor Orlikowski wrote:
> 
> I have re-read the source for Eventlet's get() on a queue.
> I am willing to give the code a try in the lab without the timeouts.
> I will report back my results - but I believe that the timeout code in get() 
> is not significantly different from the code used to schedule an unlock of a 
> blocked getter on a put().

Did you report the original eventlet bug to the developer community?
IMO, eventlet bugs should be fixed in eventlet, but at the same time,
ryu can have bug workarounds as a interim solution.

> That still leaves me needing a judgment call on the CLI option.
> 
> My reasons for believing that this option can be altered into being a config 
> file option only:
> 1) That option was included at my request, and has only been present in one 
> release.
> 2) Because of that fact, I believe that there will be very few users of that 
> option currently (perhaps only me - and I do not use it directly).
> 3) The option was chosen to have a "good default" that few users were 
> expected to have to change. This strengthens the case that few users would 
> have set it directly, and that no/vanishingly few user applications are 
> expected to be broken by this change.
> 
> I will report back on my testing of the code without timeouts on get().
> Please let me know how to proceed on the command line option; it is my 
> opinion that user applications are highly unlikely to break as a result of 
> this change, and that it is better to make the correct change now (so as to 
> make the behavior correct, and further decrease the likelihood of causing any 
> issues with user applications).

I have no strong opinion on this.
Maybe putting a clear notice on the changelog on incompatibility is
fine.

> Best,
> Victor
> --
> Victor J. Orlikowski <> vjo@[cs.]duke.edu
> 
> > On Feb 22, 2016, at 10:46 PM, Victor Orlikowski <[email protected]> wrote:
> > 
> > 1) We have a fundamental disagreement about the necessity of the timeouts 
> > on get().
> > If I understand your position - you contend that they have no merit 
> > whatsoever.
> > I contend (perhaps incorrectly) that they add minimal overhead and ensure 
> > that the get() operations have a regular "wakeup" to prevent potential 
> > missed wakeups. My desire for regular wakeups is predicated on my testing 
> > experiences, which *did* prove a need for the semaphores - but may *not* 
> > have proved a need for the timeouts.
> 

------------------------------------------------------------------------------
Site24x7 APM Insight: Get Deep Visibility into Application Performance
APM + Mobile APM + RUM: Monitor 3 App instances at just $35/Month
Monitor end-to-end web transactions and take corrective actions now
Troubleshoot faster and improve end-user experience. Signup Now!
http://pubads.g.doubleclick.net/gampad/clk?id=272487151&iu=/4140
_______________________________________________
Ryu-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/ryu-devel

Reply via email to