> On Nov. 12, 2012, 10:32 p.m., Andrew Stitcher wrote:
> > /proton/trunk/proton-c/include/proton/engine.h, line 288
> > <https://reviews.apache.org/r/8021/diff/1/?file=188827#file188827line288>
> >
> >     Do you really mean prior? surely if the engine is invoked prior to the 
> > deadline it will decide to do nothing since no timers have expired.
> >     
> >     I think you mean "as soon after the deadline as possible"

My wording bad english, not goodly written down.

;D

In practice, the driver should be invoked before that timeout, given that I/O 
events could be pending regardless of the timeout.  Technically, if the driver 
is not invoked until _after_ the timeout, timers will expire and possibly lead 
to a connection drop (e.g. keepalive not sent in time). 

A more accurate comment should probably read "invoked again at least once at or 
before the timeout expires". 


> On Nov. 12, 2012, 10:32 p.m., Andrew Stitcher wrote:
> > /proton/trunk/proton-c/src/driver.c, line 843
> > <https://reviews.apache.org/r/8021/diff/1/?file=188831#file188831line843>
> >
> >     gettimeofday() is obsolete (read man 2 gettimeofday) use clock_gettime 
> > in preference (although this then requires -lrt, but that's not a bad thing 
> > in itself).

Ok - the original code used "gettimeofday" in a couple of places - specifically 
in messenger.  Seemed like that functionality was better abstracted to the 
driver.

I'll try clock_gettime()...


- Kenneth


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/8021/#review13372
-----------------------------------------------------------


On Nov. 12, 2012, 9:27 p.m., Kenneth Giusti wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/8021/
> -----------------------------------------------------------
> 
> (Updated Nov. 12, 2012, 9:27 p.m.)
> 
> 
> Review request for qpid, Andrew Stitcher and Rafael Schloming.
> 
> 
> Description
> -------
> 
> Attached is an attempt at adding Connection idle-time-out support to 
> Proton-c, as defined by the AMQP 1.0 spec. 
> 
> Notes:
> 
> 1) Allows configuration of the AMQP Connection idle timeout locally, and 
> support for receiving the remote's configured idle timeout.
> 2) Will send periodic keepalive (empty) frames to satisfy remote's idle 
> requirements, if needed.
> 3) Will close the connection if remote violates the locally configured idle 
> deadline
> 
> 
> This addresses bug proton-111.
>     https://issues.apache.org/jira/browse/proton-111
> 
> 
> Diffs
> -----
> 
>   /proton/trunk/proton-c/bindings/python/proton.py 1408412 
>   /proton/trunk/proton-c/include/proton/driver.h 1408412 
>   /proton/trunk/proton-c/include/proton/engine.h 1408412 
>   /proton/trunk/proton-c/include/proton/types.h 1408412 
>   /proton/trunk/proton-c/src/dispatcher/dispatcher.h 1408412 
>   /proton/trunk/proton-c/src/dispatcher/dispatcher.c 1408412 
>   /proton/trunk/proton-c/src/driver.c 1408412 
>   /proton/trunk/proton-c/src/engine/engine-internal.h 1408412 
>   /proton/trunk/proton-c/src/engine/engine.c 1408412 
>   /proton/trunk/proton-c/src/messenger.c 1408412 
>   /proton/trunk/proton-c/src/util.h 1408412 
>   /proton/trunk/proton-c/src/util.c 1408412 
>   /proton/trunk/proton-j/src/main/scripts/proton.py 1408412 
>   /proton/trunk/tests/proton_tests/engine.py 1408412 
> 
> Diff: https://reviews.apache.org/r/8021/diff/
> 
> 
> Testing
> -------
> 
> Added an engine test case to verify that the timers are tracked, and the 
> action when timers fire, but I need to add a test at the driver level, too.
> 
> 
> Thanks,
> 
> Kenneth Giusti
> 
>

Reply via email to