> On Jan. 28, 2013, 8:08 p.m., Vinod Kone wrote:
> > third_party/libprocess/src/statistics.cpp, line 50
> > <https://reviews.apache.org/r/9091/diff/1/?file=251530#file251530line50>
> >
> >     Why virtual? Are you expecting this to be subclassed?
> 
> Ben Mahler wrote:
>     virtual so that deletion of a Process pointer pointing to a 
> StatisticsProcess will call this destructor. I think destructors should 
> always be virtual when dealing with inheritance: 
> http://www.learncpp.com/cpp-tutorial/123-virtual-destructors-virtual-assignment-and-overriding-virtualization/

Yes. I was just wondering if you ever anticipate passing StatiscticsProcess* as 
Process*. I guess virtual is always safe.


> On Jan. 28, 2013, 8:08 p.m., Vinod Kone wrote:
> > third_party/libprocess/src/statistics.cpp, line 176
> > <https://reviews.apache.org/r/9091/diff/1/?file=251530#file251530line176>
> >
> >     so is it ok to mutate the list and use its iterator at the same time!?
> 
> Ben Mahler wrote:
>     This was the original code from Statistics, but reading up on the map 
> erase:
>     "Iterators, pointers and references referring to elements removed by the 
> function are invalidated.
>     All other iterators, pointers and references keep their validity."
>     
>     So technically, we would need to advance to the next element to avoid 
> invalidation. (see: http://stackoverflow.com/a/5673209)
>     
>     But I've included a simpler fix here.

clever :)


- Vinod


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/9091/#review15763
-----------------------------------------------------------


On Jan. 28, 2013, 11:32 p.m., Ben Mahler wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/9091/
> -----------------------------------------------------------
> 
> (Updated Jan. 28, 2013, 11:32 p.m.)
> 
> 
> Review request for mesos, Benjamin Hindman and Vinod Kone.
> 
> 
> Description
> -------
> 
> This adds context into the Statistics model. Contexts provide a clean way to 
> logically group stats (e.g. contexts: "master", "slave", "monitor", 
> "libprocess", etc).
> 
> 
> This addresses bug MESOS-324.
>     https://issues.apache.org/jira/browse/MESOS-324
> 
> 
> Diffs
> -----
> 
>   third_party/libprocess/include/process/statistics.hpp 
> 9e3041a6e2a8ef022eacacad00bc4d974a8e33c9 
>   third_party/libprocess/src/statistics.cpp 
> 2fe8af83c6c63a0fa8cb2e9636f9289f0e3d7f2f 
>   third_party/libprocess/src/statistics_tests.cpp 
> 0aaab3526618171c7cfbd11d40d614344bcbfd0a 
> 
> Diff: https://reviews.apache.org/r/9091/diff/
> 
> 
> Testing
> -------
> 
> make check + libprocess make check
> 
> 
> Thanks,
> 
> Ben Mahler
> 
>

Reply via email to