> On Feb. 3, 2016, 10:45 p.m., Michael Park wrote:
> > src/tests/resources_tests.cpp, line 957
> > <https://reviews.apache.org/r/42751/diff/2/?file=1231620#file1231620line957>
> >
> >     We used to have a `Resources::size()` function which essentially did 
> > this, but intentionally removed it so that people don't rely on number of 
> > `Resource` instances. Is there a reason why we want to check for this?
> >     
> >     Here and below.
> 
> Neil Conway wrote:
>     The # of `Resource` instances is part of the public API of `Resources` 
> (e.g., clients can iterate over every `Resource`). If it is part of the 
> public API, it seems like something it would be worth testing.
>     
>     In this particular case, it doesn't matter that much, but in other test 
> cases (e.g., `AdditionDynamicallyReservedWithDistinctLabels`) it seems useful 
> to check.
> 
> Michael Park wrote:
>     Ok, synced with Jie on this as well. Let's re-introduce 
> `Resources::size()` and use that instead.
>     
>     I agree that the # of `Resource` instances is part of the public API 
> since as you say, one can iterate and count.
>     However, this still does not mean that such iterator math works. For 
> example, we could internally store the `Resource` objects
>     in a `std::set` (or any other data structure that does not guarantee 
> contiguous memory), rather than `std::vector`.
>     
>     Providing a `Resources::size()` would more accurately capture this 
> intention anyway. Would you be ok taking that on?

Sounds good.


- Neil


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/42751/#review117698
-----------------------------------------------------------


On Feb. 3, 2016, 11:04 p.m., Neil Conway wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/42751/
> -----------------------------------------------------------
> 
> (Updated Feb. 3, 2016, 11:04 p.m.)
> 
> 
> Review request for mesos and Michael Park.
> 
> 
> Repository: mesos
> 
> 
> Description
> -------
> 
> We should check that two reservations with the same role but different
> principals are considered distinct.
> 
> 
> Diffs
> -----
> 
>   src/tests/resources_tests.cpp 4b25e82c13e4f46c73803f773db90f269c09c48a 
> 
> Diff: https://reviews.apache.org/r/42751/diff/
> 
> 
> Testing
> -------
> 
> make check
> 
> 
> Thanks,
> 
> Neil Conway
> 
>

Reply via email to