Geometry Boost List: Re: [geometry] Regression in the Variant support for distance()

Subject: Re: [geometry] Regression in the Variant support for distance()
From: Menelaos Karavelas (menelaos.karavelas_at_[hidden])
Date: 2014-05-26 08:08:41

Hi all.

On 26/05/2014 02:55 ??, Barend Gehrels wrote:
> Hi,
>
> Menelaos Karavelas wrote On 26-5-2014 13:28:
>> Hi all.
>>
>> On 26/05/2014 01:53 ??, Menelaos Karavelas wrote:
>>> BTW, there seems to be a problem not only with variant support on 
>>> distances, but also on variant support for assign.
>>> Samuel, could you please look this up.
>>>
>>
>> Fixed those errors as well. The PR is here: 
>> https://github.com/boostorg/geometry/pull/42
>
> Thanks for finding and fixing. Merged #41/#42.
>
> It is was broken a few days, all is looking quite yellow now.
>
> Not a big problem, these things happen, often with new contributions, 
> which are of course very welcome. I did not notice because busy with 
> other things last days.
>
> However, two more things about this, now this triggered it:
>
> 1) does anybody know if Boost regression matrix can mail notifications 
> about broken tests (which were earlier)? Or does anybody know 
> otherwise the most perfect tool (even if it runs locally) which checks 
> this? There must be an abundance on the web of course.
>
> 2) Adam, in the Contribution Tutorial, section "Modify the branch", 
> could you add that it is recommended to
> - run all unit tests before submit (even the tests you did not work 
> on)? (we had a problem with this earlier)

I would recommend to run the entire set of unit tests at least in the 
algorithms directory. I remember making changes in one algorithm in the 
past, that was working okay for that algorithm, and then the entire 
library was broken.... I am sure that I was part of the "earlier 
problem" mentioned by Barend above ;-)

> - check your contribution with two compilers (VirtualBox/gcc is really 
> easy to install on Windows machines, and if Ubuntu is your local 
> platform, use gcc/clang)
> It would be good if this is done before sending the merge/pull request.

One more suggestion: in case you want to be "pedantic" with your code, 
try also the following option on g++/clang++: -Wextra
It produces even more warnings especially with respect to unused variables.
Another option that I found useful at some point was -Wconversion.
BTW, from my point of view, these two options are not meant to be used 
at the development stage, but rather at the 
before-creating-a-pull-request stage.

- m.

>
> Regards, Barend
>
>
>
> _______________________________________________
> Geometry mailing list
> Geometry_at_[hidden]
> http://lists.boost.org/mailman/listinfo.cgi/geometry