[vtk-developers] Request for a gentler tone in commit message verbiage...

David Lonie david.lonie at kitware.com
Wed Apr 29 14:50:17 EDT 2015


On Wed, Apr 29, 2015 at 2:40 PM, David Cole <DLRdave at aol.com> wrote:

> (1) what was the problem with double and make_pair -- or was that a
> misleading commit message because the problem was always with the use
> of NULL ... ?
>

Nope, it was with the double. See below.


> (2) Will you please rephrase the comment here in a future commit?
>
> https://gitlab.kitware.com/vtk/vtk/commit/e3d0653de443435533ebfb2db3fa6513e3b5c86b#7d40499821cd2f098d779d9b0da192039787a70a_587_594


No can do, it was merged before the complaint was made:
https://github.com/Kitware/VTK/commit/e3d0653de443435533ebfb2db3fa6513e3b5c86b


because it's simply not true. My code snippet proves that MSVC 2013's
> make_pair does NOT have a problem with double -- it has a problem with
> arguments of poorly defined type. NULL is possibly just "0" under some
> circumstances, which has type "int." If you construct a local variable
> of the correct type, with value NULL, then make_pair has no problem
> deducing the correct type.


The first issue was with double:

 https://open.cdash.org/viewBuildError.php?buildid=3784595

error C2664: 'std::pair<double,vtkTextProperty *>
std::make_pair<double,vtkTextProperty*>(_Ty1 &&,_Ty2 &&)' : cannot convert
argument 1 from 'double' to 'double &&'

Try the change I suggested to your test snippet in the previous email and
it should trigger this. There was a discussion on some windows support
forum where they said their implementation did strictly follow the standard
and they won't change the compiler, which is fair enough. But I don't think
anyone would find that behavior intuitive or expected :P

On Wed, Apr 29, 2015 at 2:37 PM, Will Schroeder <will.schroeder at kitware.com>
 wrote:

> I would say to Dave Cole's point that I have personally seen several
> problems in the very modest development efforts I have been doing on
> Windows (e.g., vtkSMPTools related) mostly in the CMake files, etc. Every
> compiler/platform has something good to say and we need to listen. This
> ensures that our software remains world class, despite our personal
> religious beliefs :-) Actually I actually think more of us need to build on
> Windows/MSVC because stuff is falling through the cracks.
>

Agreed that all platforms should be well tested and supported. I think the
real question with your issue is, why didn't the dashboards catch these
problems? Is there some way we could have caught them with a
nightly/continuous test?

Dave
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/vtk-developers/attachments/20150429/6e434d17/attachment.html>


More information about the vtk-developers mailing list