[Dune-devel] Replace Dune C++11 by plain C++11

Martin Nolte nolte at mathematik.uni-freiburg.de
Wed Mar 12 19:32:46 CET 2014


Hi all,

I think the missing traffic on the mailing list proves that there is little 
interest in the topic and it seems everyone can live without the imports.

As to deprecating the imports: I don't think you can easily do that. Moreover, 
we are heading for 3.0, which "might contain non-backward compatible changes". 
IMHO we should just drop them (if so desired).

Moving the compatiblity headers does not sound like a good idea to me, here. 
We would need to deprecate the old ones anyway.

As to using std:: in the core modules: From my point of view, the import 
serves the purpose to easily use a fallback implementation in case the std:: 
variant is buggy (e.g., std::array on xlc). However, I agree that there is no 
need to support such compilers and a simple sed-command (combined with a find) 
will replace all std::array by Dune::array (or whatever name I would like to use).

Personally, I do (maybe no longer) have an opinion on the matter. If the 
majority of interesed people think it is simpler to use std::array instead of 
the import Dune::array, then so be it.

But: I do dislike reverting Christoph's work just for an extremely tiny bit of 
potential compatiblity (not to say out of pure spite). Therefore, I do have 
one request: Can we please either vote on the matter or allow Christoph to 
drop the imports without any further discussion?

Best,

Martin

On 03/12/2014 11:07 AM, Oliver Sander wrote:
> Am 12.03.2014 11:04, schrieb Christian Engwer:
>> Hi Oli,
>>
>> I just checked again.
>>
>> a) with respect to static_Assert I was wrong and we actually decided
>>     to drop dune_static_assert completely.
>> b) if we follow you suggestion
>>     > We cannot drop the namespace imports without deprecating them first.
>>     > But we can start using the std:: versions explicitly inside the
>>     > core modules, which is what Christoph did, IIRC.
>>     we basically revert the decision to have namespace import to ease the
>>     use of unsupported compilers. Which is something, where I don't have a
>>     strong opinion, besides we should stick to what we decided until we
>>     decided differently.
>
> I think it is a little bizarre to officially state a list of supported
> compilers, and then jump through loops to (kind-of) support the unsupported
> ones, too.  But I don't have a strong opinion on this, either.
> --
> Oliver
>
>>
>> Ciao
>> Christian
>>
>
>
>
>
> _______________________________________________
> Dune-devel mailing list
> Dune-devel at dune-project.org
> http://lists.dune-project.org/mailman/listinfo/dune-devel
>

-- 
Dr. Martin Nolte <nolte at mathematik.uni-freiburg.de>

Universität Freiburg                                   phone: +49-761-203-5630
Abteilung für angewandte Mathematik                    fax:   +49-761-203-5632
Hermann-Herder-Straße 10
79104 Freiburg, Germany




More information about the Dune-devel mailing list