[Kmymoney-devel] KMyMoney Dependencies and Build Requirements

[Cristian Oneț]

2014-07-03 16:32 GMT+03:00 Christian Dávid <christian-david at web.de>:
> Am Mittwoch, 2. Juli 2014, 21:43:29 schrieb Fernando Vilas:
>> I think I remember reading that for either Qt5 or KF5, that C++11 was a
>> requirement. Even as far back as RHEL5, you could get the gcc44 toolchain
>> installed and use "-std=c++0x" for most of what they use. Based on the
>> state  of affairs with GCC, LLVM, and MSVC, I have no problem with c++11
>> being a requirement. This is further supported if I am remembering
>> correctly above.
>
> Both, Qt5 and KF5 do not require C++11. I do not about KF5, but Qt5 has C++11
> support. If you build it with C++11 enabled some classes contain constexpr,
> move constructors, initialization lists and other nice improvements.
>
> But I do not know if we can make use of it, as the distributions decide if the
> shared libraries are compiled with C++11 enabled or not.
>
> Anyway we should enable C++11 because *we* can make our lives easier with it.
> I am sure KF5 will require it in near future (Qt probably not). Also I want it
> enabled because it is more fun for me :)
>
> With the compilers you are right, most features are supported by the main
> compilers for all systems (gcc, clang and MS Visual Studio) for a long time.
>
> Also I think it is questionable if someone who has a distribution which is
> several years old is interested in compiling a new (and quite untested)
> version of KMyMoney — which will be released at the end of 2014. She/he needs
> to compile huge packages like Qt5, KF5 and further dependencies as well. Who
> is willing to do that cannot be stopped with an old compiler.

I've started to do some profiling again and found another argument for
the case that enabling C++11 would be a great benefit to the project.

1. We could really benefit from move semantics, our API is a perfect
fit for that (think about all the MyMoney* objects)
For example while loading a file the method which takes most CPU
cycles in the file storage object is MyMoneySeqAccessMgr::price [1].
You might ask why is that, well... read [2] (I know it's long) and
you'll notice that the compiler can't perform RVO [3]. We need to
improve the performance of the code we write but we should not
sacrifice the readability to do this. We can fix stuff using our
current C++03 but to do it properly without hurting readability I
would suggest switching to C++11.

2. We could benefit from auto, range-based for and other stuff to
improve readability
Line like [4] would simply go away.

3. It would allow us (the developers) to learn new stuff and write better code
Although I'm a C++ developer I didn't write a single line of C++11
code yet. We should be learning and having fun while working in our
free time.

So, I totally agree with you Chris, let's enable C++11 in KMyMoney, I
know that KDevelop already needs C++11 so it shouldn't restrict us
from distributing KMyMoney in any way. I'll take care of the necessary
build system changes.

Regards,
Cristian

[1] https://projects.kde.org/projects/extragear/office/kmymoney/repository/revisions/master/entry/kmymoney/mymoney/storage/mymoneyseqaccessmgr.cpp#L1773
[2] http://www.drdobbs.com/move-constructors/184403855
[3] http://en.wikipedia.org/wiki/Return_value_optimization
[4] https://projects.kde.org/projects/extragear/office/kmymoney/repository/revisions/master/entry/kmymoney/mymoney/storage/mymoneyseqaccessmgr.cpp#L1568

>
> Greetings
> Chris
> _______________________________________________
> KMyMoney-devel mailing list
> KMyMoney-devel at kde.org
> https://mail.kde.org/mailman/listinfo/kmymoney-devel