Re: Hugin build

[Kornel Benko]

Am Samstag, 31. Januar 2015 um 09:31:09, schrieb T. Modes <Thomas...@gmx.de>
> Hi Hugin builder,
>
> the changes to Hugins code base of the last months results in changes to
> the build process.

[snip]

>
> 2.) hugin_executor branch
> This branch contains a lot of changes. They are mainly behind the scene,
> so not directly visible to the end user.

[snip makefile related]

>
> This branch contains also work to unify the multi-threading code in
> different parts of Hugin source code. The dependency on (internal or
> external) zthread has been removed. Hugin is now using OpenMP or
> boost::thread (or c++11 thread). I hope that this unification simplifies
> the further development.
>
> But!! this requires an OpenMP capable compiler. This is currently only
> checked, but not forced. If the compiler is *not* OpenMP capable nona
> and cpfind will run only single-threaded, which will be slower.
> Should we force an OpenMP enabled compiler? Or leave this optional?
> (CMake will report some informations.)

I'd say let it be optional, with warning.

> This branch contains now also some C++11 code. I could therefore reduce
> the dependency on Boost by using C++11 code instead of boost features.
> CMake checks if the compiler is C++11 capable and activate this code
> (From boost then only filesystem, graph and spirit is used).
> If the compiler is not C++11 capable, the boost functions will be used
> instead (this will require filesystem, graph, spirit, thread, date_time,
> iostreams, random, shared_ptr from Boost). I tested with 3 different
> compiler on 2 operating systems and there it works fine. I could not do
> more testing because I don't want to install further systems.
> So this needs some testing on different system if all works.
> CMake should report which code parts are used: C++11 std code or boost code.
> I added also an override switch: Set USE_BOOST=true in CMake will
> override the auto-detection and will always use boost (as in older
> versions).
>
> I hope that I described all good enough. So the question is now: can we
> merge the hugin_executor branch or does it breaks something? Please test
> and report back. Comments are also welcomed.

There are also 67 new po-strings to be translated.

I tested with some of my projects, but have not seen any flaws in my hugin usage.
I'm very much in favour to merge this branch to master.

> Thomas

Kornel