[Kmymoney-devel] Review Request 119234: Transform KHolidays into an optional dependency.

[Jack]

On 2014.07.14 08:16, Alvaro Soliverez wrote:
> 
> 
> > On July 11, 2014, 7:58 p.m., Christian David wrote:
> > > Is this only used until the pim libs are KF5 ready? Because I  
> think this function is so important that it justifies having the pim  
> libs a obligatory dependency. There will be many bug reports if a  
> package maintainer forgets to compile with kdepimlibs.
> >
> > Cristian Oneț wrote:
> >     No, I would like to see this as a permanent change. I'm not  
> that sure about which of the features are so important that they must  
> be a mandatory dependency. I like the idea of having the freedom to  
> build without kdepimlibs if necessary and if that can be obtained  
> easily why not have it? I'm not sugesting that packagers should do  
> this but for example in source based distributions users would be  
> able to have a choice, which is nice.
> >
> > Alvaro Soliverez wrote:
> >     I'm against this being a permanent change. It affects how  
> schedules are handled, which impacts all over the application.
> >     And I certainly don't want to track down a bug reported by an  
> unsupecting user just to figure out later a packager left the  
> dependency out because kde pim libs was "too big".
> >
> > Cristian Oneț wrote:
> >     I hate to disagree but I don't see it that way. The only effect  
> to the way schedules are handled is that processing dates are not  
> obtained from the holliday region thus the "old" (and documented)  
> behavior of considering only weekends as non-processing days will be  
> in effect.
> >
> >     If you ask me is this feature really that crucial that we  
> should make everyone build kdepimlibs in oder to use kmymoney I would  
> say no. Again, I'm not removing the feature, just making it optional.
> >
> >     The feature isn't even documented so I'm not sure if we would  
> start receiving bug reports about it. But just to be on the safe side  
> we could finish the feature in main.cpp:69 which would provide a list  
> of optional features that were built (or found as plugins) in the  
> about data.
> >
> >     If it's something I've learned from and appreacited in the KF5  
> model is that the take everything or leave it approach is wrong.  
> Think about it, kdepimlibs has 22 libraries plus their dependecies ,  
> of which we only use 4 (QGpgme, KHolidays, PIMIdentities, Akonadi)  
> for pretty slim features. Knowing this wouldn't it be wiser to allow  
> those who don't need those features to be able to use the application  
> without these dependencies?
> >
> > Marko Käning wrote:
> >     Cristian, I am very much in favour for what you're stating  
> here. This is something which one wouldn't be too worried about on  
> Linux, but on OSX as on Windows in your case it is a real  
> show-stopper to be forced into having far too many dependencies for  
> seemingly little functionality gain.
> >
> >     Thus, I also vote for making things optional!
> >
> > Christian David wrote:
> >     I think, on the long term kdepimlibs will be split up anyway.  
> So the size problem might be solved just by waiting.
> >
> >     Please see my comments as not mainly targeting this review  
> request but more as a general concern about making kdepimlibs  
> optional.
> >
> >     From my point of view this patch adds more complexity to  
> KMyMoney without a real benefit.
> >
> >     All future development needs to wrap the usage of kdepimlibs,  
> which means a lot of work. Also it is hard for a new developer to  
> understand/change it as he has to learn kdepimlibs and the KMyMoney  
> wrapper. Also using #if HAVE_KDEPIMLIBS is uncomfortable from my  
> point of view but maybe I am just too sensitive to that (I worked in  
> project which overdid this — bad memories).
> >
> >     If this has a real benefit for Windows or Mac users it could be  
> a fair deal until kdepimlibs got actually split up. Especially as  
> long as it is a small patch like this one.
> >
> > Cristian Oneț wrote:
> >     I agree that preprocessor definition based conditonal compiling  
> is a PITA. KMyMoney already has KMM_DESIGNER which gave us a lot of  
> headaches but in this case I think that there is a real benefit as  
> long as the conditional code can be kept fairly simple and future  
> development is not hindered.
> >
> > Cristian Oneț wrote:
> >     This weekend I checked out the state of kde framemeworks 5 on  
> Windows. Once I passed the hurdles and managed to install all the  
> necessary frameworks (by fixing some issues) I would have found it  
> much easier to start working on KMyMoney, if this patch would have  
> been in master and frameworrks. Knowing the state of framworks  
> kdepimlibs on Linux I didn't even tried to build it on Windows, just  
> went ahead and ifdef-ed the kdepimlibs dependencies so I could get to  
> the more important issue of KMyMoney on frameworks on Windows.
> >
> >     I guess the situation is similar on OS X. The sooner we would  
> be able to setup a CI build of KMyMoney on all three systems (Linux,  
> Windows, OS X) the better because KMyMoney could act as a real proof  
> that frameworks are really cross platform like Qt.
> 
> That's not a reason to have this as a permanent change. We found the  
> same kind of issues during our port to Qt4, and these will be fixed  
> too.
> 
> If this is going to go in, I'd like a very clear notice in the UI of  
> the features disabled when the dependency is not included. No silent  
> disabling.

I strongly agree with this, because I also support keeping this  
dependency optional.  While I currently run a full KDE system, there  
are pieces I'm getting closer and closer to wanting to avoid.  While  
this may not be realistic for most users, I'd like to minimize the  
dependencies for non KDE users.  (For example, I use Balsa for my  
email, which is a Gnome application, but I have a relatively minimal  
set of Gnome libraries and apps installed.

I think someone else already suggested having a new feature (or is it  
partly existing already) that could produce a run-time list of compile  
information, specifically including which the include/exclude status of  
all optional dependencies.  We could easily require this information be  
included in all bug reports.

Jack