[Ktechlab-devel] Project components dependencies

[P Zoltan]

On Tue, 26 Jan 2010 11:28:47 +0100, Julian Bäume <julian at svg4all.de> wrote:

> On Sunday 24 January 2010 01:50:34 P Zoltan wrote:
>>   Meanwhile I've found a huge problem with the qvector header: Qt has a
>> header with the same. And it also tests inside if QVECTOR_H is defined.  
>> So
>> I've got several very strange build errors... Due to this, I've renamed
>> the math files to quickvector and quickmatrix.
> Eeks, that's indeed a problem ;) another solution would have been to use  
> a
> proper namespace for that. But I'm fine with renaming the files, also.
>

The problem is that namespace won't help with defines. Anyway, it works  
now.

> Let me comment on your work. At first, it didn't compile for me.
> quickvector.cpp uses memset, which is declared in <cstring>. After  
> including
> it, it compiles and runs fine.

  Added it to the repository. I'll upload it soon.

>
> Next step would be to integrate this into the kde4-port branch. This  
> should be
> done without the eigen sources. We should just add a (soft) dependency to
> eigen for the tests. And built the test, when eigen is found. I could do  
> that
> in the cmake files. Shouldn't be much of a problem. Then we should merge  
> your
> changes to the maths classes into the kde4-port branch.

The situation is that Eigen is purely header-based implementation; it has  
no lib to link against. All the classes are built inside the final program  
(this allows serious optimisations). So the chances of having Eigen  
sources at somebody's system are very close to zero. I consider a waste of  
time to separately download Eigen and set its path.

  As now I consider the tests good-enough, I'll try to integrate it with  
the rest.

  The next interesting operation should be creating a test for the document  
and simulator, together, so simple circuits can be tested. If those tests  
are ready, we could try to integrate the simulator with the qt4 port; the  
problems should be revealed fast by using the testcases.

>
> Another minor issue, i mentioned, while reading the patches (git log -p).
> There are a lot of trailing white-spaces in your code. Your should try to
> avoid them. What editor are you using? Kate can be configured to  
> visualise
> trailing-white-space and automatically remove them on save. Same counts  
> for
> vim. You can also see these in git diff (before commit, just have a  
> short look
> into the diff). You need to do some git configuration to see it, though:
>
> color.ui=true
> core.whitespace=fix,-indent-with-non-tab,trailing-space,cr-at-eol
> core.pager=less -FXRS
>
> I think, these options should do the trick.

  I should check those configurations; also learning git more deeper would  
be necessary, as there are some open questions for me, like how to remove  
a file from a planned commit (in case of an accidental git add); also how  
to keep track of remote branches and push / pull from them in an efficient  
manner. All these should be described in the wiki. Also an idiot-proof git  
checkout guide is needed, as there is some working code in the repository.

  Another question about sourceforge: is there any read-only access mode  
with uses some more common port? git:// port seems blocked from me, so I  
have to use ssh in order to download the changes; that's not very  
convenient for me.

  About the trailing spaces: I created them by using gedit :D Now I'll  
start using kate; hopefully the visibility of the trailing spaces will  
reduce the number of such problems.
>
> Well :) enough for now.
>
> bye then
> julian

  Zoltan