[Digikam-devel] Re: [Bug 267789] Make face detection (kface) and geo tagging (kmap) optional features

Alex Jironkin alexjironkin at gmail.com
Tue Mar 15 22:45:59 GMT 2011


K I have just pushed my copy. Everything compiles but still crashes.
To reproduce is simple: just try running face detection from test GUI.


Alex


On Mon, Mar 14, 2011 at 6:37 PM, Marcel Wiesweg <marcel.wiesweg at gmx.de> wrote:
>
>> K, So here is a quick low down on what I did and why. There was a memory
>> leak as the pointer wasn't released when libface::Face was destroyed. So I
>> have added that, which ultimately meant that no Face objects could be on
>> stack, as they would be destroyed and subsequently pointer released. So I
>> have changed the vectors to only contain pointers to the Face objects,
>> meaning pointer wasn't released unnecessary, and would also cut down on
>> amount of memory allocated.
>
> Yes, I think this is a good thing to do. With Qt, we are used not to worry
> about this stuff, but with STL we need to worry.
>
>>
>> This has resulted in downstream changes to interface. Subsequently some
>> changes had to be made in libkface to accommodate for the same reasons. So
>> I have changed a lot of QList<Face> references internally to QList<Face*>
>> . I have made all necessary changes to the libkface and everything
>> compiles fine. Except as I have described before there is a segfault in
>> detection, which I haven't figured out yet. The patch is rather large to
>> submit here because of all the changes, none the less I can still make it
>> and post it here so everyone can patch it up.
>
> Just commit to the libface-2.0 branch of libkface. It's all right to have a
> crash in there. Now that we are on git, we need to take the benefit of easy
> branching ;-)
> If you have your changes as normal local changes in your libkface clone, just
> proceed with
> git stash
> git checkout -b libface-0.2 origin/libface-0.2
> git stash pop
> git commit ...
> git push
>
>>
>> On the different note I have merged the 0.2 branch with trunk so the latest
>> copy is there. There is a 0.1 folder in the tags, if anyone needs the 0.1
>> copy.
>
> Ok
> _______________________________________________
> Digikam-devel mailing list
> Digikam-devel at kde.org
> https://mail.kde.org/mailman/listinfo/digikam-devel
>



-- 
"The generation of random numbers is too important to be left to
chance" - Robert R. Coveyou



More information about the Digikam-devel mailing list