[Digikam-devel] [digikam] [Bug 338176] Face Recognition memory allocation crash application

[Gilles Caulier]

https://bugs.kde.org/show_bug.cgi?id=338176

--- Comment #10 from Gilles Caulier <caulier.gilles at gmail.com> ---
>Is there any way to test the recognition.db for whether it is "intact"? Then I >could maybe find out just at which point it is corrupted again.

Not that i know

>In any case, even with a corrupt database (if this is indeed a part of the >problem), digikam should not behave the way it is doing it now. It should >detect the corruption or the memory excursion and stop whatever it is doing >before it gets too extreme. Maybe somehow fix the database (if the problem is >in the database) or otherwise reset the face recognition if it doesn't "work" >within a reasonable time?

Typically, with my recent changes introduced with digiKam 4.1.0 and 4.2.0, i
wrapped all libkface code that call OpenCV API which handle Histograms (memory
allocation, calculation) with C++ exception. If something is wrong, typically,
digiKam mustnot crash and stop recognition process...

Sound like i forget something or code is incomplete somewhere...

Gilles Caulier

-- 
You are receiving this mail because:
You are the assignee for the bug.