Need suggestion on how to fix the common crash in plasma-desktop (kdelibs related)

[David Faure]

On Wednesday 21 March 2012 09:06:05 Lamarque V. Souza wrote:
> Em Wednesday 21 March 2012, David Faure escreveu:
> > On Tuesday 20 March 2012 22:31:57 Lamarque V. Souza wrote:
> > > WeatherEngine::~WeatherEngine() calls
> > > WeatherEngine::unloadIons(), which tries to use the invalid
> > > Plasma::DataEngineManager::self().
> > 
> > Since DataEngineManager uses K_GLOBAL_STATIC internally, just use the
> > isDestroyed() method to know if it has already been destroyed, before
> > calling self().
> 
> 	I thought about doing that, but I would like to prevent an overhead that
> is needed only when plasma-{desktop,device} are exiting.

Checking one bool can hardly be considered "overhead".

> > More precisely, either add a isDestroyed method to the public class, which
> > calls the one in the K_GLOBAL_STATIC, or let self() return 0 when
> > privateDataEngineManagerSelf.isDestroyed().
> 
> 	The Plasma::DataEngineManager::valid() I suggested is exactly returning
> the negation of privateDataEngineManagerSelf.isDestroyed(). I just wanted to
> know if there were other options. Ok, let's add a
> Plasma::DataEngineManager::isDestroyed(). Can I backport that to 4.8? Wrong
> question, I will have to it to 4.8 so that somebody forward ports it to
> master, duh.

Yep.

> > On a more philosophical note: this is exactly why "intelligent
> > destructors"
> > are to be avoided at all costs. All this wouldn't happen if the
> > WeatherEngine destructor didn't call methods that do stuff.
> 
> 	At least stuff with other object's pointer.

Yep.

-- 
David Faure, faure at kde.org, http://www.davidfaure.fr
Sponsored by Nokia to work on KDE, incl. KDE Frameworks 5