[Nepomuk] [Kde-pim] nepomukqueryservice crash

[Volker Krause]

On Sunday 08 July 2012 09:40:32 David Faure wrote:
> On Saturday 07 July 2012 11:13:08 Volker Krause wrote:
> > On Friday 06 July 2012 14:44:30 David Faure wrote:
> > > I found the reason for the kmail composer window being disabled forever,
> > > in
> > > some cases.
> > > 
> > > A blocking nepomuk query from an akonadiserver thread never terminates,
> > > because nepomukqueryservice crashes meanwhile.
> > > 
> > > This leads to two bugs:
> > > 
> > > 1) queryserviceclient should kill blocking queries when
> > > _k_serviceUnregistered tells us the service crashed.
> > > The attached patch works for me, OK if I commit it to all appropriate
> > > branches? (kdelibs 4.8, kdelibs 4.9, nepomuk-core, akonadi master --
> > > LOL)
> > 
> > +1 as far as Akonadi is concerned. We don't have _k_serviceUnregistered at
> > all though, so this might require backporting a few more changes of the
> > queryserviceclient code.
> 
> Yep, I have these commits ready for akonadi. Just waiting for an OK from the
> nepomuk side.

great

> BTW, does search/nepomuksearchengine.cpp ever delete QueryServiceClient
> instances? It seems to me that reloadSearches() leaks all the
> already-running queries.

reloadSearches() calls removeSearch() for all existing QueryServiceClients, 
which ends with a call to deleteLater(), so this looks ok to me.

There is a theoretical scenario (existing query but no existing corresponding 
collection) where it would leak indeed, but that should never happen, so 
adding extra safety checks for that is probably not worth it. Or did you see 
any indication it's actually leaking?

regards,
Volker
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 190 bytes
Desc: This is a digitally signed message part.
URL: <http://mail.kde.org/pipermail/nepomuk/attachments/20120708/02501669/attachment.sig>