[Nepomuk] Review Request 109747: ResourceWatcher: switch to using KDbusConnectionPool

[David Faure]

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
http://git.reviewboard.kde.org/r/109747/#review30038
-----------------------------------------------------------


Found one more issue, I'm afraid.


libnepomukcore/resource/resourcedata.cpp
<http://git.reviewboard.kde.org/r/109747/#comment22385>

    Don't you need to unlock the data mutex first? You wrote in that method's docu not to call it with m_dataMutex locked (OK, unless the RM mutex is already locked, but I don't see that being the case here).
    
    Same problem in propertyRemoved, in fact.


- David Faure


On March 28, 2013, 7:57 p.m., Simeon Bird wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> http://git.reviewboard.kde.org/r/109747/
> -----------------------------------------------------------
> 
> (Updated March 28, 2013, 7:57 p.m.)
> 
> 
> Review request for Nepomuk, David Faure and Vishesh Handa.
> 
> 
> Description
> -------
> 
> ResourceWatcher: switch to using KDbusConnectionPool
> 
> 
> Tidy up ResourceData::propertyAdded.
> 
> If an entry is not found in a QHash, it will return an empty list,
> so we only need to check for contains one way.
> 
> Rework determineUri so that the RM lock isn't held during the executeQuery
> 
> 
> fix lock order in resetAll
> 
> 
> early return if nothing to do (noop, just makes the code clearer)
> 
> 
> Fix bad threading in ResourceManagerPrivate::addToWatcher dbus usage.
> 
> Instead of doing MoveToThread, take the resourceManager mutex before
> talking to the ResourceWatcher. This is a lot clearer and less racy.
> It should not be too contended now the rm mutex is not held over the
> socket communication.
> 
> BUG: 305024
> FIXED-IN: 4.10.3
> 
> ResourceData: Change updateKickOffLists to take three arguments, the uri, old and new values.
> 
> This means that it does not need to be under the dataMutex anymore,
> which in turn means that we no longer need to remember to lock the RM
> mutex before calling it, just to remember to unlock the dataMutex.
> This in turn means that we can add a property that isn't NIE::url() or
> NAO::identifier to the Resource without taking the mutex at all.
> 
> Remove unneeded parameter from ResourceData::resetAll
> 
> 
> This addresses bug 305024.
>     http://bugs.kde.org/show_bug.cgi?id=305024
> 
> 
> Diffs
> -----
> 
>   libnepomukcore/datamanagement/resourcewatcher.cpp f394ae8705fa882b9f4329f93ea7d18c1cfabc73 
>   libnepomukcore/resource/resource.cpp 7158802cf1e1ba48da86f103aa9311c7acbe24fe 
>   libnepomukcore/resource/resourcedata.h 39508764682b798290c1ae5b74a2384f9cbcbf58 
>   libnepomukcore/resource/resourcedata.cpp 7a979745ca22c88a3a73921d9c0e64b51e064c38 
>   libnepomukcore/resource/resourcemanager.cpp 2b32be01176059932cc1c2cf3b1f348ed91e982b 
> 
> Diff: http://git.reviewboard.kde.org/r/109747/diff/
> 
> 
> Testing
> -------
> 
> Compiled, ran, used activities for a while.
> 
> 
> Thanks,
> 
> Simeon Bird
> 
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/nepomuk/attachments/20130328/7a875b26/attachment-0001.html>