Review Request 125259: Support multiple X servers in the NETWM classes

[Martin Gräßlin]

> On Sept. 18, 2015, 2:21 p.m., Thomas Lübking wrote:
> > src/platforms/xcb/netwm_p.h, line 31
> > <https://git.reviewboard.kde.org/r/125259/diff/2/?file=404639#file404639line31>
> >
> >     errrhemmm ... the idea was actually to keep things more in sync and not have to match variable, string assignment and netAtomCount by hand, so I would have actually gone for including an atom header twice on different macro definition.
> >     
> >     Also the presented system drops scope (KDE_NET_WM, NET_WM, WM and ... UTF8_STRING ;-)
> >     
> >     Sorry :-(

sorry, now you have lost me with what you actually want.


- Martin


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://git.reviewboard.kde.org/r/125259/#review85619
-----------------------------------------------------------


On Sept. 18, 2015, 10:19 a.m., Martin Gräßlin wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://git.reviewboard.kde.org/r/125259/
> -----------------------------------------------------------
> 
> (Updated Sept. 18, 2015, 10:19 a.m.)
> 
> 
> Review request for KDE Frameworks.
> 
> 
> Repository: kwindowsystem
> 
> 
> Description
> -------
> 
> So far on first creation of a NETRootInfo or NETWinInfo a static
> initialization of atoms was performed. This meant that the NET classes
> are only able to interact with the X server the first NET class is
> connected to.
> 
> Normally applications don't need to interact with multiple X servers.
> An exception is kwin_wayland which needs a connection to its Xwayland
> server and on the x11 backend to the X server it renders to. So far
> KWin could not use the NET classes for it, causing the rendering window
> to e.g. not have a window title.
> 
> This change removes the implicit constraint on one X server by
> creating a hash of connection and atoms. For each created NET class
> we check whether we have already resolved the atoms, if yes we reuse
> otherwise we create them.
> 
> For the normal use case of one X11 connection the change should be
> rather minimal. Instead of performing the check whether the static
> atoms have been created, it now is a check whether the atoms for the
> connection have been created. The atoms are kept in a
> QSharedDataPointer ensuring that we don't needless copy the atoms into
> each class.
> 
> CHANGELOG: Allow interacting with multiple X servers in the NETWM classes.
> 
> [autotests] NETWM tests get a new X server per test
> 
> Making use of new feature of allowing multiple X connections:
> start X server in init() instead of initTestCase().
> 
> 
> Diffs
> -----
> 
>   autotests/netwininfotestwm.cpp a98e12fd690b0250337c7588e1525af1d0dda38c 
>   autotests/netrootinfotestwm.cpp e918873a5281f6ecbcc1769de3dadcf8f6222f6a 
>   autotests/netwininfotestclient.cpp a5b10376b943ea914a9521b5c07f9ef13a65d2f1 
>   src/platforms/xcb/netwm.cpp d99a925ad2b99d5e60ab1dafcb01400b4a6a4c93 
>   src/platforms/xcb/netwm_p.h 917a86ed5b6c83f36e73bbc346360b943d457243 
> 
> Diff: https://git.reviewboard.kde.org/r/125259/diff/
> 
> 
> Testing
> -------
> 
> see adjusted unit tests. Also tried it in kwin_wayland
> 
> 
> Thanks,
> 
> Martin Gräßlin
> 
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/kde-frameworks-devel/attachments/20150918/429ff82c/attachment-0001.html>