Review Request 123075: do not require X11 on Mac OS X

[Marko Käning]

> On March 20, 2015, 8:07 a.m., Martin Gräßlin wrote:
> > as in other similar requests: -2 from my side
> 
> Martin Gräßlin wrote:
>     To extend: I think the way is wrong. If it now builds on MacOS the required is wrong. It should be an optional find_package properly ifdefed.
> 
> Christoph Cullmann wrote:
>     Actually, you don't want that it is optional as you really don't want that it ever is found on MacOS. If you install an XQuartz for legacy apps, it will be found, and you will have a completly mess as result ;=)
> 
> Martin Gräßlin wrote:
>     Christoph, that argument is wrong. The same would be the case with Windows and any other platform which optionally offers X11 (this includes Linux!). CMake can handle the situation quite well to disable an unwanted build dependency. If a user installs XLib headers (which is not the same as just installing XQuartz) we should expect the user to disable the cmake build option.
> 
> Christoph Cullmann wrote:
>     That means that nobody will get a senseful build on Mac, if he doesn't disable that optional dependency. Thats the opposite of a normal optional dependency, that leads to missing features if not found but not complete mess if found.
>     I tried to compile frameworks stuff on Mac, and IMHO, really, that makes it close to unusable, that you need to tweak each cmake call just to have something usable, if you have too much stuff installed. I never had to do that on Linux/Other Unices.
> 
> Martin Gräßlin wrote:
>     > That means that nobody will get a senseful build on Mac
>     
>     What since when is XLib as a build dependency available by default on OS X?
> 
> Christoph Cullmann wrote:
>     Actually, if you work with Frameworks there, you will install something over MacPorts or Homebrew, and yes, you will have XQuartz installed, to run some legacy apps and there is really no user understandable way to install XQuartz without its devel headers, the .dmg will just install everything, I was suprised myself that I have X devel headers just by installing an X-Server. My first workaround was just to deinstall XQuartz, later I started to tweak CMake options or do exactly the same fixes as above. And yes, I think, per default, without any magic options, frameworks should just build to a usable state. And I see 0 reason to ever have even an optional "with x11" build of an frameworks application. But I might be wrong. Therefore I don't vote here or say ship it, just wanted to state my concerns ;=)
> 
> Martin Gräßlin wrote:
>     my concern is that we make our CMakeLists.txt way more complex (it's not the only framework which recently saw such a proposed change) and work around broken systems in our CMakeLists. That's something I do not want to see - if the downstream packaging sucks, it needs to be fixed there. We would tell the same to every Linux distribution.
>     
>     I do not want to see such OS specific changes go in as X11 is not OS specific - all operating systems we support in KDE can provide X11 and on all OS there are alternative windowing systems. What I don't want to see is something like:
>     if (NOT APPLE AND NOT WINDOWS AND NOT LINUX_ANDROID AND NOT LINUX_UBUNTU_PHONE AND NOT LINUX_SAILFISH AND NOT LINUX_WEBOS AND NOT ...
>     
>     if we start with one platform, where do we end? Is adding the check for Apple OK and Windows not?
> 
> Christoph Cullmann wrote:
>     We could wrap the X11 search in a own module that will not do anything on Apple, to avoid the if stuff.
>     On the other side, even with the if, it still avoids that we have some combinations feasible in the frameworks, that we need will not test anyway, e.g. apple + X11.
>     That avoids complexity, too. Actually I would be OK do have the same for WIN, too, yeah, still better than a code path that you can configure in, that actually will not work.
> 
> Martin Gräßlin wrote:
>     I'm not worried about the code paths - everything X11 specific should be runtime wrapped anyway in a if (QX11Info::isPlatformX11()). If not: that needs fixing.

What's the status of this one?


- Marko


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://git.reviewboard.kde.org/r/123075/#review77779
-----------------------------------------------------------


On March 19, 2015, 11:59 p.m., Harald Fernengel wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://git.reviewboard.kde.org/r/123075/
> -----------------------------------------------------------
> 
> (Updated March 19, 2015, 11:59 p.m.)
> 
> 
> Review request for KDE Frameworks and Michael Palimaka.
> 
> 
> Repository: kdesu
> 
> 
> Description
> -------
> 
> do not require X11 on Mac OS X
> 
> 
> Diffs
> -----
> 
>   CMakeLists.txt 9623483d6f11f9cdb7d7dc19decfd7cf5e86d079 
> 
> Diff: https://git.reviewboard.kde.org/r/123075/diff/
> 
> 
> Testing
> -------
> 
> 
> Thanks,
> 
> Harald Fernengel
> 
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/kde-frameworks-devel/attachments/20150613/494f6fa0/attachment-0001.html>