kde4support: kstandarddirs_unittest fix

[Ben Cooksley]

On Tue, Dec 24, 2013 at 5:37 AM, David Faure <faure at kde.org> wrote:
> On Monday 23 December 2013 14:42:45 Alex Merry wrote:
>> On 23/12/13 14:23, Alex Merry wrote:
>> > On 23/12/13 12:52, David Faure wrote:
>> >> On Monday 23 December 2013 11:43:39 Alex Merry wrote:
>> >>> Either way, the tests will fail is KDEDIRS is not set correctly; I think
>> >>> the right approach there is to do a bit of a sanity check in
>> >>> initTestCase() and bail (with either QSKIP or QFAIL) if it is either
>> >>> empty or does not contain at least the install directory.
>> >>
>> >> Or we can just set it with a qputenv? ;)
>> >
>> > Setting it correctly is the tricky part.  CMake doesn't seem keen on
>> > giving you the install prefixes of other packages.
>>
>> Ah, I should say that the reason setting it correctly is tricky is
>> because it turns out it doesn't actually need to contain the
>> installation directory.  The important thing is that it contains the
>> directory where KConfig is installed, unless that *is* the installation
>> directory.  Basically, it only needs to be set if you're installing the
>> frameworks in different directories, like Jenkins does.
>
> Sounds to me like jenkins itself should be setting it.
>
> I'm actually pretty sure it does already, otherwise a lot of stuff would never
> be found (like the kdoctools entity file...)

The CI scripts set KDEDIRS, XDG_DATA_DIRS and XDG_CONFIG_DIRS as
expected for each prefix.
This includes the prefix where the installation is staged to prior to
it's full deployment.

Those interested in the specifics of what it does should look at
kde:websites/build-kde-org, branch production, file tools/kdecilib.py
from line 431 onward.

>
> Ben?
>
> --
> David Faure, faure at kde.org, http://www.davidfaure.fr
> Working on KDE, in particular KDE Frameworks 5
>

Thanks,
Ben