Summary:
When using a toolchain file one often would want to set a
CMAKE_FIND_ROOT_PATH to "re-root" lookups into a staging area to prevent
native host artifcats from being found. This in particular is usually the
case for lib/include/package finders as one would want the target
systems build artifacts rather than the hosts.
find_file() unfortunately draws its re-rooting rule from
CMAKE_FIND_ROOT_PATH_MODE_INCLUDE, which as mentioned would usually be
set not to look in the actual host paths to avoid picking up wrong includes
which would also prevent find_file to work here.

In the case of the coreaddonstooling it makes sense to look in the host
though. For one the lookup only runs when CROSS_COMPILING is set (which
usually is the case for toolchains), AND one explicitly has to specify a
HOST_TOOLING path. So in a way by specifying that one is already opting
out of whatever rooting rules one has set elsewhere as setting that var
is a clear indication that we have *specific* host tooling and we want
to use that rather than anything else.
As a result NO_CMAKE_FIND_ROOT_PATH is the way to go when looking for stuff
under a KF5_HOST_TOOLING conditional.

Reviewers: mdawson, apol

Reviewed By: apol

Subscribers: #frameworks

Tags: #frameworks

Differential Revision: https://phabricator.kde.org/D5218

Spotted because it leads to race conditions due to tzset
(surely a Qt bug, but anyway no point in finding out lastModified
for nothing).

While at it, remove the unused size field as well.

For some reason some people have kdeglobals entries like
   Close=Ctrl+W; Ctrl+Esc; Ctrl+W; Ctrl+Esc;
having the same shortcut more than once in the shortcut declaration is clearly bogus so fix it

BUGS: 375555
REVIEW: 129987

Differential Revision: https://phabricator.kde.org/D4941

Summary:
It can't possibly have been used anywhere, because it's not in the KConfig
API anywhere. The intended way to use this API was by providing a plugin
that would derive from KConfigBackend but the plugin loading code
in KConfigBackend::create has been disabled since before KF 5.0.

The reason I want to stop exporting this class is to be able to
optimize it (e.g. the QDateTime in it is completely unused but leads
to data races due to tzset)

Test Plan: Note, I'll rename it to _p.h if we agree, it would just have made this diff too big.

Reviewers: mdawson

Reviewed By: mdawson

Subscribers: #frameworks

Tags: #frameworks

Differential Revision: https://phabricator.kde.org/D4604