BTW, am I missing something or shouldn't this entire kioslave lookup be made static? It only needs computing once but currently is run for every new slave that gets started (in fork-mode).

Would it make sense to just install it in libexec and remove the difference between platforms?

Generally yes, I am not sure we can do that very well for 5.x though. The LIBEXECDIR_KF5 where kioslave is installed is controlled by ECM, that's where the conditional split between suffix-or-not happens, so unless we add another LIBEXECDIR_KF5_REAL which does away with the conditional and then explicitly install kioslave using LIBEXECDIR_KF5_REAL I don't think we can fix it. (and a second LIBEXEC var may only be confusing; the thought of it certainly doesn't excite me).

from KDEInstallDirs.cmake

if(WIN32)
    _define_relative(LIBEXECDIR BINDIR ""
        "executables for internal use by programs and libraries"
        LIBEXEC_INSTALL_DIR)
    _define_non_cache(LIBEXECDIR_KF5 "${CMAKE_INSTALL_LIBEXECDIR}")
else()
    _define_relative(LIBEXECDIR LIBDIR "libexec"
        "executables for internal use by programs and libraries"
        LIBEXEC_INSTALL_DIR)
    _define_non_cache(LIBEXECDIR_KF5 "${CMAKE_INSTALL_LIBEXECDIR}/kf5")
endif()

Oh, we could maybe use GNUInstallDirs.cmake's LIBEXECDIR on !Win32?

like so:
https://phabricator.kde.org/P317

I am not sure if there are any downsides to this. CMake code-wise I am not sure we should workaround implementation details of ECM in such a way.