I'm fixing the problem differently in D24609

Looks fine to me, thanks for the patch.

In D16610#418802, @rakuco wrote:

I've sent a patch upstream with a more general approach: https://sourceforge.net/p/libpgf/patches/12/

I don't think this should be a FreeBSD-specific check; even on other systems where the build is not currently broken we shouldn't be redefining nullptr anyway. I've sent a patch upstream with a more general approach: https://sourceforge.net/p/libpgf/patches/12/

In D17606#377854, @dvratil wrote:

Maybe we could just get rid of the optional in this agent until we can switch to C++17 proper. I hate not having it, but it's just one or two occurrences here and probably not worth the magic to satisfy all compilers.

I've finally posted a more comprehensive patch in D14146 -- so much time passed that I lost my original branch and did something similar in spirit :-)

I agree this can be abandoned -- whatever solution we agree upon should probably be done in plasma-desktop.

This isn't just a problem on KDE Neon though, is it? I thought FreeBSD is also affected?

Per my previous comment, I still don't see how changing this to a target would solve anything.

@kossebau's right: the problem lies in lookandfeeltool depending on kcm.cpp, while the kcoreaddons_desktop_to_json() call makes the kcm_lookandfeel target depend on the generation of the json file. It's pretty easy to reproduce this bug by running make lookandfeeltool_autogen with a fresh build directory.

@jriddell @kossebau: ping?

In D8062#150767, @adridg wrote:

FindGraphviz.cmake uses DOT as well for finding the version of GraphViz that is installed -- not that the current module or code supports checking hthat version, but it's there.

Given the amount of commits to the repository in the past years I'm not sure you'll be able to review this patch. Perhaps someone with a KDE Multimedia hat?

See also: https://git.reviewboard.kde.org/r/129246/

Note that at least glibc defines strtod_l in stdlib.h, I think it's safer to include both stdlib.h for glibc and xlocale.h for OSX and the BSDs.

Isn't it better to use check_cxx_compiler_flag to see if the flag is supported and enable it in case it is?

Looks great to me now; I'll leave it to Martin to accept the diff.

Looks good to me, but @cordlandwehr should probably take a look at this too.

The patch makes sense to me. Since QTGSTREAMER_INCLUDES already contains QTGSTREAMER_INCLUDE_DIR, you can probably remove the latter from include_directories().

How about using the same calls present in Modules/memory/memory_fbsd.cpp (basically read the contents of the hw.physmem sysctl)?

I've committed this to the 1.7 branch, with the required changes. Will you guys merge it to 5.0 or should I take care of it?