Some CMake methods needed adjustments:
https://invent.kde.org/frameworks/kpackage/-/commit/ed65dc7b7a173490044d411267526273a41d0c2f
https://invent.kde.org/frameworks/plasma-framework/-/commit/6e53dcc2d25089f35988eacbbfe6e3e189690db4

https://invent.kde.org/utilities/kate/-/merge_requests/445
https://invent.kde.org/multimedia/k3b/-/merge_requests/13

https://invent.kde.org/sdk/dolphin-plugins/-/commit/ce564d08fe31dd52790897a084bb0a81441f7c09

One aspect to consider is that desktop files are IMHO a bit nicer for humans to edit than json, so we might want to still use those as source but install only the json files

I disagree that it is nicer to edit, especially because some of the property names are changed after the JSON conversion and some are not convertet at all. Like the mentioned X-KDE-Library in T14757.

I think that the keys used should be the same as their accessor in KPluginMetaData. Like the comment key in the desktop files is called description in the JSON metadata.
Also in the JSON files multiple authors can be put inside of the JSON array, int the desktop file there is only a single X-KDE-PluginInfo-Author field.

Also in quite a lot of places there are already JSON files in VSC, converting the rest of them would make it more consistent.

And as discussed in the other tasks the service types should disappear, but they are currently (almost exclusively) used for the JSON file conversion. There are also issues with the desktop conversion, for example translated strings are not handled very well or the correct service types could not be found during the build. This was an issue when I introduced a new KRunner property which was a bool.

Notes from KF6 weekly:

• https://phabricator.kde.org/T14564: There are a number of good reasons to consistently use JSON files, desktop files for editing might be nicer, but those files are rarely edited and mainly contain scripty generated content.

I'm confused, what should I do if I'm porting an old KCM?
Should I remove the desktop file all together and install json instead? If so, where can I get the json file.
Or should I convert the desktop file to json but still install the desktop file?
Or should I convert the desktop file to json and install json, but still keep the desktop file?

I've seen your MRs, but I'm confused why you added desktop to json conversion and then remove it in the next MR

You should install a desktop file with the translated name and a few other keys, see https://invent.kde.org/plasma/plasma-workspace/-/blob/master/kcms/autostart/kcm_autostart.desktop#L40 until the end of the file.

The KCM should have embedded json metadata, which contains additional info like the systemsettings categories, description & keywords.

If so, where can I get the json file.

desktoptojson -i myfile.desktop

but I'm confused why you added desktop to json conversion and then remove it in the next MR

I wanted to do the json conversion in one go, otherwise life would have been more difficult for translators. The json conversion code was needed to port the deprecated systemsettings KCM querying. Because this was a big change across multiple repos already, I wanted it to get merged before working on the conversion.

If you need further help, I can push a change to a Mr/branch if you like

If you need further help, I can push a change to a Mr/branch if you like

FWIW I haven't found editing a JSON file to be any worse than editing a .desktop file. My preference would be to autogenerate all the relevant autogeneratable .desktop files and have a single JSON file as only one source of truth.

For reference: The CMake function kcmutils_generated_desktop_file landed a while ago. For KCMs we now only have a json file for the metadata.

I think we can consider this done from a frameworks perspective. There are still apps that need to be fixed though