Hosted by Kevin Ottens and the KDE Frameworks crew.
Notes taken during the meeting on etherpad (not really, network not stable enough).
There is no branch, everything happens in master.
Feature branches: Should be very short lived, if they live longer than 2-3 weeks look at splitting them into smaller changes. Might not be best suited for ReviewBoard review, alternative review approaches are allowed. Rebase/split into independent commits is possible, force pushing is currently disabled on the main repos (possible on personal clones, or with new branches).
Lack of long term stable branches caused concerns from packagers initially, try it nevertheless for now and see how well it works.
Marco suggested a hook to check for the REVIEW tag, with a way to by-pass it quickly with "REVIEW: trust-me", just to force you to think about review. We want that.
Mandatory all-time review vs. non-review bypass option?
Still low on tests, people would want more.
Push for more tests during reviews, for new stuff.
Make sure automated tests work without installation. Tricky due to ksycoca and QStandardPaths.
Extra/betters tools for tests? See Shantanu's talk from yesterday and Zanshin for mock examples, currently not used in KF5 (yet).
Coverage not currently enabled on Jenkins, we would like to have that enabled for KF5. Needs an option to set the coverage build flags (got removed from ECM). Does Jenkins have enough computing power? Aleix will look into it.
Unstable tests? Either make stable or disable, there is no other way around that.
Currently Qt code can't be copied into KDE lacking the LGPLv3 compatibility, but that is going to change. Do we need/want to adjust the license policy? Currently not really needed, wait until the need for this actually comes up.
Followup to the discussion in Randa on this topic.
CI/Jenkins skills missing, for:
Script for automagic environment setup to install distro-provided dependencies and build the rest, possibly based on kdesrc-build.
Documentation should be linked inside Qt documentation/Assistant.
kdesrc-build focus on contributor use, inqlude focus on KF5 users.
KDE specific tools in KDevelop (access to recent bugs, reviews, etc).
Aleix working on CMake support.
Needs CI, running tests is tricky though.
Distribution channels: Aleix has a proof-of-concept for F-Droid. KDE could have its own F-Droid repository. Alternative: Play Store, less FOSS-friendly, might still have legal risks, but we could charge money there. Publish via single KDE account (via KDE e.V.)? In any case, we want a single recommended way for getting KDE apps into some Android app store. Details to be discussed, includes technically, legal and organizational issues.
Small Patrick is working on Windows support on Jenkins (using an old workstation at Large Patrick's place, might need more power later), biggest challenges are porting the Unix CI scripts, and KDE CI specific Qt patches.
Someone is working on the same for Mac.
Work on book started in Randa, goal is one chapter per framework. So far: 4-5 chapters/36 pages exist. Can be found in Git, has CMake build system and will get CI. Aim at having a printed book at Qt Dev Days, more chapters needed though.
Marco, Martin G., Martin K., Luigi, David E., David F. each volunteer to write at least one chapter. Tutorials on techbase might be a good source of content. Sebas wrote readmes per framework, might provide useful intro texts for the chapters.
techbase vs. code distance, can the tutorials be maintained next to the code in Git and docs generated from that?
Future of techbase, does the book replace it? No, but some content is being superseded/archived for KDE4. Some restructuring needed.
Partially covered by the reworked api.kde.org and the book. Improved landing page? Websites tend to get outdated quickly though.
Better home pages for frameworks instead of pointing to projects.kde.org (there's a task for that already).
Code sharing between file managers. David F. working on merging this into KIO.
FolderView has a copy of libkonq, goal is to remove that.
Still left to do:
Plan from fd.o summit, same approach from the shared mime database. Needs glib-based code to be written, progress so far only during fd.o summit. Problem due to delay: compatibility with already released KF5 code. This so far only covers application desktop files, but plug-ins and services aren't covered (but we assume to can reuse the application part for that). Alternatively we could also move that part forward independently, as there is no point in sharing that with the glib world. Plasma packages also need that. Becomes more of an issue as more and more Plasma packages/plugins appear/are installed.
For KF5 split of kdepimlibs plan see Notes from PIM Bof earlier today.
How and where do we handle KF5 communication with kdepimlibs (and possibly others) joining? Domain discussions can stay where they are, framework coordination/structural discussion can stay together.
Fredrik wants to kill the system bill and simplify the a11y KCM, some files miss license headers there though.
Runtime part currently in workspace, X11-only (no Wayland) at the moment. Win/Mac not supported/ported. LxQt wants to use this, that would need to move out of workspace. Would increase tier from 1 to 3 though, tier 2 without KCrash (needed for restart). Daemon only needed for Mac, would work on Linux/Windows without that (thus removing the D-Bus dependency there). On Linux, the crash restarting could be done via systemd.
Upstream crash auto-restart feature to Qt would address the KCrash dependency.
Let's move it, or is the tier increase a problem for anything using kglobalaccel on a lower tier (unlikely)? GPL for the daemon is not a problem.
Windows teams is working on removing the old complicated UI from the installer and replacing it with something much easier.
See parallel BoF.
Kevin K replaces Kevin O as KF5 project manager.