Frameworks/Epics/Splitting kdelibs

kdelibs splitting epic

Important Forewords

If you are working on splitting a framework, make sure to consult our list of common issues. It will give you an overview of the solutions you're supposed to use to deal with some of the dependencies.

Also, when splitting out a framework, it is part of being done to comply with the KDE Frameworks Policies, it will also need to follow this Epic Policies. Any framework in the staging area cannot move to its final place without following those policies.

And last but not least, if you create a new framework, please consider being its maintainer as well. We need people to ensure the stewardship of all the frameworks.

Existing frameworks

Definition of done:

• No dependency on kde4support, kdecore, kdeui, kio, kparts...
• Library name is in CamelCase
• Modularized CMakeLists.txt [using ECM]
• Installs a Config.cmake file for itself [this is done automatically by extra-cmake-modules]
• Does not install any Find*.cmake modules itself
• Calls feature_summary at the end of the toplevel CMakeLists.txt (see kcoreaddons for an example)
• All Find*.cmake modules needed by the framework and used also by other frameworks have been upstreamed to extra-cmake-modules (via the kde-buildsystem mailing list) or directly to cmake (via the cmake list). To make sure of this, do not append to CMAKE_MODULE_PATH, but set it.
• Does not set EXECUTABLE_OUTPUT_PATH anymore.
• Follows the policy on directory organization
• Unit tested
• Maintains source compatibility over kdelibs
• Does not use any Q_WS_* defines (for Qt 5).
• Does not include config.h.
• Does not use QObject::tr(), but only QCoreApplication::translate() with an empty context, for gettext compatibility (po->ts, no context available). A bit ugly, could be removed later on, with either a PoTranslator (to load .po files directly) or whatever Oswald has in mind for a new Qt translation system.
• See Frameworks/Epics/kdelibs_cleanups for more details
• Document all dependencies in CMakeList between frameworks (example in kauth)

Status	Framework	Tier	Type	Maintainer	Comment
DONE	itemmodels	Tier 1	Functional	Stephen Kelly
DONE	kdbusaddons	Tier 1	Functional	Kevin Ottens
DONE	kplotting	Tier 1	Functional	Benjamin Port
DONE	solid	Tier 1	Integration	Alex Fiestas
DONE	threadweaver	Tier 1	Functional	Mirko Boehm
DONE	kauth	Tier 2	Integration	Dario Freddi
DONE	kcoreaddons	Tier 1	Functional	Romain Perier/David Faure
DONE	kwindowsystem	Tier 1	Integration	Martin Graesslin
DONE	karchive	Tier 1	Functional	Mario Bensi
DONE	kconfig	Tier 1	Functional	David Faure	Contains KConfigGui + KConfigCore
DONE	kidletime	Tier 1	Integration	Dario Freddi
IN PROGRESS	kservice	Tier 3	Solution	David Faure	in staging, tier 3 because of ki18n, could be tier2 otherwise
IN PROGRESS	itemviews	Tier 2	Functional	??	Tier 2 because of natural compare (kcoreaddons) otherwise could be Tier 1
IN PROGRESS	kguiaddons	Tier 1	Functional	??
IN PROGRESS	kjobwidgets	Tier 2	Functional	Kevin Ottens
DONE	kjs	Tier 1	Functional	David Faure
IN PROGRESS	ki18n	Tier 2	Functional	Chusslove Illich	in staging
IN PROGRESS	kemoticons	Tier 3	Solution	??
IN PROGRESS	xmlgui	Tier 3	Integration	??
DONE	sonnet	Tier 1	Functional	Martin Sandsmark	Contains SonnetCore + SonnetUi
IN PROGRESS	frameworkintegration	Tier 4	N/A	Kevin Ottens
IN PROGRESS	kio	Tier 3	Solution	David Faure	incomplete, will contain KIOCore + KIOWidgets
IN PROGRESS	knotifications	Tier 2	Integration	??
IN PROGRESS	kde4support	Tier 4	N/A	??	Can't be completed until kde_file and kglobalsettings are moved into kde4support (they induce a cyclic dependency which prevents turning it into a proper framework)
IN PROGRESS	kde4attic	Tier 4	N/A	??
DONE	kcodecs	Tier 1	Functional	Stephen Kelly
DONE	kwidgetsaddons	Tier 1	Functional	Stephen Kelly
IN PROGRESS	plasma-framework	Tier 3	Solution	Aaron Seigo, Marco Martin	libplasma2 already well under way need to identify and break out candidates that belong in lower tiers (e.g. ConfigReader -> kconfigxt++) QGraphicsView and QML libraries Eventually the runtime bits from kde-runtime and the Plasma QML pieces Source compat is priority, but more relaxed than the rest of Frameworks as QML destroys any realistic possibility of that Porting notes Various identified issue/improvements to be made
TODO	kbookmarks	Tier ?	Functional	??	Right now in kio/bookmarks.
TODO	kcmutils	Tier 4	N/A	??	Right now in kutils/kcmutils
TODO	kparts	Tier 4	N/A	??	Right now in kparts

Backlog

For reference, find the class by class analysis produced during Platform11 on the kdelibs dependencies page

This list is non-final, hence why it is not integrated in the table above, when the scope of a lib gets defined and worked on, it is removed from this list and go in the table above.

Yet to be produced frameworks (foreseen tier and type, not set in stone):

• dnssd (easy one!)
• libkdebug (Tier 3 / Functional) / Depends on the results coming from qDebug() in Qt5
• kcompletion (all the KComboBox, KLineEdit completion features + KCompletion* classes)

And probably more...

Finalization tasks

To be executed to complete the epic when all the frameworks are "done":

• Make sure it builds and runs against Qt 5.2 final
• Split into its own git module; pre-existing history will remain in the kdelibs repository and be accessible with git graft
• Produce a script or cmake file to recreate a kdeexamples tarball out of a set of examples directory in the frameworks

Policies

Strive for Source Compatibility

Avoid removing API from the frameworks branch, even if it is replaced or deprecated. Prefer to implement the old API in terms of the new API.

In some cases this is very easy. For example, many APIs in KGlobal will be replaced as part of the frameworks effort.

• KGlobal::charsets() -> KCharsets::charsets()
• KGlobal::dirs() -> KStandardDirs::global() (or something?)
• KGlobal::ref() -> qApp->ref()
• KGlobal::deref() -> qApp->deref()

etc.

Removing this API from KGlobal is not necessary in most cases, and it introduces a porting burden which is best avoided.

There is unlikely to be a 1:1 mapping of old API to new API in every case, so how to implement the old API may not always be obvious. However, attempting to ensure that the older API remains useful should be at the front of our efforts and standards in the frameworks branch.

Binary compatibility policy relaxed

During the execution of this epic, the binary compatibility contraints from the global policy list is lifted.