aurupdater

plasma-framework5

Build times (auto-updated)

2026-09-21: qt5-doc makedepends investigated (not removed) + kirigami2 gap found & fixed

User asked to check whether qt5-doc (in makedepends, alongside kdoctools5/doxygen, supporting this PKGBUILD’s -DBUILD_QCH=ON) is actually needed to build. Investigated in a scratch copy (/tmp/.../scratchpad/plasma-framework5-qt5doc-test, NOT /data/INSTALL, NOT this tracked PKGBUILD) with qt5-doc stripped from makedepends.

Result: not required. Build succeeds cleanly without it. CMake’s ecm_add_qch() just logs and skips Qt’s own cross-reference targets: -- No such target Qt5Gui_QCH defined when calling ecm_add_qch(), ignored. – KF5Plasma.qch/.tags still get generated (via doxygen) and installed fine. Losing qt5-doc only means the generated .qch docs lack working hyperlinks into Qt5’s own API docs (e.g. QObject references won’t cross-link) – everything else is unaffected. The real PKGBUILD was deliberately left unchanged (investigation-only, per how the task was framed) – qt5-doc is still in makedepends as of this writing. Remove it if/when desired.

Side discovery while running that scratch test: kirigami2 (a hard depends=) turned out to be completely missing from both the official extra repo (dropped in KDE’s KF5->KF6 migration) and this arch/ tree – the scratch build could only resolve it by injecting a stale cached .pkg.tar.zst still sitting in this host’s pacman cache (makechrootpkg -I). That meant the real published plasma-framework5-5.116.0-2 was itself silently relying on that same stale cache the last time it was built, undetected until now. Fixed properly: built and published kirigami2 from its (orphaned but current) AUR package as [kirigami2] – plasma-framework5 rebuilds now resolve it from the real repo instead of a host-local cache artifact that could vanish at any time (e.g. a pacman -Scc).

2026-09-21 (same day): qt5-doc actually removed and confirmed (pkgrel 2 -> 3)

User asked to go ahead and remove qt5-doc from the real makedepends and confirm via a real rebuild. Done: makedepends=(extra-cmake-modules qt5-tools kdoctools5 doxygen), pkgrel 2 ->

  1. First rebuild attempt hit a transient, unrelated snag: the real pipeline now needed to install the freshly-published archlinuxaba/kirigami2 as a runtime dep, but a stale/corrupted kirigami2-5.116.0-2-x86_64.pkg.tar.zst was sitting in the host’s own /var/cache/pacman/pkg/ (leftover from this same session’s earlier scratch-test makechrootpkg -I injection of the pre-drop official build, byte-different from the freshly-published AUR rebuild despite matching pkgver-pkgrel) – pacman flagged it as corrupted and aborted. The file was gone on the very next check (pacman’s own “delete it?” prompt must have removed it), so a bare retry succeeded with no further changes needed.

Confirmed: real-pipeline build succeeded cleanly without qt5-doc, published plasma-framework5-5.116.0-3-x86_64 to archlinuxaba. No further qt5-doc/QCH-related issues.