aurupdater

python-pyqt5-datavisualization

Build times (auto-updated)

2026-09-21: SPDX license fix applied, but build currently broken (unrelated)

Fixed license=(GPL3) -> license=('GPL-3.0-only') (no open AUR comment about this specifically – PyQt is GPLv3-only per Riverbank’s licensing, not “or later”; fixed proactively while auditing SPDX issues). pkgrel 5 -> 6.

Discovered a separate, pre-existing, unrelated build failure while testing: sip-build: ABI v12 is being targeted but the PyQt5.QtDataVisualization module doesn't support it. Currently installed sip 6.16.1 / pyqt-builder 1.19.1 default to a newer ABI than this old 5.15.6 binding supports. Not caused by the license edit (pure metadata, no functional change) – build failed identically before and after. Not fixed as part of this pass; would need either pinning sip/pyqt-builder to older versions (messy, affects other python-pyqt5-* siblings sharing the same makedepends) or forcing an older ABI via sip-build --abi-version in build(). Left pkgrel=6/the license fix in place (harmless, correct either way); nothing published this session – still needs the ABI fix before it builds again.

2026-09-24: AUR comment confirms the same ABI v12 issue

jghodd commented on AUR (2026-09-21 21:13) hitting the exact same sip-build: ABI v12 failure independently. No new information – same root cause as above, still needs the sip/pyqt-builder pin or --abi-version decision. No reply posted.

2026-09-24 (2): root-caused and fixed – pkgrel 6 -> 7

Traced the real cause instead of guessing at --abi-version/pinning sip: patched a local copy of sipbuild’s parser (generator/parser/parser_manager.py, _finalise_abi_version) with a debug print (via PYTHONPATH override so the actual installed /usr/lib/python3.14/site-packages/sipbuild was never touched – never modify host-installed software to debug/fix a build) and reran sip-build directly against the extracted source. Confirmed: spec.minimum_abi_versions was an empty list, and the target major (12) correctly matched the installed PyQt5’s own ABI (QtGui.toml says sip-abi-version = "12.15") – so this was never actually an ABI mismatch between our toolchain and PyQt5. minimum_abi_versions is populated only by an explicit %MinimumABIVersion "X.Y" directive inside the binding module’s own top-level .sip file (sipbuild/generator/parser/rules.py, p_minimum_abi_version) – it is not derived automatically from %Import resolution. PyQtDataVisualization-5.15.6’s sip/QtDataVisualization/QtDataVisualizationmod.sip simply never gained that directive – Riverbank’s last release (2022-era) predates when sip-build started enforcing it. So every build with any reasonably modern sip/PyQt5 combination was always going to hit this, regardless of exact version – not an environment-drift issue, a genuine upstream omission.

Fix: fix-minimum-abi-version.patch adds %MinimumABIVersion "12.0" immediately after %Module(...) and before %Import QtGui/QtGuimod.sip – position matters, the parser rejects the directive (“move it closer to the %Module directive”) once %Import has already triggered ABI finalisation. Applied via a new prepare() as a standalone patch file (this project’s convention: source fixes go in .patch files, not inline sed in the PKGBUILD). Verified locally against the extracted source before touching the PKGBUILD: with the patch, sip-build generates cleanly and gets all the way to qmake (which only then fails locally because qt5-datavis3d isn’t installed on this host – irrelevant, it’s a real chroot depends). pkgrel 6 -> 7, pkgver unchanged.

Real chroot test build (repo_build_staged.sh, x86_64) confirms: builds clean, namcap only flags implicit/expected deps (glibc, libstdc++, qt5-base – all transitively via qt5-datavis3d/python-pyqt5) and a false-positive “python-pyqt5 may not be needed” (it’s an API-level dep, not a linkage one, same as always). checkpkg shows no soname differences vs. the currently-published 5.15.6-5.

Built+signed+published cleanly via repo_release.sh: 5.15.6-7 + -debug, x86_64.