qgpgme1-qt5
- Path: arch/maintained/qgpgme1-qt5
- 2026-07-28: failed in the routine
repo_release_changed.sh batch –
Missing dependencies: gpgme-1, qt5-base. gpgme-1 is a package in
this same tree (arch/maintained/gpgme-1), not an official-repo
package, so archlinuxaba-x86_64-build can only resolve it if it’s
already published to the private repo. It failed to build in the same
batch run (see memory/gpgme-1.md – transient upstream 429 rate
limit on dev.gnupg.org, not a real defect), so it was never
published, so this package’s dependency resolution failed too. This
is the documented intra-tree ordering gotcha
([[private-repo-changed-only-builds]] / TOOLING_NOTES.md), not a
problem with qgpgme1-qt5 itself – retry once gpgme-1 is
successfully published.
- 2026-07-29: retried after
gpgme-1 1.24.3-5 (github-mirror fix)
published for x86_64. Dependency resolution now fine, but hit a real
packaging bug: package() only ran make DESTDIR=... install inside
lang/qt, never installing lang/cpp. The install-time libtool
relink of libqgpgme.so needs -lgpgmepp resolvable under
$pkgdir/usr/lib – gpgme-1 (this package’s depends=) does ship
the runtime libgpgmepp.so.6.21.3 + versioned symlink, but
deliberately no unversioned dev symlink (see memory/gpgme-1.md’s
package() – headers moved to a nonstandard gpgme1/ prefix, no
pkgconfig/cmake, runtime-only by design), so linking -lgpgmepp
failed with cannot find -lgpgmepp.
- Fix (pkgrel 1 -> 2):
package() now runs make -C lang/cpp
DESTDIR="${pkgdir}" install first (installs lang/cpp’s
libgpgmepp into $pkgdir purely so the relink step can find it),
then installs lang/qt as before, then strips the cpp bindings
back out (usr/include/gpgme++, usr/lib/libgpgmepp*,
usr/lib/cmake/Gpgmepp, usr/lib/pkgconfig/gpgmepp.pc) before the
existing QGpgme -> QGpgmeQt5 cmake rename – gpgme-1 already
ships the actual runtime lib, no need to duplicate/conflict with it.
Verified the built x86_64 package only contains libqgpgme.so*,
no leftover libgpgmepp*. Published successfully.
- Worth checking whether
qgpgme-qt5 (the modern gpgme-2.x sibling,
which uses cmake --install build for the whole tree instead of a
subdirectory make install) is naturally immune to this same class
of bug, or just hasn’t hit it yet.