gpgmepp

Category: local-install (rebuilt only so archlinuxaba/the cluster gets it too; not maintained/adapted/owned by us — see arch/scripts/install_only_packages). Built to unblock qgpgme-qt5’s i686/pentium4 extension, which in turn unblocks trojita-qt5-git’s.

Official Arch gpgmepp PKGBUILD (Antonio Rojas), C++ bindings for GPGME.

Build notes (2026-08-20)

• User pulled the official upstream PKGBUILD via pkgctl themselves (already had arch=(x86_64 i686 pentium4) set).
• Source unreachable: dev.gnupg.org (upstream’s git host) refused connections outright (verified directly from this host, not chroot-specific). Switched to the official release tarball instead (https://gnupg.org/ftp/gcrypt/gpgmepp/gpgmepp-2.1.0.tar.xz, reachable), re-hosted on our own src/ mirror per [[repo_src_mirror]] rather than left depending on gnupg.org directly. build()’s -S path adjusted from gpgmepp (git checkout dirname) to gpgmepp-2.1.0 (release tarball’s top-level dir).
• x86_64 built and published cleanly. pkgrel 1 -> 1.1 (local-install, X.Y scheme) for the source swap.
• i686/pentium4 initially blocked: gpgmepp itself requires Gpgme >= 2.1.0 (its own CMakeLists.txt find_package requirement), but archlinux32 only ships gpgme 2.0.1-1.0 for 32-bit (confirmed via pacman -Qi in both chroots and the archlinux32 mirror listing) – a genuine version gap on their side, one level deeper than the qgpgme-qt5/gpgmepp gap this build was meant to fix. Resolved by building gpgme 2.1.2 ourselves for 32-bit too (see memory/gpgme.md) – once that published, gpgmepp’s i686/pentium4 builds succeeded cleanly on the first retry, no further changes needed. All three arches now publish.

Infrastructure note: repo host /tmp too small for repo-add

While publishing this package, repo-add failed with Write failed: No space left on device – NOT the repo’s actual storage (/data, 626G free), but the repo host’s /tmp, a tiny 32M tmpfs that repo-add/bsdtar extracts the whole package database into to rewrite it. This session’s cumulative package growth apparently pushed the extracted db size past 32M for the first time. Fixed by adding TMPDIR=/data/tmp (created once, chmod 1777) to both repo_add_remote() and repo_remove_remote() in arch/scripts/repo_lib.sh. Verified no earlier publish this session was silently lost – spot-checked several recent packages (php74, trojita-git, smp_utils, qgpgme-qt5, ack-git) directly in the remote archlinuxaba.db.tar.gz listing, all present at the correct version. gpgmepp’s own x86_64 publish had partially failed this way (file uploaded, db not updated) before the fix – caught and re-published manually via repo_publish.sh directly once fixed.