perl-crypt-rijndael

• Path: /data/INSTALL/perl-crypt-rijndael (plain, non-submodule git checkout; listed in scripts/install_only_packages as local-install/perl-crypt-rijndael – not tracked under arch/ at all, see the dated entries below for why). Vendored from AUR 2026-08-29/30 alongside mod_fcgid to unblock thruk – the only 2 of thruk’s 18 dependencies not already in official repos at that time.
• Plain CPAN-style Perl module (CPANPLUS::Dist::Arch-generated PKGBUILD), no PGP source signature, no local modifications at all, no build issues. Built and published cleanly on i686/x86_64 (its declared arch=('i686' 'x86_64' 'armv7h' 'aarch64'); armv7h/aarch64 skipped – the whole ARM cluster was down for the entire 2026-08-29/30 session).

2026-08-30: maintained/ -> adapted/ -> local-install/ (two wrong

categories before landing on the right one)

1. Vendored via scripts/add_maintained.sh, hardcoded to always add under maintained/ – but we’re neither maintainer nor co-maintainer here (AUR shows maintainer jose1711, last packager jose1711, no mention of andreas_baumann on the actual package page). git mv maintained/perl-crypt-rijndael adapted/perl-crypt-rijndael as a first fix.
2. That was also wrong: adapted/ packages are never git submodules, and more fundamentally are specifically for packages with a real local diff on top of upstream (see every entry in arch/adapted/README.md – each names a concrete local patch/fix). This package has zero modifications from jose1711’s current AUR PKGBUILD, so there’s nothing to diff/track – confirmed via diff against a live fetch. Deinit’d and removed the submodule entirely; the already-existing /data/INSTALL/perl-crypt-rijndael staged checkout (from the original build) became its only home instead, added to scripts/install_only_packages as local-install/... – the existing convention for “someone else’s unmodified PKGBUILD, kept out of arch/ on purpose” (previously only used for yay-installed personal packages; broadened to also cover an AUR dependency pulled in purely to unblock another package’s build, which this is).

See memory/mod_fcgid.md’s matching (longer, since that one had an orphaned local commit to untangle) entry, and memory/TOOLING_NOTES.md for the scripts/add_*.sh maintainer-check fix this whole episode motivated.

2026-08-31: ARM catch-up – armv7h/aarch64 built once the cluster came back

Both were skipped at initial build time since the whole eurobuild3/4/5/14 cluster was down that session. Once it came back up, built with REPO_BUILD_ONLY_ARCH="armv7h" then "aarch64" (run directly against /data/INSTALL/perl-crypt-rijndael, same plain-passthrough path repo_build_staged.sh already uses for any package with no arch/ counterpart). Both built and published cleanly, no issues – confirmed via repo_arch_status.sh: all 4 of this package’s declared archs (i686/x86_64/armv7h/aarch64) now show published (1.16-3).

Noted in passing: REPO_BUILD_ONLY_ARCH only restricts which archs repo_build.sh attempts to build – repo_publish.sh afterward still globs and (re-)publishes every *.pkg.tar.* sitting in the pkgdir, including old i686/x86_64 files left over from the original build (harmless here since the version hadn’t changed, just a redundant re-upload of identical bytes – but worth remembering this isn’t a “touch only this arch” guarantee end-to-end).