perl-test-tcp

Build times (auto-updated)

• any: 38s (2026-11-01 09:50 UTC, MAKEFLAGS=-j6)

---

• Path: arch/maintained/perl-test-tcp (new – vendored from AUR 2026-08-02, transitive checkdepends of perl-plack -> thruk). Its dependency perl-test-sharedfork was already vendored (see its own memory file).

• 2026-08-02: built+published cleanly (2.22-2), any package.

• No longer an actual arch/maintained/ submodule, but still bucketed under the maintained/ category in scripts/install_only_packages (see [[perl-filesys-notify-simple]] for the same note).

• 2026-10-29: perl-5.44 ABI rebuild batch (see [[mtop]]). First attempt failed: the chroot’s dependency install hit perl-test-sharedfork-0.35-2-any.pkg.tar.zst is corrupted (invalid or corrupted package (checksum)) – that exact package had been rebuilt+republished to archlinuxaba moments earlier in the same batch run, and this build’s own dependency download raced the mirror/db propagation. Plain retry (no changes) succeeded immediately. Same symptom as [[perl-test-time]] below.

• 2026-11-01: another perl-ABI batch – but this time upstream had genuinely moved on its own: fresh clone showed pkgrel=3 (new maintainer ryoskzypu, real packaging change – SPDX-style license string), while we’d last published 2.22-2. Not our own content-free rebuild – built upstream’s PKGBUILD as-is at 2.22-3, no extra X.Y bump (nothing local to protect against; this is a real upstream pkgrel). Built [perl-test-sharedfork] earlier in this same batch run this time specifically to avoid repeating the 2026-10-29 mirror-propagation race – worked, clean build.

  repo_arch_status.sh falsely reported “not published” right after a confirmed-successful publish (checked the live remote db.tar.gz directly: perl-test-tcp-2.22-3 really is there). Root cause: this PKGBUILD’s pkgname=perl-${_dist@L} is a bash parameter expansion, which a naive sed capture can’t evaluate – it literally captured perl-${_dist@L} as “the pkgname” and looked up that non-existent string in the db. Fixed in repo_arch_status.sh itself (falls back to the directory’s own basename when the parsed pkgname still contains an unexpanded $) – see memory/TOOLING_NOTES.md. Not a problem specific to this package once the script was fixed; just the one that surfaced it.