munin-libvirt-plugins

Build times (auto-updated)

• any: 1m10s (2026-10-15 08:42 UTC, MAKEFLAGS=-j6)

---

• Path: maintained/munin-libvirt-plugins. New package, created 2026-10-15, initial AUR push done by the user the same day – now a real git submodule (ssh://aur@aur.archlinux.org/munin-libvirt-plugins.git), AUR confirms the user as maintainer. Converting from the earlier plain untracked directory to the submodule hit a snag worth remembering: the user had also already run git submodule add themselves locally (setting up their own AUR push), so scripts/add_maintained.sh failed with fatal: 'maintained/munin-libvirt-plugins' already exists in the index after an rm -rf of the plain directory – the script doesn’t check that command’s exit code, so it printed “added” anyway despite failing. Recovered cleanly with git submodule update --init maintained/munin-libvirt-plugins (the .gitmodules entry was already correct from the user’s own setup) – no data lost, but don’t rm -rf a package directory before confirming via git ls-files/git status whether it’s already a tracked gitlink, especially right after telling the user “I’ll drop this in, you handle the AUR side” – they may do more locally than just the AUR-side push.
• Real build bug caught by the actual chroot pipeline, not the earlier plain makepkg -s test: first repo_release.sh attempt failed with ERROR: Cannot find the git package needed to handle git sources – git was missing from makedepends. My earlier validation build (plain makepkg -s directly on this host, before the chroot pipeline existed for this package) didn’t catch it because git was already installed on the host itself, masking the missing makedepends entry. Fixed by adding makedepends=('git'); pkgrel not bumped since this was fixed before any version of the package was ever actually published (not a “fix a published package” bump). Lesson: a plain non-chroot makepkg -s test is not a substitute for a real chroot build when validating a brand-new PKGBUILD – it can silently pass due to host-side packages the chroot doesn’t have.
• 4 Python3 Munin plugins (libvirt-cputime, libvirt-blkstat, libvirt-ifstat, libvirt-mem) to monitor libvirt-managed VMs, plus a munin-libvirt-plugins-detect autoconfig helper. GPL-2.0-only.
• Upstream is agx/munin-libvirt-plugins (Guido Günther, github.com/agx/munin-libvirt-plugins), not the similarly-named MikeQG/munin-libvirt-plugins fork – the latter has zero tags and no real commits since 2012; agx’s is what Debian actually packages (munin-libvirt-plugins 0.0.7-1.2 in sid), confirmed via packages.debian.org. No existing AUR package under this or a similar name – confirmed directly against the real AUR RPC API (https://aur.archlinux.org/rpc/v5/search/...), not a web-search summary, which falsely claimed a munin-libvirt-git package already existed (hallucinated/wrong – always verify AUR package-existence claims against the RPC API directly, not a search summary).
• Pinned to the exact git tag Debian builds from (debian/0.0.7-1), via a git+...#tag=debian/0.0.7-1 source – upstream doesn’t publish GitHub Release tarballs, only tags, so a VCS source made more sense than guessing at a tarball URL. sha256sums=('SKIP') since it’s an unsigned plain git source (no ?signed, no validpgpkeys – upstream doesn’t sign tags for this project).
• Upstream’s own Makefile has a real gap worth knowing about: install: build depends on a build target that’s declared .PHONY but has no rule at all, and does not list the sed-generated munin-libvirt-plugins-detect file as an install prerequisite either – make install alone, cold, would try to install a munin-libvirt-plugins-detect file that doesn’t exist yet. Debian’s debian/rules works around this implicitly (plain dh $@ -> dh_auto_build runs bare make, which builds the Makefile’s first target, all: $(DETECT), before dh_auto_install runs make install separately). Mirrored that split explicitly in the PKGBUILD: build() runs plain make (generates the detect script from its .in template), package() runs make DESTDIR="${pkgdir}" SBINDIR=/usr/bin install (overriding upstream’s default SBINDIR=/usr/sbin – Arch convention, /usr/sbin is just a symlink to /usr/bin here anyway but no reason to rely on that).
• depends=('munin-node' 'libvirt-python' 'libxml2' 'python') – matches Debian’s Depends: munin-node, python3, python3-libvirt, python3-libxml2 exactly, translated to real Arch package names. Confirmed on this host: core/libxml2 already ships the python3 libxml2 module directly (no separate python-libxml2 package exists or is needed on Arch – checked pacman -Ql libxml2 | grep python), and both import libvirt / import libxml2 succeed with libvirt-python + libxml2 installed.
• Build-tested clean via plain makepkg -s (not yet run through this project’s chroot/repo_release.sh pipeline, since it’s not going to be published to the private archlinuxaba repo – this is a genuine AUR-bound package, built here only to prove the PKGBUILD/Makefile interaction works). Resulting file layout verified by hand (tar tf): plugins under /usr/share/munin/plugins, detect helper under /usr/bin, license + README under /usr/share/licenses / /usr/share/doc. namcap wasn’t available outside a chroot on this host to double-check further – worth a quick namcap pass once this does go through a real chroot build (e.g. after the AUR push, via the normal maintained/ pipeline).