Hello Michael, sorry for the delay! Some national/personal holidays came in between.
Michael Tokarev [2026-05-23 19:12 +0300]: > Package: cockpit-machines > Version: 351-1 > Severity: serious > > qemu dropped support for 32bit hosts in 11.0 version, this support > is not going back. Due to this, libvirt daemon dropped qemu-related > packages on i386 and armhf (and other 32bit architectures). I suppose this is related to https://salsa.debian.org/libvirt-team/libvirt/-/commit/f7c3ccddabded4b5919f0f39fcb1b43d4290366b#line_58ef006ab_A810 But I don't understand this. I get that libvirt-daemon-driver-qemu was dropped, but libvirt can manage more than just QEMU, e.g. also containers with LXC. That should still work on 32 bit machines. So why dropping libvirt-daemon-system entirely, as opposed to just reducing its dependency to libvirt-daemon-driver-qemu (as the patch also did)? This is unrelated to cockpit-machines, but relevant for libvirt users. > Due to this, with new libvirt your package fails autopkgtests, and > hence whole set of packages can't migrate to testing. I see: https://ci.debian.net/packages/c/cockpit-machines/testing/i386/ https://ci.debian.net/packages/c/cockpit-machines/testing/i386/71625328/ At least it doesn't seem to block libvirt migration: https://qa.debian.org/excuses.php?package=libvirt In theory it should, not sure why .. it's blocked on gnome-boxes rather. > Please consider dropping (test) dependency on libvirt-system-daemon > on 32bit architectures, or stop building cockpit-machines on these > architectures entirely. Hmm, c-machines is Architecture: all. Conceptually, building on just the supported architectures is right, but it'd have to become arch: any then and that's difficult because reasons (esbuild yadayada). I'll make the dependency architecture specific as a quick fix to unblock migration. Thanks, Martin
