Hello, I've been doing some testing on this and first tried a no-change rebuild of 3.5.0-1build1 on questing proposed. This can be seen in the PPA here: https://launchpad.net/~bryanfraschetti/+archive/ubuntu/ovs- proposed/+packages
Looking at the build log on amd64, s390x, and arm64, 3.5.0 is now failing in a similar way to 3.5.1 (on OVS datapath actions parsing and formatting - userdata overflow FAILED (odp.at:468) with some additional failures). Here's the build log on amd64: https://launchpadlibrarian.net/806517147/buildlog_ubuntu-questing- amd64.openvswitch_3.5.0-1build2_BUILDING.txt.gz. So a previously successfully built package is now failing on -proposed. It seemed like the problem is not with the package but maybe with one of it's dependencies that's in -proposed causing the breakage. I decided to check locally to try and isolate which package in -proposed was causing the failure. To test locally I - created a VM from the daily amd64 questing iso: http://www.cdimage.ubuntu.com/daily-live/current - enabled proposed - pinned questing-proposed to high priority Package: * Pin: release a=questing-proposed Pin-Priority: 1001 - apt source openvswitch, which pulled in 3.5.1 - debuild -us -uc Which succeeded with minor bashcomp failures I created a script to compare the package versions from the CI with the installed versions on my system and found that there were no meaningful differences. See the differences below. "NOT INSTALLED" means the package was present in the CI but not locally. Similarly, "CI" indicates what was in the CI while Installed indicates what's on my VM. The differences are because I'm running this locally a few days later than when the CI runner executed. NOT INSTALLED: libapt-pkg6.0t64 (proposed: 2.9.27ubuntu1) MISMATCH: libreadline8t64 CI : 8.2-6 Installed: 8.3-1 MISMATCH: libxml2-16 CI : 2.14.4+dfsg-0exp1 Installed: 2.14.5+dfsg-0exp1 MISMATCH: readline-common CI : 8.2-6 Installed: 8.3-1 MISMATCH: rust-coreutils CI : 0.1.0-0ubuntu1 Installed: 0.1.0+git20250711.2ba3a33-0ubuntu1 NOT INSTALLED: sbuild-build-depends-main-dummy (proposed: 0.invalid.0) I noticed that my VM was using the 6.15 kernel, while the CI was using 6.8.0-63 #66, so I downgraded the kernel and built again. This succeeded as well and there was no difference compared to the 6.15 kernel (see attached build log). Since it doesn't seem to be the package, the dependencies, the system package versions in -updates or -proposed, or the kernel, I'm not sure what could be causing the CI failure. Maybe it's a problem with the base image or some LP-specific customizations to the build environment if there are any? The following are just things I'm curious about - Where LP pulls the image (if I can pull this image maybe I can reproduce the failure locally) - In what ways it differs from the questing isos - How often is it updated (maybe a newer base image will solve the failure) To summarize: Rebuilding 3.5.0 (the ovs which was released on questing) in the LP CI against -proposed fails in a similar way to 3.5.1 Building 3.5.1 succeeds on the current questing server iso even with -proposed enabled. ** Attachment added: "openvswitch_3.5.1-0ubuntu1_amd64.build" https://bugs.launchpad.net/ubuntu/+source/openvswitch/+bug/2109823/+attachment/5892285/+files/openvswitch_3.5.1-0ubuntu1_amd64.build -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/2109823 Title: [SRU] openvswitch 3.3.4 (noble) 3.5.1 (plucky/questing) point release To manage notifications about this bug go to: https://bugs.launchpad.net/cloud-archive/caracal/+bug/2109823/+subscriptions -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
