Cobbler
Linux installation and provisioning server for bare metal and virtual systems
Cobbler's 3.3 backport branch ships Leap 16 and Agama support, then has to re-cut the release
◆Recent moves
- 17d ago
3.3.8 re-cut after fixing the PyPI publish workflow
The only change is a fix to the pypa/gh-action-pypi-publish SHA-pin that left 3.3.8's packaging workflow unrun on tag pushes. Nothing new ships here, but this is the tag that actually delivers 3.3.8's Leap 16, Agama and XCP-ng work to users.
View source ↗ - 17d ago
Leap 16 and Agama support, withdrawn as defunct
The substantive half of this pair: openSUSE Leap 16 and Agama installer support, interactive and automatic XCP-ng 7.x and 8.x network installs, and basic Python 3.13 support, alongside BIND zone and Windows import fixes. It was marked defunct the same evening when its packaging workflow failed, so the content reached users through 3.3.9 instead.
View source ↗ - 2mo ago
CVE-2024-47533 backported to the 3.2 branch
The 3.2-branch twin of 3.3.7, cut for the same CVE-2024-47533 admin-API bypass, with its own fix and feature lists left as TBD. Its feed date sits well after the 2024-11-17 release it describes, so the stamp reads as a re-capture rather than new work.
View source ↗ - 1y ago
CVE-2024-47533: API admin authentication bypass fixed
The 3.3-branch fix for an issue that let anyone connect to the API as admin, bundled with BIND cname generation, RHEL9 build and Windows fixes. Backports like this are the clearest justification for keeping 3.3 alive for downstreams.
View source ↗ - 2y ago
EFI support added to cobbler buildiso
EFI support in cobbler buildiso is the notable addition, with extra_settings_list giving operators a supported way to carry custom settings. A typical rollup for this branch: a couple of real capabilities wrapped in a long fix list.
View source ↗ - 2y ago
Lazy collection loading and UUID, MAC and IP indices
Startup lazy-loading plus new indices for UUIDs, MACs, IP addresses and hostnames - work aimed at installations large enough for linear scans to hurt. It marks the point where the backport branch starts absorbing performance work, not just fixes.
View source ↗