Hello, On Fri, 2024-11-01 at 22:22 +0100, Paul Gevers wrote: > > No, this can't resolve itself automatically as this is a hard-encoded build > > dependency. What you're seeing here is just the cruft mechanism of DAK in > > action which doesn't exist in Mini DAK which Debian Ports uses. > > Can you elaborate more then? A problem on ports is not "serious", but > I'm not sure if I follow you if there now is or is not a problem for the > release architectures.
The Debian Archive Kit, short DAK, has a mechanism called cruft which makes sure that packages never become BD-Uninstallable when build dependencies are removed or changed by keeping obsolete binary packages around as long as they are needed for building. This is to ensure that transitions never make packages BD-Uninstallable so that manual intervention is required. This mechanism does not exist for Debian Ports because it doesn't use DAK but Mini-DAK. However, cruft is not intended as a permanent solution as the binary packages kept around are obsolete versions of a source package which causes problems with Multi-Arch parallel installability and also poses a security risk. In this particular case, src:hdf5 [1] no longer provides the binary package libhdf5-openmpi-dev for any 32-bit architecture since version 1.10.10+repack-4 which means the package is out-of-date for these architectures. If it wasn't for this bug in src:vtk9, DAK would have purged the 1.10.10+repack-3.3 packages already but they are kept around by the cruft mechanism until src:vtk9 has been fixed. Adrian > [1] http://ftp.debian.org/debian/pool/main/h/hdf5/ -- .''`. John Paul Adrian Glaubitz : :' : Debian Developer `. `' Physicist `- GPG: 62FF 8A75 84E0 2956 9546 0006 7426 3B37 F5B5 F913 -- debian-science-maintainers mailing list debian-science-maintainers@alioth-lists.debian.net https://alioth-lists.debian.net/cgi-bin/mailman/listinfo/debian-science-maintainers
