I've reproduced the lacking conflict with the latest server and virtual kernels in karmic-updates, 2.6.31-20.58, and can confirm that the kernels in karmic-proposed, linux-image-2.6.31-21-{server,virtual} 2.6.31-21.59, correctly conflict on amd64. I also verified that the linux-image-2.6.31-21-{generic-pae,virtual} kernel packages continue to conflict with each other.
I also verified that none of the 2.6.31-20.58 kernels (generic, server, generic-pae for {i386,amd64}) introduce no regressions in the qa-regression-testing kernel tests. Marking verification-done.
The reason these packages need to conflict is that the linux virtual kernel is not separately built from the server/generic-pae kernel images. The same kernel image is used as both, it's just that the virtual kernel packages include fewer kernel modules. Re- architecting the kernel build process is out of scope for a karmic SRU as well as for lucid, given how late it is in its development cycle.
I've reproduced the lacking conflict with the latest server and virtual kernels in karmic-updates, 2.6.31-20.58, and can confirm that the kernels in karmic-proposed, linux-image- 2.6.31- 21-{server, virtual} 2.6.31-21.59, correctly conflict on amd64. I also verified that the linux-image- 2.6.31- 21-{generic- pae,virtual} kernel packages continue to conflict with each other.
I also verified that none of the 2.6.31-20.58 kernels (generic, server, generic-pae for {i386,amd64}) introduce no regressions in the qa-regression- testing kernel tests. Marking verification-done.
The reason these packages need to conflict is that the linux virtual kernel is not separately built from the server/generic-pae kernel images. The same kernel image is used as both, it's just that the virtual kernel packages include fewer kernel modules. Re- architecting the kernel build process is out of scope for a karmic SRU as well as for lucid, given how late it is in its development cycle.