If a recipe uses update-alternatives in postinst/prerm, this will cause failures with rpm packaging during upgrade as we expect (based on assumptions from opkg) that prerm/postrm occur after postinst. For an example see the original report on the mailing list:

https://lists.yoctoproject.org/pipermail/yocto/2011-November/005559.html

Mark Hatle has suggested using guard wrappers for the prerm/postrm scripts within package_rpm.bbclass that check the calling argument - rpm specifies how many versions of the package are still installed; if the answer is not 0 then according to the behaviour of opkg in this situation we should be able to skip all prerm/postrm scripts, although careful examination of all of these scripts as well as ones in meta-oe should be done to ensure this won't cause any issues.