For 1.3, we should look at the amount of extra space we currently allow for images; in particular it might be nice when an image has package_management enabled to give it some additional space to allow for additional packages to be installed.
This is really up to the downstream distros, there was a brief time before the postinstalls were run on the host, a minimal/systemd build was not large enough, but that is now solved. I we really wanted we could check form package-management in the image creation and increase the IMAGE_OVERHEAD_FACTOR from 1.3 to say 1.5. Assigning it to Laurentiu since he is about to look at the code also. Thoughts?
I've been looking on this and, apparently, for rpm we already add a 50MB extra space if we use package management (http://git.yoctoproject.org/cgit.cgi/poky/commit/?id=7061af381277a20e4df455daac9754b73fcbe467). Also, I've seen a logic error in image size computation. Right now, it is done like that: base_size = COMPUTED_SIZE * ${IMAGE_OVERHEAD_FACTOR} base_size = ((base_size > ${IMAGE_ROOTFS_SIZE} ? base_size : ${IMAGE_ROOTFS_SIZE}) + ${IMAGE_ROOTFS_EXTRA_SPACE}) The problem I spotted is when COMPUTED_SIZE * ${IMAGE_OVERHEAD_FACTOR} is less than but very close to ${IMAGE_ROOTFS_SIZE}, in which case the resulting image size will be always ${IMAGE_ROOTFS_SIZE} and there will be no free space on the image. So, in this case, IMAGE_OVERHEAD_FACTOR is useless. Even delayed postinstalls might fail in this case. Unless IMAGE_ROOTFS_EXTRA_SPACE is set...
wic uses ROOTFS_SIZE variable to set partition size. This variable is calculated in image.bbclass get_rootfs_size function using IMAGE_ROOTFS_ALIGNMENT, IMAGE_OVERHEAD_FACTOR, IMAGE_ROOTFS_EXTRA_SPACE, IMAGE_ROOTFS_MAXSIZE and the actual size of rootfs directory returned by 'du -ks <path to rootfs directory>'. This should be enough to handle size-related variables correctly.