On Fri, Mar 01, 2013 at 11:56:32PM +0000, Burton, Ross wrote: > On 1 March 2013 23:32, Martin Jansa <martin.ja...@gmail.com> wrote: > > Can you answer this > > http://lists.linuxtogo.org/pipermail/openembedded-core/2013-February/036223.html > > and how this solution helps with upgrade paths? > > Having split packages can break the upgrade path - say your distro > goes from sysvinit to sysvinit rescue + systemd main. How does your > foo-daemon package get the right init script package on upgrade?
By right RRECOMMENDS like meta-systemd did. Plus simple way to exclude some at image creation time with BAD_RECOMMENDATION or explicit entries pulled with packagegroup for each type of image. > I proposed a solution for distributions that care - inject the > migration path dependencies though meta-systemd. I still maintain > that oe-core shouldn't have to bend over backwards to maintain > compatibility with every recipe that migrates. Obviously we don't > want to deliberately break where we have a choice but equally so http://lists.linuxtogo.org/pipermail/openembedded-core/2013-February/036222.html > > Coding was contributed to meta-systemd which was working fine for both > > use-cases. Maybe explaining hidden benefits of not splitted packages > > would motivate some people.. > > The advantage of having init scripts in the daemon package is > simplicity. For the cost of two init scripts (what, 1K between them?) > you remove lots of complexity, including the upgrade path breakage I > described above. There is no upgrade path breakage and complexity is in postinst scripts which need to support both update-rc.d and systemctl calls. Lets discuss this in that oe-core thread so that other people not subscribed to yocto ML can also comment. -- Martin 'JaMa' Jansa jabber: martin.ja...@gmail.com
signature.asc
Description: Digital signature
_______________________________________________ yocto mailing list yocto@yoctoproject.org https://lists.yoctoproject.org/listinfo/yocto