Currently initscripts will not work correctly if /usr is mounted by initramfs-tools. Firstly, fsck will fail and the boot process will stop. If you press Ctrl-D to continue, the boot process will succeed but /usr will not be remounted read-write if necessary.
upstart (or mountall) and wheezy's version of systemd also fail to remount /usr as read-write. This issue *must* be fixed *now*. I am extremely annoyed that this delicate change has been handled so badly and I intend to upload some kind of fix by the end of this week. My current ugly fix is to mount /usr if init is systemd, and add a versioned Breaks to ensure systemd is upgraded if necessary. The actual fix might not look quite like this: * If there is a fixed initscripts and/or util-linux available then the check for systemd can be replaced with a versioned Breaks on initscripts. * As for upstart, since it installs /sbin/init I don't see a way to distinguish it from sysvinit. So if initscripts is fixed and upstart is not then I'm afraid I may have to add an unversioned Breaks: upstart. If they are both fixed, then that can be another versioned Breaks:. * If anyone can present a better solution (not an idea, not a plan, but an actual implementation), I would be very happy because this one is not at all nice. (Aside from 'init: Fix validation of the real init program' which is a bug fix in its own right.) Ben. Ben Hutchings (5): init: Decide what the real init is immediately before mounting /usr init: Fix validation of the real init program init: Resolve both absolute and relative symlinks in validate_init init: Only mount /usr if the real init is systemd debian/control: Add Breaks: systemd-sysv (<< 186) debian/control | 2 +- init | 90 ++++++++++++++++++++++++++++------------------------------ 2 files changed, 44 insertions(+), 48 deletions(-) -- Ben Hutchings Logic doesn't apply to the real world. - Marvin Minsky
signature.asc
Description: This is a digitally signed message part