summaryrefslogtreecommitdiff
path: root/scripts/live-functions
diff options
context:
space:
mode:
authorChris Lamb <chris@chris-lamb.co.uk>2008-04-15 18:59:26 +0100
committerDaniel Baumann <daniel@debian.org>2011-03-09 17:47:59 +0100
commit55d0448c2b0bef163092e251879047d92a58c91e (patch)
tree9dc73e0da6992f2f4188cf72a2c6d29623502f83 /scripts/live-functions
parent570998533ee43582fd73beee3e206c52ee3494dc (diff)
downloadlive-boot-55d0448c2b0bef163092e251879047d92a58c91e.tar.gz
live-boot-55d0448c2b0bef163092e251879047d92a58c91e.zip
Use triggers when calling update-initramfs in postinst and postrm
One complication with the interaction between live-initramfs and update-initramfs is completeness - in a normal Debian system, we do not mind that not all the initrds are updated, so one would simply call "update-initramfs -u" when upgrading (for example) mdadm. However, live-initramfs needs to update *all* initrds (ie. '-u -k all'), but calling this would not use the trigger functionality as 'update-initramfs' will only accept triggering with a single '-u' argument. This is sensible behaviour, as the trigger callback cannot (and shouldn't) be parameterised. The result is this is that we must detect how many kernel flavours we have and revert to the non-triggered call ('-u -k all') if we have more than one. My tests show that this is behaviour is sound, even in the slightly contrived example of updating live-initramfs whilst installing a second kernel, regardless of the interleaving of the calls to postinst. We additionally attempt the same logic in the postrm so that upgrading live-initramfs (for example, in chroot_local-packages) does not result in two calls expensive calls in update-initramfs.
Diffstat (limited to 'scripts/live-functions')
0 files changed, 0 insertions, 0 deletions