I think these notes cover most of the diff, in varying levels of
detail.
Ben.
## compcache/zram
16:13 < apw> yeah a lot of that is compcache adding support for zram iirc
16:14 < infinity> The compache stuff can disappear, I need to tear it out.
16:14 < infinity> We should just support zram, so I'll fix all of that.
16:14 < apw> we also restore a few older compatibility options which i am sure
we longer use, but without interlocking with infinity on d-i i had
not yet ripped
## wait-for-root
16:21 < bwh> At the moment Debian's initramfs-tools doesn't assume all device
setup runs in udev hooks
16:22 < bwh> so it calls initramfs-tools hooks periodically
16:22 < apw> no indeed, and in the last merge up i believe i did restore the
use of the loop when wait-for-root does not work
16:22 < apw> which isn't great either
16:23 < apw> i guess to make it more compatible i should have used wait-for
root as the sleep, but that is rather ugly
16:26 < bwh> So I agree that wait-for-root *is* doing something useful still,
I'm just not sure how that ought to be combined with the polling
we do
16:26 < apw> well right now we poll and sleep and wake and run hooks and poll
16:27 < apw> it seems it could most usefully allow short circuiting of those
sleeps as it has a timeout
16:27 < apw> so if root does appear we find out straight way
16:27 < bwh> We still need something like that for (a) mdadm timeout of
incremental assembly (b) lvm2 activation (given the current
opposition to activating all VGs auatomatically)
16:28 < apw> i assume we handle both of those out of udev else we would not
have been able to not have the repeated hook calls
16:28 < apw> but likely by activating the lot unconditionally
...
16:32 < bwh> I think I'd be happy to include wait-for-root and then Debian and
Ubuntu may differ in what the itmeout for that is, i.e. how often
they run the local-block hook
16:33 < bwh> ...and then later we may be able to get rid of the need for that
hook
## keymap loading
No keymap hook in Ubuntu; instead it's part of keyboard-configuration
(built from console-setup, which is very different in Ubuntu). keymap
is loaded on panic or if framebuffer is used (?). Password
entry always uses framebuffer (?).
## fixrtc
Disgusting hack... specific to ext[234], needs to compare dates in shell
script, duplicates device lookup (and wait!) code.
## loopback device support
Was meant for WUBI which is dead, but there is some other interest in
using it.
Could it be separated from scripts/local?
## framebuffer setup
Lots of code, and infinity's not sure what it's for. Harmless if not
enabled by default?
## OPTION= hook/script selection
Sensible to allow filtering this way, but dislike the syntax.
Within initramfs-tools only the framebuffer hook/script use this;
if no other package uses it then it could easily be changed.
## Upper-case UUID fixup
libblkid is matching UUIDs case-sensitively which I think breaks
`root=PARTUUID=` in some cases.
See <https://bugs.launchpad.net/ubuntu/+source/initramfs-tools/+bug/1548120>.
Currently worked around by retrying with conversion to upper-case,
but should really be fixed in libblkid.
--
Ben Hutchings
For every action, there is an equal and opposite criticism. - Harrison
Attachment:
signature.asc
Description: This is a digitally signed message part