Some news from auxiliary debian packages concerning debian-live
- Subject: Some news from auxiliary debian packages concerning debian-live
- From: daniel at debian.org (Daniel Baumann)
- Date: Sat, 09 Feb 2008 12:28:29 +0100
- Message-id: <47AD8E5D.email@example.com>
in the last days since 2.6.24 upload to sid, I was giving some love to
the kernel modules. Some comments wrt/ debian-live:
* aufs is usable and working for normal setups (newer aufs seems to
have issues with nfs and plain images, though. not fully
investigated about this yet). prebuild binary modules are in the
any tests and bug reports welcome, please fill bugs directly against
the aufs package if it is not related to a bug in live-helper or
live-initramfs. i'm subscribed to aufs, so i'll help distinguish
in case of live specific bugs/help/$whatever.
* squashfs as of 3.3-3 has lzma default-off, which means, it produces
non-corrupted normal images again. there's no need for the -no-lzma
workaround anymore. prebuilt binary modules are in the archive.
as of squashfs version 3.3, there seems to be new, other issues
unrelated to lzma. those are not investigated yet. as with aufs,
any tests/bugs welcome, but send them against squashfs package
squashfs plus lzma enabled seems to be unusable, both because it's
too bugy and because it's not integrated well in debian (e.g. no
prebuilt lzma-modules). i've tried to integrate it, but wasn't
welcome to do so. so, i personally wont care about this anymore.
* unionfs 2.2.3 is now in sid, but it needs a kernel patch, which
i've applied. so, for the time beeing until 2.6.24-4 makes it into
sid, unionfs will not work (module is refused to be loaded).
the recommended combination of modules for lenny/sid is still
aufs and squashfs.
* user-setup 1.18 is now in testing. Thanks again to Jordi for
discovering the regression. Bugs like 'there's no user
created in my live system' are fixed with this, the workaround
of putting the 1.18 .deb into config/chroot_local-packages is
not needed anymore.
* Probably for linux 2.6.26, we'll see a read-only bindmount
implementation. The lack of it has always been a mayor problem in
debian-live, as it requires to copy around the chroot in order
to built clean, not-hostsystem-tainted images.
If we have ro-bindmounts, image buildtime decreases dramatically
from about 20min to 5min for a kde image on my system.
As ro-bindmounts are a new thing and will require a new enough
kernel version on the host system, this will be third, additional
method to manage the build (currently we have just two,
chrooted-with-full-copy and unchrooted).
Address: Daniel Baumann, Burgunderstrasse 3, CH-4562 Biberist
Email: daniel.baumann at panthera-systems.net