[arch-dev-public] [signoff] mkinitcpio 0.7-1, device-mapper/lvm2 2.02.85-3, dmraid 1.0.0.rc16.3-2, mdadm 3.2.2-2, v86d 0.1.10-2

[Dave Reisner]

On Thu, Jun 30, 2011 at 09:40:07PM +0300, Evangelos Foutras wrote:
> On 30 June 2011 21:15, Thomas Bächler <thomas at archlinux.org> wrote:
> > This is a huge release and should stay in testing for a while. I want to
> > point out two significant changes:
> >
> > 1) We now have a build() function instead of an install() function - the
> > latter was an unfortunate conflict with the install utility. The hooks
> > in the repository have been adjusted, but backward-compatibility code is
> > in place (a deprecation warning will be printed). All these packages
> > have conflicts=(mkinitcpio<0.7).
> >
> > 2) We now use bsdcpio instead of gen_init_cpio. This obsoletes the
> > gen_init_cpio package and adds a dependency on libarchive. This also
> > speeds up the image generation considerably.
> >
> > Other changes include massive cleanups and fixes, code refactoring, and
> > changes that should finally make the -b (basedir) option useful.
> >
> > Please test and sign off.
> 
> Signoff both.
> 
> (Tested mkinitcpio on x86_64 and {mkinitcpio,device-mapper,cryptsetup} on i686.)

Signoff i686.

Also wanted to highlight the change to the -k flag, which now lets you
specify a kernel version as a path to a kernel image itself. This makes
the .kver file distributed with kernel26 package superfluous.

Found one minor display bug that occurs when you add a
binary through BINARIES in /etc/mkinitcpio.conf. My error handler is
tripped because the 'while read' loop correctly returns 1 at EOF, and is
read as a failure. This results in an ominous warning being printed
without any other evidence that anything has gone wrong.

Fix is already on my working branch:

https://github.com/falconindy/mkinitcpio/commit/a594ffca592612f9293454198e56d7ea72b75d14

d