[Pkg-sysvinit-devel] Bug#390069: what about Etch?

[Petter Reinholdtsen]

I must admit the etch base freeze caught me by surprise.  I believed I
had a few more weeks left to fix sysvinit and ifupdown.  Especially
ifupdown should have had a few fixes.  Some of the changes I have done
to initscripts has been to prepare an environment for ifupdown to work
better with network coldplug events.

[Henrique de Moraes Holschuh]
> And this is not and will not be in Etch.

Unless we talk to the release managers about it, yes.
> 
> > Right.  I hadn't considered that.  This problem was fixed in version
> > 2.86.ds1-23, when I changed checkroot.sh to use /lib/init/rw/ instead
> > of mounting a tmpfs over /tmp/.
> 
> We will have to revert these changes in Etch, along with the ones in
> /dev/shm that are breaking UML.

If we drop nodev for /dev/shm/, checkroot.sh can use it to store the
root device node as before.  I believe it is better to introduce
/lib/init/rw/ into etch and let checkroot.sh use is to store the
device node.  To unbreak UML, I believe it is enough to drop noexec
for /dev/shm/.

All of these things are fixed in -28.

I'm not yet sure what is breaking vserver, so I do not have a
suggestion there.

Personally, I suspect it the best option is to get -28 into Etch.