[Mario 'BitKoenig' Holbe]
> This claim is obviously not true for 2.4 systems without any udev
> support at all (nor is it for early 2.6 systems where udev support is
> not available anymore - at least not by the Debian package).
Sure. The issue at hand is that the approach used by discover to
track kernel modules is flawed in principle, as it is always lagging
behind the kernels view on things as represented in
/lib/modules/*/modules.pcimap. So I believe it is best to stop trying
to run behind the kernels picking up the pieces, and leave the kernel
module loading to those systems using the information provided by the
kernel modules themselves. This is the reason I dropped the init.d
script.
discover do have some merit when it come to providing information not
available elsewhere, like X module and debian package information. So
I want to focus on doing that part well, and leave the kernel module
issue to someone else.
I'm not going to reject updates to discover-data to add kernel
modules, but I am not going to spend much time keeping the kernel
module information up-to-date either.
If your system is using 2.4 or early 2.6 system, I wish you good luck
and hope you have time time available required to keep it secure. But
I am not going to spend much time making it easier for you, because I
believe the sensible approach is to move to a 2.6 kernel with reliable
/lib/modules/*/modules.pcimap and to use that information instead.
> Furthermore, enforced removal of the init-script could also break
> previously running systems (since I don't know such a system I'll
> keep the current severity of this bug at wishlist, but expect it to
> be raised sooner or later :)).
Well, as far as I can see, upgrading will also enforce an upgrade of
the kernel, and newer kernels will pull in udev and thus continue to
work.
> Btw: Currently, not even the sample init-script is shipped within
> the package, so interested users have to dig into the package
> sources to find it.
I'll consider copying the init.d script into
/usr/share/doc/discover/examples/ or similar, to make it easier
available for those wanting to spend time duplicating the work of the
kernel developers. It is not as it is hard to recreate for those in
need:
#!/bin/sh
test -x /sbin/discover-modprobe || exit 0
case "$1" in
start|restart)
if [ "$VERBOSE" = "no" ] ; then
discover-modprobe
else
discover-modprobe -v
fi
;;
stop|reload|force-reload)
exit 0
;;
*) echo "usage: /etc/init.d/discover [start|stop|force-reload]"
exit 1
esac
> And further btw: The actual init-script (from the sources :)) runs
> discover-modprobe twice: once conditional with or without -v, once
> unconditional with -v.
Ah, thank you for the heads-up. I'll fix that in svn.
Friendly,
--
Petter Reinholdtsen
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
