Kel Modderman wrote:
It breaks the dependency on a locally installed package, that has no
dependencies or control data? Please further convince me that this is
actually a problem in the packaging of this module, and that you should not
use equivs to satisfy these depends when installing local software.
I personally like this dependency on the linux-image. The modules are
worthless without the linux-image. It also prevents the modules from sticking
around, wasting space, when the linux-image is removed.
My point here is that if modules are built for a custom installed kernel
(so no make-kpkg one), it should either be general for all module source
packages to depend on [linux|kernel]-image-[kernelversion] or recommend
it. What I'm seeing here is that for fgrlrx it's recommended, and for
madwifi it's dependant on it.
I briefly discussed this in #debian-kernel:
<infernix> is a module-assistant compiled package supposed to assume
that the current kernel is a make-kpkg generated one and depend on it?
<Manoj> no
<Manoj> users are not required to install image .debs
<Manoj> I mean, the package works just fine without the dependency for
you, so the dependency is spurious
As a consequence, I filed this as a bug for madwifi-source, because I
couldnt find any policies or guidelines. If you are positive that any
module built with module-assistant is supposed to depend on
[linux|kernel]-image-[kernelversion], I will file a bug with fglrx.
However, I have checked out ipw2100 and ndiswrapper and they did not
depended on [linux|kernel]-image-[kernelversion] either. Haven't tried
more but it might be worthwhile to check them all out.
In either case, if there's no current stance on this, I think one should
be made and it should be defined in modules-assistant's HOWTO-DEVEL so
that all module source packages are doing the right thing.
Regards,
infernix
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
