Re: [PROPOSAL] update-binfmts - manages the binfmt_misc kernel module
you make a valid point, but i don't think that outweighs the benefit of
allowing packages to depend on a certian kernel feature.
nobody expects, say, fbset to work on a kernel compiled without framebuffer
support. with a few exceptions, most people only have on kernel compiled on
their system. i keep a backup bzImage in lilo.conf, but i really only use
one image.
perhaps the componet of the packaging system that checks for capbilites
could check lilo.conf (or milo, or grub) for the default kernel and check
that. i'm not clear on recent developments. i expect that we will end up
looking in /proc, or devfs when that comes about.
this is certianly a tricky subject, but one worthy of discussion.
Wichert Akkerman (wichert@cistron.nl) wrote:
> Previously Jacob Kuntz wrote:
> > perhaps the answer is to not use the Depends field as it exists today.
> > create a new field called Kernel-Depends. i'm sure there's some way to probe
> > a kernel/module tree for which options a given kernel has compiled in
> > without having the configuration file that was used to build it.
>
> But often I switch between kernels and install something for a kernel
> I'm not currently running.. there is just no way you can make a
> packaging-system support that properly without going through all kind
> of nasty sillyness.
>
> Wichert.
>
> --
> ________________________________________________________________
> / Generally uninteresting signature - ignore at your convenience \
> | wichert@liacs.nl http://www.liacs.nl/~wichert/ |
> | 1024D/2FA3BC2D 576E 100B 518D 2F16 36B0 2805 3CB8 9250 2FA3 BC2D |
--
(jacob kuntz) jpk@cape.com jake@{megabite,underworld}.net
(megabite systems) "think free speech, not free beer."
Reply to: