Re: s,/sbin/update-grub,update-grub,g transition
On Sun, 6 Aug 2006 09:11:52 +0200, Robert Millan <rmh@aybabtu.com> said:
>> As for #361929, I overlooked that kernel-package also happens to be
>> the most appropiate place for fixing the user scripts. Would you
>> please also add something like:
I reject this assertion. There is no way of knowing when the
end user will not be subject to kernel images built with
kernel-package << 10.051 -- which, incidentally, means every sing
official image at the moment. So a relative path is a really bad
idea.
And there is no reason for update-grub to go right now to a
relative path -- update-grub is unlikely to ever move again.
On the other hand, we do know, when the path for
update-sbin moves -- that is, when the new version is
installed. kernel-package will have to test the location to see which
path is valid for the current machine.
So, since a relative path for the postinst hook script would
break all current images, and just changing the absolute path would
work for the foreseeable future, the best place for this change is in
grub.
Post Etch, at some point, one could try relative paths. Unless
you plan on moving update-grub _again_, there is no point doing it,
though.
>> sed -i /etc/kernel-img.conf -e
>> "s,\(.*\)/sbin/update-grub$,\1update-grub,g"
>>
>> for linux-image-* postinsts ?
>>
>> (or whatever the perl equivalent is)
> Do you think it could be put in kernel-package as well ? We could
> handle the transition in grub, but then we'd have to work out when
> do our kernel postinsts support relative paths, and it's not trivial
> because of the way kernel-package is used to generate postinsts in
> other packages.
This is the wrong fix.
Indeed, expect a grave bug from me (for breaking unrelated
packages) if anything puts this scheme in place, since it would break
an image which otherwise works just fine. If you must edit
configuration files in place, go from one absolute path to another
absolute path, depending on the path of the binary shipped in grub.
manoj
--
You are sick, twisted and perverted. I like that in a person.
Manoj Srivastava <srivasta@debian.org> <http://www.debian.org/%7Esrivasta/>
1024D/BF24424C print 4966 F272 D093 B493 410B 924B 21BA DABB BF24 424C
Reply to: