Re: Vanishing /usr/doc symlink
Manoj Srivastava wrote:
> >>"Raul" == Raul Miller <email@example.com> writes:
> Raul> If it's possible to install woody against potato without errors caused
> Raul> by a missing /usr/doc then the tech committee's migration plan is no
> Raul> longer needed.
> Unless one does an audit of the packages (and not jsut
> packages that happen to be installed on ones machine at the moment),
> there is no way to be sure of this. As I have said before, current
> policy makes it perfectly legal to do thisng like
> if [ -d /usr/doc/$package ]; then
> rm -rf /usr/doc/$package;
> ln -sf ../share/doc/$package /usr/doc/$package
> I, for one, have not done this audit, nor do I have the time
> to do so. Additionally, several months into the freeze, isn't it kind
> of late to be changing plans midstream? Policy is frozen. And there
> is a must requirement for packages to provide the symbolic link --
> unless you are saying we should just blow the release manager off and
> the say the hell with the freeze and start over just so we don't ship
> /usr/doc/ full of symbolic links.
What about putting "packages must not depend on /usr/doc being present"
in policy and aiming for a /usr/doc-less woody+1 fresh install, then?
Regarding the code in prerm and postinst, most people uses debhelper,
or cut and pasted from the policy manual, or wrote completely robust
postinst from their own, so we can't make a package audit but we can
reasonably assume that this policy will not force a large number of
packages to suddenly be in violation (which is one of the requirements
of policy changes).