[pacman-dev] -Sp is broken?

[Dan McGee]

On Nov 19, 2007 1:55 AM, Xavier <shiningxc at gmail.com> wrote:
>
> On Sun, Nov 18, 2007 at 11:47:49PM -0600, Dan McGee wrote:
> > On Oct 3, 2007 1:03 PM, Xavier <shiningxc at gmail.com> wrote:
> > > On Wed, Oct 03, 2007 at 07:52:44PM +0200, Nagy Gabor wrote:
> > > > -Sp simply seems to be broken.
> > > > If I do pacman -Sp evince, I get the following error:
> > > > 'error: evince': transaction not initialized
> > > > error: failed to release transaction (transaction not initialized)
> > > >
> > >
> > > See http://bugs.archlinux.org/task/6812
> > >
> > > git is currently broken because of this commit :
> > > http://projects.archlinux.org/git/?p=pacman.git;a=commit;h=b86479cf9b584c62dbda967902cee2a81678c9ef
> > >
> > > However, after reverting this commit, this command still needs to be run as
> > > root for working. That wasn't the case with pacman 3.0 ..
> >
> > So who wants to tackle this one? Hopefully we can get a real solution
> > to this crap in 3.2, but we may have to hack something out here for
> > 3.1.
> >
>
> It took me a while to understand this problem again, though I already
> explained it in this old thread :
>
> "Oh... I just figured out why this happens. In pacman 3.0, the lock was in
> /tmp/pacman.lck , while it's now in /var/lib/pacman/db.lck.
> So previously, a normal user could lock. And not anymore..   "
>
> So, I don't know, should we explictly skip locking the db in alpm_trans_init
> in case of PRINTURIS ?
> That sounds really ugly.
>
> But if you do that, besides putting the PRINTURIS check back in
> needs_transaction , src/pacman/util.c , then it might work.

The best solution would be to factor print URIs operation out of
needing a transaction at all. This may be helped by me actually
merging your upgrade patch first.

-Dan