[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Bug#985749: apt: "apt-mark hold" flag lost on package upgrade using --ignore-hold



Control: reassign -1 apt

Hi!

On Tue, 2021-03-23 at 00:06:21 +0100, Sven-Haegar Koch wrote:
> On Mon, 22 Mar 2021, David Kalnischkies wrote:
> > On Mon, Mar 22, 2021 at 10:21:59PM +0100, Sven-Haegar Koch wrote:
> > > I hold the package, and with normal upgrade/dist-upgrade it works
> > > exactly as expected.
> > > 
> > > But when I then upgrade these single package later using --ignore-hold,
> > > the hold flag is lost afterwards.
> > 
> > holds are stored by dpkg as a "selection state", which e.g. install or
> > deinstall are, too, and which will override the old selection state sort
> > of by design.
> > 
> > It is also this way since the dawn of time, so that is kinda unlikely to
> > change – resolving this bug might be as "simple" as adding a note that
> > holds will be (potentially) lost if they are ignored.
> > 
> > Sorry, as that is probably not what you wanted to hear.
> 
> Thanks.
> 
> I think a note in the apt-get manpage on the options (--ignore-hold and 
> --allow-change-held-packages) would already have helped me a lot - it 
> took very long to figure out that my problem was indeed loosing the 
> mark after upgrading the package - as usually the availability of the 
> next version of a marked package happens weeks or months later.
> 
> And when you discover the mark missing on a dist-upgrade call then, you 
> mostly think "I thought I set it, did I really also on this server?" :)
> 
> Never even got the idea that the hold would not be supposed to 
> survive, neither from apt-get nor from apt-mark manpages.

In dpkg this got clarified with bug #926472, I've just pushed a commit
to dpkg git HEAD to further clarify a confusing wording in the man page
for the --force-hold description, but otherwise I think the stuff in
the «Packages selection states» subsection should be clear enough
already.

Given that you request improvements to the apt documentation, I'm
reassigning back.

Thanks,
Guillem


Reply to: