[Aptitude-devel] Bug#346321: aptitude: offers upgrade to exp version (pri -10) instead of unst version (990)

[Manuel A. Fernandez Montecelo]

Control: tags -1 - moreinfo
Control: close -1

2016-03-18 0:59 GMT+00:00 Henrique de Moraes Holschuh <hmh at debian.org>:
>> >However, aptitude does this:
>> >libarts1c2a recommends libarts1-akode
>> >--\ The following actions will resolve this dependency:
>> > -> Upgrade libarts1c2a [1.4.3-3 (now) -> 1.5.0-2 (experimental, experimental)]
>> > -> Keep libarts1c2a at version 1.4.3-3 (now)
>> > -> Remove libarts1c2a [1.4.3-3 (now)]
>> > -> Install libarts1-akode [4:3.5.0-2 (experimental, experimental)]
>> > -> Leave the dependency "libarts1c2a recommends libarts1-akode" unresolved.
>> >
>> >i.e it prefers to install the experimental version, even if it is priority
>> >-10. The correct solution is to install libarts1c2a 1.5.0-3, and leave the
>> >dependency unresolved.  OR to hold everything.  But installing anything that
>> >has a negative priority is a no-no.
>> >
>> >For reference, my /etc/apt/preferences is:
>> >Package: *
>> >Pin: release a=experimental
>> >Pin-Priority: -10
>>
>> This should not happen with "recent" versions of aptitude, recent as in
>> the last 6 years at least, because solutions involving installations /
>> upgrades of non-default versions are kept in a different level and only
>> offered last, if at all, below upgrading to default versions, removing
>> it or keeping everything at the same version.
>>
>> Have you experience this behaviour recently?
>
> No, but that doesn't mean anything: sometime after I reported the bug, I
> disabled experimental, and I don't think I had any reason to reenable it in
> the last decade (if I did, I most certainly used a disposable chroot, got
> the job done, dropped the chroot in the bitbucket and promptly forgot about
> it).
>
> One would have to test current apt (or better yet, stable apt) to be sure.

>From my reading of the code, tests that I was doing lately, as well
using experimental myself for many years, and from the lack of
"secondings" to this bug or similar complaints in hundreds (or more
like about a thousand) bug reports that I have triaged in the last few
years, I know that this was not a case that happened at least in this
decade, not under normal conditions in any case.  I was just trying to
get confirmation from you as submitter.

There are still hundreds of open bugs.  If the submitters are not
interested if they are still happening, I am not all that interested
in getting down to the last detail of every one of them, when there is
no reasonable indication that ancient bugs continue present.

Still, I just tested this specifically with experimental priority -10
and -1000 and with git-man which caused the full resolver to be
involved, and the offerings to upgrade to experimental always come
last (Remove, Upgrade to unstable, Keep, Upgrade to experimental).

So closing the report now.


Cheers.
-- 
Manuel A. Fernandez Montecelo <manuel.montezelo at gmail.com>