Package depends on epoch, but why

[Panu Matilainen]

On Sun, 30 Apr 2006, Richard Bos wrote:
> Thanks Panu,
>
> Op zondag 30 april 2006 13:27, schreef Panu Matilainen:
>>> AFAICS, no references to epoch (0:...).  Why does apt determine that
>>> dbus-1-* depends on 0:0.60...?
>>
>> Because createrepo stomps zero epochs everywhere if epoch doesn't exist
>> and that doesn't mix well with reality on pre 4.2.1 rpm versions. To add
>> to that, lorg3 tries to hide away those false epochs partially, mixing up
>> the situation even more for old rpm versions. What version of rpm does
>> 10.1 have, is it still rpm-4.1+patches?
>
> suse101> rpm -q rpm
> rpm-4.4.2-40
>
> That looks sufficiently new.

Indeed, rpm-4.4.2 should behave just fine with lorg3 as is, without 
patching up createrepo or anything else. The fact it doesn't means that 
Suse has incompatible patches in their rpm - or something like that. What 
a mess :(


>> See the thread starting at
>> http://lists.laiskiainen.org/pipermail/apt-rpm-laiskiainen.org/2006-April/0
>> 00024.html, RHL 9 has similar problems because of the false zero epochs
>> reported by repomd data.
>>
>> Unless I'm missing something, the only solution is to have the metadata
>> not lie about epochs, but that breaks smart and yum which both expect the
>> epoch property to be always there and be zero for non-existing epochs. The
>> remaining option is to add a switch to createrepo to make it possible to
>> disable adding those false epochs in the metadata:
>> https://lists.dulug.duke.edu/pipermail/rpm-metadata/2006-April/000627.html
>
> Well hopefully this patch makes it into the mainstream package....
>
> I guess for now I'm stuck to manually upgrade the packages....,

You can try applying this patch, it'll make apt treat repodata exactly 
the way it is:
svn diff -r83:82 http://svn.laiskiainen.org/apt-rpm/trunk/
(note that it's reversed on purpose)

I'll probably have to revert that patch anyway but it'll cause other 
issues elsewhere :(

> I think that apt will not complain anymore once the packages have been 
> installed....

It shouldn't, because the installed packages have untampered dependency 
information.

Oh and btw Richard, please subscribe to the list so I don't need to 
manually approve your posts :) I'd prefer keeping this subscriber-only 
list to avoid spammers.

 	- Panu -