2009/7/21 Goswin von Brederlow <[email protected]>:
> Michal Suchanek <[email protected]> writes:
>
>> 2009/7/21 Goswin von Brederlow <[email protected]>:
>>> Michal Suchanek <[email protected]> writes:
>>>
>>>> Package: ia32-apt-get
>>>> Version: 22
>>>> Severity: important
>>>>
>>>>
>>>> I tried installing wine and it forces installation of libc6-i386 from
>>>> the i386 repository as opposed the amd64 one which fails.
>>>>
>>>>
>>>> E: Couldn't configure pre-depend libc6-i386 for libc6-i386, probably a
>>>> dependency cycle.
>>>> A package failed to install.  Trying to recover:
>>>> Reading package lists... Done
>>>> Building dependency tree
>>>> Reading state information... Done
>>>> Reading extended state information
>>>> Initializing package states... Done
>>>> Writing extended state information... Done
>>>
>>> It has resolved itself now as 2.9-21 has been autobuild on amd64. But
>>> the situation can be avoided in the future.
>>>
>>> libc6-i386 (and the other lib32* packages) need to be filtered out for
>>> amd64. There shouldn't be a libc6-i386 i386 package on amd64, only on
>>> ia64. My bad.
>>>
>>> If you pin 32bit packages lower than 64bit packages then the problem
>>> goes away too. See the docs for example entries.
>>>
>>>
>>> I wonder though why aptitude picks the uninstallable package in its
>>> solution. I've seen it pick a lower version and even lower pin for a
>>> package because that would make it installable. I would have thought
>>> it would pick libc6-i386 2.9-20 (amd64) instead of 2.9-21~22 (i386)
>>> here too.
>>>
>> I would guess it's because wine depends on 2.9.21 so the 2.9.20 is not
>> sufficient. Otherwise there would be no reason to upgrade from 2.9.20
>> after it was pulled in by gcc-multilib.
>>
>> Thanks
>>
>> Michal
>
> That would do it.
>
> MfG
>        Goswin
>

Which would mean that every time the libc6 in i386 is ahead of libc6
in amd64 packages compiled against the new libc won't be installable.

Thanks

Michal



--
To UNSUBSCRIBE, email to [email protected]
with a subject of "unsubscribe". Trouble? Contact [email protected]

Reply via email to