Broken dependencies: netcdf (gcc-gfortran(x86-32))

Orion Poplawski orion at cora.nwra.com
Mon Dec 20 19:27:42 UTC 2010


On 12/20/2010 11:32 AM, Jussi Lehtola wrote:
> On Mon, 20 Dec 2010 10:20:49 -0700
> Orion Poplawski<orion at cora.nwra.com>  wrote:
>
>> On 12/13/2010 11:04 AM, Orion Poplawski wrote:
>>> On 12/13/2010 09:24 AM, Kevin Fenzi wrote:
>>>> On Mon, 13 Dec 2010 08:47:54 -0700
>>>> Orion Poplawski<orion at cora.nwra.com>  wrote:
>>>>> Ah, I didn't catch the multilib aspect (32-bit missing on x86_64).
>>>>> So, this needs to be fixed in the EPEL multilib compose?
>>>> It can't be. It's the way it is in the rhel repos. We could copy
>>>> and re-create the rhel repos, but thats not something I think we
>>>> want to do.
>>>>
>>>> kevin
>>> What I meant was - could we exclude items that require
>>> gcc-gfortran(x86-32) from the EPEL x86_64 repos?
>>>
>> Does anyone know how I/we can fix this?
> Drop the Requires: gcc-gfortran%{_isa} and it will work.
>
> Maybe someone should mention this in the guidelines; this won't work on
> any Fedora or RHEL until x86_64 gfortran is made to provide
> gcc-gfortran(x86-32)... [Although I don't know if 64-bit gfortran ran
> with -m32 produces the same modules as 32-bit gfortran]
This works perfectly fine on Fedora because gcc-gfortran.i686 is present 
in the x86_64 repo.
I guess dropping the %{_isa} would prevent the broken dep at the expense 
of perhaps being useless.




More information about the epel-devel-list mailing list