Re: [ATrpms-users] Building nvidia 310 driver for Fedora 16 x86_64

On Wed, Dec 12, 2012 at 12:16 PM, George Galt <[email protected]> wrote:

> On Wed, Dec 12, 2012 at 11:36 AM, Brian Long <[email protected]> wrote:
>
>> On Wed, Dec 12, 2012 at 9:05 AM, George Galt <[email protected]>wrote:
>>
>>> Brian:
>>>
>>> As I noted above, there appears to be a mismatch between what RPM
>>> believes the package provides and what it requires.  For the 304.51 driver,
>>> the -libs file "provides" libnvcuvid.so.1, and "requires" libnvcuvid.so.1.
>>> However, for the 310.19 driver, the -libs file provides "libnvcuvid.so.1"
>>> but requires "libnvcuvid()(64-bit).  The list of what is "provided" by a
>>> package and "required" by a package is generated by RPM.  It seems that
>>> somehow RPM inspects the 310-libs package and gets the list wrong, but got
>>> it right on the 304-libs package.
>>>
>>> I don't know if there is a way other than the macros to alter the
>>> "requires" list to properly state that the 310-libs package requires
>>> libnvcuvid.so.1, but that is where things are failing.  I hope this helps.
>>
>>
>> Could you just use the following inside the .spec?
>> Autoreq: 0
>>
>> This would make the RPM rely solely on the "Requires:" entries specified
>> in the .spec, but it wouldn't search for automatic requirements.
>>
>> Just an idea.
>>
>> /Brian/
>>
>> _______________________________________________
>> atrpms-devel mailing list
>> [email protected]
>> http://lists.atrpms.net/mailman/listinfo/atrpms-devel
>>
>
> Brian:
>
> As I mentioned somewhere in this thread, I'm new to building rpms, so I'm
> not in a position to evaluate your proposal.  However, assuming it does
> what you say, and the SPEC was populated with a reasonable list, it would
> seem to be a good option.
>

Understood, George.  Look for autoreq in this page:
http://rpm5.org/docs/api/specfile.html

I've used it many years ago when packaging Oracle RDBMS into an RPM.   Boy
was that fun.  :-)

/Brian/