[Rpm-maint] [PATCH] - Added support for ppc64p7 arch (Power7 and higher optimized)

Panu Matilainen pmatilai at laiskiainen.org
Tue Apr 17 06:36:09 UTC 2012

On 04/16/2012 08:14 PM, Phil Knirsch wrote:
> On 04/13/2012 01:32 PM, Panu Matilainen wrote:
>> On 04/13/2012 01:25 PM, Phil Knirsch wrote:
>>> On 04/12/2012 11:48 AM, Panu Matilainen wrote:
>>>> On 04/11/2012 12:58 PM, Phil Knirsch wrote:
>>>>> Attached is the updated version of the ppc64p7 support patch. I've got
>>>>> rid of the AUXV env reading and doing it properly now via reading and
>>>>> parsing /proc/self/auxv. Additionally the patch now checks if the
>>>>> AT_PLATFORM is power7 or higher via a sscanf(), allowing future power
>>>>> architectures to benefit from it as well.
>>>> The patch leaks file descriptors, the fd it opens is never closed.
>>>> In the cosmetics department you might want to split the long if with
>>>> sscanf() into two separate if's (or otherwise reformat) to avoid the
>>>> large line-wrap which is easily avoidable.
>>>> Other than that, looks okay to me.
>>>> - Panu -
>>> Alright, latest version attached. I've basically extracted the auxv
>>> parsing into a separate function, added a static struct that stores the
>>> info, fixed the fd closing problem and using the struct now in the
>>> defaultMachine().
>> Cool, this should be useful for various other architectures as well so
>> we wont end up with a dozen slightly differing variants of
>> /proc/self/auxv and /proc/cpuinfo parsing in the tree.
>> One more request :) Can you split the patch in two: one that adds the
>> auxv parsing and another which adds the new arch bits? Oh, and I think
>> it'd be useful to initialize rpmat.platform to "" so its users wont have
>> to bother checking for NULL to avoid segfaulting in case the auxv
>> parsing happened to fail.
>> - Panu -
> Alright, split it in 2 patches now and i've initialized rpmat.platform
> in the parse_auxv() function to "".

Thanks, applied + pushed to master now.

> Would it be possible to get an updated version of rpm out for Fedora 17
> with the changes so we can use this for building packages there on PPC?

I dont see any major reason why not...

	- Panu -

More information about the Rpm-maint mailing list