Re: freshrpms-list Digest, Vol 762, Issue 1

On Jun 7, 2006, at 12:00 AM, [email protected] wrote:
>
> Message: 2
> Date: Tue, 6 Jun 2006 13:04:19 +0200 (CEST)
> From: Dag Wieers <[email protected]>
> Subject: Re: Latest build oops. (Dag Wieers)
> To: "General freshrpms.net discussion list"
> 	<[email protected]>
> Message-ID: <[email protected]>
> Content-Type: TEXT/PLAIN; charset=US-ASCII
>
> On Mon, 5 Jun 2006, Dean Takemori wrote:
>
>> I generally build all packages myself (with --target=athlon).  90% of
>> the time (apart from the athlon part), it makes no difference -  
>> especially
>> with the perl/noarch stuff.
>
> Does it make a big difference when you build them with -- 
> target=athlon ?
> Even if it doesn't, it's nice to get bugreports from someone who's
> building in parallel. I'd encourage more people to do that and report
> problems.

Truth is that I've never benchmarked my builds vs the standard i386  
ones.
It's mostly a "if I'm going to build them myself anyway, I might as well
try to squeeze a few percent performance" kind of thing.  Besides, the
process has taught me quite a bit, and I can turn around and pay the
community back in a small way by reporting the minor bugs I find.

>> But once in a while I've caught a small
>> .spec bug, missing dependency or some such anomaly, so I report  
>> them when
>> I find them.  (In a few rare cases, I _must_ build myself due to some
>> incompatibility I introduced with my customization.)
>
> I'm interested to know what this means :) What customization need  
> custom
> builds ? Maybe there's some broader use for this so we can  
> accomodate this
> customization ?

Generally speaking, things that depend on particulars of the kernel.
(I run off a heavily modified FC1 kernel+modules with the release
versions - from tarball/patch - of ALSA, StrongSWAN, lm_sensors and
a driver patch for one of my ethernet cards).  If I recall correctly,
ALSA dependencies were the reason I started rolling my own binary rpms
two years (give or take) ago.

-dean takemori