Roman Gaufman wrote:
>There is certainly potential here, but let me list why your method
>will be flawed:
>
>1) Non single user mode means you cant predict what deamon is doing
>what at what time.
>
>
Damn, I knew there was something I missed ;)
>2) CFLAGS in make.conf are usually changed by users often. You need to
>test against the age of /var/log/emerge.log and test against the
>ebuild. Many ebuilds filter out several CFLAGS.
>
>
Sure, I understand.. hopefully they would be the same across the board,
so that would induce a kind of consistent baseline, however I'm sure
they differ across ebuilds of the same software
>3) There's too many entries to fill in, prone to user error, false
>assumptions and ofcourse the people that will just fill in junk for
>the sake of it.
>
>
Yea, good point again.. it's not supposed to be a be-all-and-end-all,
just more of a (very limited) point of reference.. all the same notes
above apply to the forums I suppose.. just depends what people take at
face value
>I think a much better solution is to make a script that will auto
>reboot into single user mode, perform all tests, reboot into mult user
>mode and upload results.
>
>
I agree!
>I could help write some of those tests and benchmarks.
>
>
Cool! It'd be great to have standardisation... everyone singing from the
same benchmark script ;)
Save them to XML and fire them up to a db for everyone to assess.
>I've worked on something similar earlier this year, but its really
>quite a major, complicated task thats a lot harder to achieve than
>just making a simple php page where people fill in values.
>
>Thats why I suggested a performance forum, so things can be discussed
>and polished before a good quality, comprehensive database can be
>made.
>
>
Yea, don't get me wrong I agree wholey with you, was just thinking about
interim solutions as I'm sure to setup the scripts et al. would be a
big, time consuming job!
>
>On Tue, 12 Oct 2004 12:45:27 +0100, Joel Merrick <joel@...> wrote:
>
>
SNIP 8<
--
gentoo-performance@g.o mailing list
|
