develooper Front page | perl.perl5.porters | Postings from September 2011

Re: A summary of Chip patches and potential impact

Thread Previous | Thread Next
From:
Steffen Mueller
Date:
September 5, 2011 12:28
Subject:
Re: A summary of Chip patches and potential impact
Message ID:
4E6522CD.9010504@cpan.org
On 09/05/2011 05:56 PM, George Greer wrote:
> On Mon, 5 Sep 2011, Steffen Mueller wrote:
>
>> I assume that at around 20k distributions, it'll be done by the end of
>> the working week. It may be a bit slower if I have to restart it
>> often. I would not feel entirely comfortable spreading this out over
>> multiple machines since the environments would be subtly different.
>> But if we had a fast Intel octu-core (16 threads), with plenty of
>> local storage we'd probably rip through CPAN in two days.
>
> Anybody happen to have US$4,000 lying around to buy one of those
> processors? I think you'd be better off buying a dual-socket motherboard
> (~US$260) and putting in two 6-core CPUs (~US$600 each). Imagine
> 24-threads! Although if you're going big then why not use that
> dual-socket for two 8-core CPUs for 32-threads? Ah, whimsy...

I was actually thinking of two quad-core Xeons here. I will be able to 
lay my hands on such a system for this purpose soon. Stay tuned. My 
musings weren't completely random.

> On a more realistic side, I have an old Core i7 4-core/8-thread that I'd
> use for the cause if you can automate it in some fashion beyond the
> current babysitting. I run 'make -j' smoke runs and with them nice'd I
> generally don't notice they're running even with load averages in the 80s.
>
>> * The blacklisting needs rethinking. I'm simply stripping the bad
>> distributions out when assembling the original input todo list. This
>> can mean that those distributions are pulled in by others as
>> dependencies!
>
> I assume this means you're not using the normal CPAN smoke tools?
> (CPAN-Reporter-Smoker or CPANPLUS-YACSmoke)

I am using CPANPLUS-YACSmoke and Chris' SmokeBox.

> http://wiki.cpantesters.org/wiki/QuickStart

Will check that out.

> You'd have to tell those modules to save reports to files instead of
> delivering to the metabase though, at least for the custom branch, so
> that you don't subject the CPAN authors to failures by mad scientists.

That's what I did.

I'll just have to figure out how to tell the tools to keep their hands 
off certain dists. I never said it's not a conceptually solved problem!

--Steffen

Thread Previous | Thread Next


nntp.perl.org: Perl Programming lists via nntp and http.
Comments to Ask Bjørn Hansen at ask@perl.org | Group listing | About