On Tue, 01 Apr 2008 19:58:22 +0200, Bram <[EMAIL PROTECTED]> wrote:

> On Mon, 31 Mar 2008 15:42:28 +0200, Dave Mitchell <[EMAIL PROTECTED]> wrote:
> 
> > Any chance of getting some more smoking going on there?
> 
> 
> Are there any requirements for who can smoke?

No, not really. You have to be able to Install Test::Smoke, and have an
internet connection to fetch the perl source code and to send mail

> And are there any guides on how to start smoking?

None besides the link below.

> All I can find is a reference in perlhack which links to  
> http://search.cpan.org/dist/Test-Smoke/lib/Test/Smoke/FAQ

If you tell us where that document, along with
http://search.cpan.org/~abeltje/Test-Smoke/README
come short in explaining what to do to start, we can improve that.

> What configuration files should be used for the smoke?

Depends on the system you have, and the (compiler) environment.
e.g. It's no use to try to smoke -Duse64bitall if your system
does not support 64bit builds. It's also no use to smoke
-Duselongdouble or -Dusemorebits if your system doesn't support
long doubles.

If your system has reliable locales, esp UTF-8 locales are the
interesting ones, you can add 30% testing time by telling smoke
to test that too.

> As in, when looking at some smoke reports I see -Duse64bitint and  
> -Duseithreads and -Duseithreads -Duse64bitint in one report and  
> -Duse64bitint and -Duselongdouble and -Dusemorebits and -Duseithreads and  
> -Duseithreads -Duse64bitint and -Duseithreads -Duselongdouble and  
> -Duseithreads -Dusemorebits in another report...

The smokers base the combinations on the possibilities and the speed
of the machine. My slowest machine tests less than my fastest box, so
they can all finish withing 24 hours to start the next cycle.

> Or is that decided/asked when running configsmoke.pl?

Most of it is yes.

> Also, how long does a smoke - on average - take? (I suppose this depends  
> on the config?)

Yes, and on the machine speed. (CPU(s), Memory, availability). I myself
also smoke for two different compilers: the system native compiler and
GNU gcc. If gcc is your default compiler (linux), you might consider
smoking with g++ too.

See http://www.test-smoke.org/tinysmokedb.shtml for the combined
results. Here you can also check per OS what reports come in and
how long they took.

some of my smokes for pretty recent blead:

linux 2.6.22.17-0.1-default (x86_64/4 cpus)  - 24 test confs - parallel tests
         2 hours  2 minutes (average 15 minutes 22 seconds)
linux 2.6.22.13-0.3-default (x86/2 cpus)     - 96 test confs
        13 hours 28 minutes (average 25 minutes 17 seconds)
hp-ux B.10.20    (PA-2.0/32/32/1 cpu)  - cc  - 12 test confs
        19 hours 37 minutes (average 4 hours 54 minutes)
hp-ux B.10.20    (PA-2.0/32/32/1 cpu)  - gcc - 12 test confs
        21 hours 10 minutes (average 5 hours 17 minutes)
hp-ux B.11.00/64 (PA-2.0/32/64/2 cpus) - cc  - 24 test confs
        13 hours  5 minutes (average 1 hour 38 minutes)
hp-ux B.11.00/64 (PA-2.0/32/64/2 cpus) - gcc - 24 test confs
        14 hours 20 minutes (average 1 hour 47 minutes)
hp-ux B.11.11/64 (PA-2.0/32/64/2 cpus) - cc  - 24 test confs
         9 hours 36 minutes (average 1 hour 12 minutes)
hp-ux B.11.11/64 (PA-2.0/32/64/2 cpus) - gcc - 24 test confs
        10 hours 29 minutes (average 1 hour 18 minutes)
hp-ux B.11.23/64 (PA-2.0/32/64/6 cpus) - cc  - 24 test confs
        18 hours  2 minutes (average 2 hours 15 minutes)
hp-ux B.11.23/64 (ia64/2 cpus)         - cc  - 24 test confs
         7 hours  7 minutes (average 53 minutes 25 seconds)
hp-ux B.11.23/64 (ia64/2 cpus)         - gcc - 24 test confs
         8 hours 12 minutes (average 1 hour   1 minute)
aix   5.2.0.0/ML04 (PPC/1 cpu)         - xlc - 36 test confs
        14 hours 49 minutes (average 1 hour 14 minutes)
aix   5.2.0.0/ML04 (PPC/1 cpu)         - gcc - 36 test confs
        15 hours 24 minutes (average 1 hour 17 minutes)

Note that older OS versions probably also run on older machines (less
memory or slower CPU's), so blindly comparing on OS version would not
be fair.

Also note that above durations are from smokes that run side-by-side
with maint smokes (perl-5.8.x), so the average system load during smokes
can be very high.

For blead smokes, you might be able to speed up the smokes if you have
multiple CPU's, but that is not working 100% yet.

-- 
H.Merijn Brand         Amsterdam Perl Mongers (http://amsterdam.pm.org/)
using & porting perl 5.6.2, 5.8.x, 5.10.x  on HP-UX 10.20, 11.00, 11.11,
& 11.23, SuSE 10.1 & 10.2, AIX 5.2, and Cygwin.       http://qa.perl.org
http://mirrors.develooper.com/hpux/            http://www.test-smoke.org
                        http://www.goldmark.org/jeff/stupid-disclaimers/

Reply via email to