On Sat, Feb 27, 2010 at 09:35:25PM +0000, Nicholas Clark wrote:
> On Sat, Feb 27, 2010 at 09:14:37PM +0000, [email protected] wrote:
> > Revision: 1086
> > Author: tim.bunce
> > Date: Sat Feb 27 12:12:54 2010
> > Log: Fix so $cpu1 on entry to foo() is accurate. (Test was failing on OS X
> > with '0.39'>='0.396'. Maybe the * 0.99 for freebsd is no longer needed.)
> 
> Works on my machine.
> 
> (Well works once on not-my-machine-but-I-have-an-account-on-it)
> 
> Ship it!*
> 
> I *think* that part of the "problem" is that the test counts all CPU use
> when busy-looping to 0.4s, whereas the profile only counts time in the
> subroutine towards its tally of CPU time. So if enough "100"ths of a second
> happen to fall into the time that NYTProf was running, the sum of the count
> for the subroutine will be legitimately under.

Umm. I don't think so. The syncing to a cputime tick before entering
foo() should avoid that. (Assuming I understand what you mean.)

I think not counting sys time and having $cpu1 being one tick behind
were the key problems.

> I think that it's actually at risk of hitting the problem that sinks DProf:
> 
> http://blog.timbunce.org/2008/07/12/devel-dprof-broken-by-the-passage-of-time/
> 
> Nicholas Clark
> 
> * heck, given that it's "It compiles - ship it", running the tests is
>   positively extravagant.

:-)

We're very close. If the _96 gets good cpantesters results (excluding
windows) then I'll ship it.

Fixing it for windows would be nice but I won't hold up a release for it.

Tim.

-- 
You've received this message because you are subscribed to
the Devel::NYTProf Development User group.

Group hosted at:  http://groups.google.com/group/develnytprof-dev
Project hosted at:  http://perl-devel-nytprof.googlecode.com
CPAN distribution:  http://search.cpan.org/dist/Devel-NYTProf

To post, email:  [email protected]
To unsubscribe, email:  [email protected]

Reply via email to