On Monday, March 25, 2013 11:05:37 AM UTC-7, Tim.Bunce wrote:
>
> On Fri, Mar 22, 2013 at 10:46:12AM -0700, Ryan Moore wrote: 
> >    On Friday, March 22, 2013 9:34:37 AM UTC-7, Tim.Bunce wrote: 
> > 
> >      On Thu, Mar 21, 2013 at 03:58:13PM -0700, Ryan Moore wrote: 
> >      >    Is there any way to combine NYTProf output from multiple 
> source files to give a useful / 
> >      non-garbage 
> >      >    output? 
> >      >    The docs indicate that you might be successful running 
> nytprofmerge with two runs of the same 
> >      source, 
> >      >    but trying to combine multiple sources will probably not do 
> what you expect. I'm trying to 
> >      generate a 
> >      >    combined profile for our tests running under prove (220 files 
> in t/ and counting), 
> > 
> >      Why, specifically? 
> > 
> >    Concentrating effort on improving the test setup itself. On any 
> individual test a 1-second improvement 
> >    might not be the best thing to focus on (or biggest timesink), but if 
> that same one-second improvement 
> >    is propagated to all 220 tests that's 3+ minutes of improvement for a 
> full test run. 
>
> I see your point. I'd guess that improving a random few, focusing on 
> common code paths, would have a similar effect. 
>
> I presume you're using a modern version of prove, which has many good 
> features for avoiding rerunning all tests. 
> See --state slow/fast etc in 
> https://metacpan.org/module/OVID/Test-Harness-3.26/bin/prove 
> Perhaps that would be of some help. 
>
>
> >      What's the problem with the current output from nytprofmerge? 
> > 
> >    It seems to be only reporting a fraction of the time used overall, 
> i.e. for a 30 minute test run it 
> >    shows a function with 4 seconds of total usage as the top offender, 
> and doesn't seem to include all of 
> >    the t/* files in the output. 
> > 
> >      In theory you might be able to hack on nytprofmerge to get 
> something 
> >      more useful out. 
> > 
> >    I'm unfortunately a Perl n00b, but I'll take a look to see if there's 
> anything obvious in nytprofmerge 
> >    that applies to my use case. I'll also see if I can create a repro 
> case with less than the 
> >    several-hundred sources involved in my current failures. 
>
> In my (very quick) testing it seems to work ok. 
>
> Small (ideally very small) test cases would be a big help. 
>
> Tim. 
>

It turns out that one out of N files was aborting early and generating an 
empty/incomplete nytprof.out file, and merging that one with the other 
working files made everything unhappy. I have yet to track down which test 
is borking the rest, it might be something intermittent since testing 
[a-m]*.t and [n-z]*.t separately both show no failures.

Thanks for looking into a stupid-user-error, I'll send a repro case for the 
"bad nytprof.out merges badly" if I can isolate it.

 - Ryan

-- 
-- 
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]
--- 
You received this message because you are subscribed to the Google Groups 
"Devel::NYTProf Dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/groups/opt_out.


Reply via email to