Front page | perl.perl5.porters |
Postings from May 2016
Re: [perl #128113] [PATCH] upgrade Test-Simple in perl blead
Thread Previous
|
Thread Next
From:
Chad Granum
Date:
May 12, 2016 01:39
Subject:
Re: [perl #128113] [PATCH] upgrade Test-Simple in perl blead
Message ID:
CAJFr3kv7+LUhip0WJRqNiNZv7b40frESn=wHz8T0o+5O2QAixQ@mail.gmail.com
The change ultimately results in a much faster runtime, but a slower
startup time. Test2 takes twice as long to load, but runs twice as fast
once loaded. The number of test files vs the length of the test files
usually explains why some suites are faster or slower.
I am not familiar with imager, does it have a large number of test files?
On May 11, 2016 5:05 PM, "Tony Cook via RT" <perlbug-followup@perl.org>
wrote:
> On Tue May 10 19:24:32 2016, public@khwilliamson.com wrote:
> > On 05/10/2016 08:18 PM, James E Keenan via RT wrote:
> > > On Tue May 10 08:08:00 2016, exodist7@gmail.com wrote:
> > >> This patch will update perl blead to use the latest Test-Simple
> > >> distribution.
> > >>
> > >> -Chad
> > >
> > > Question for porters: > 1 year ago, at an earlier stage of exodist's
> > > work on Test-Simple, our practice was to apply his patches directly
> > > to blead. The rationale for that was that this distro was so far
> > > "upriver" that we wanted to get it smoked as widely as possible --
> > > and blead gets smoked more than any smoke-me branch.
> > >
> > > Shall we continue this practice with these latest patches?
> > >
> > > Pending an answer to that question, I have created this branch for
> > > smoking:
> > > smoke-me/jkeenan/exodist/128113-test-simple
> > >
> > > Thank you very much.
> > >
> >
> > It seems prudent, whatever the final outcome, to wait a bit for
> > results
> > from your branch before doing a wider smoke. Obviously, if there are
> > problems, we wouldn't want to proceed with a wider net.
>
> The results from the smoke branch looked good - the failures were typical
> intermittent failures for the smokers involved (timing issues on the VMs
> (cygwin and solaris), the occasional free2 failure on neptune, and the
> speed.t failure on neptune is reasonable (neptune is a busy box.)
>
> I benchmarked with and without the Test2 patch, and the details can be
> found at:
>
>
> https://docs.google.com/spreadsheets/d/1QV6g8q8FWTL_MiFsBe24bHViKCDU3x4Z5bryrDUVFs4/edit?usp=sharing
>
> Testing perl took ~7% longer on average for non-parallel testing and 3.5%
> longer for 6-job parallel testing, which isn't too bad.
>
> Testing Imager was a bit more worrying, non-parallel testing took ~16%
> longer and parallel testing took ~11% longer.
>
> But I can only consider that the cost of the increased flexibility that
> the new Test system will provide.
>
> In any case we're unfrozen, the extra costs aren't prohibitive, so this
> has been applied as b4514920cd5cabccad6add35edf1bef258070a11.
>
> The patch updated win32/makefile.mk and win32/Makefile, but missed
> updating win32/GNUmakefile, which I've done in
> ec35cd4c022dea519712ce60efb24e281b048471.
>
> Tony
>
> ---
> via perlbug: queue: perl5 status: open
> https://rt.perl.org/Ticket/Display.html?id=128113
>
Thread Previous
|
Thread Next