From: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Date: Tue, 10 Aug 2010 21:00:11 GMT
Subject: Re: [PATCH] git-notes: Run partial expensive test everywhere
Message-ID: <AANLkTikO6_cBBywSdvSdpWNO8OgSzmp9EB-xQ7Y_s4qf@mail.gmail.com>
In-Reply-To: <AANLkTim2tozO31HKLwxi6kWyZKYMd9egDhS5U_gb5OvZ@mail.gmail.com>

On Tue, Aug 10, 2010 at 20:29, Sverre Rabbelier <srabbelier@gmail.com> wrote:
> Heya,
>
> On Tue, Aug 10, 2010 at 14:56, Ævar Arnfjörð Bjarmason <avarab@gmail.com> wrote:
>> The git-notes expensive timing test is only expensive because it
>> either did 1k iterations or nothing. Change it to do 10 by default,
>> with an option to run the expensive version with the old
>> GIT_NOTES_TIMING_TESTS=ZomgYesPlease variable.
>
> Nice, why 10 though? Any motivation for that particular value?

The old version had "for count in 10 100 1000 10000; do". Mine has 10
as non-expensive, and "for count in 100 1000 10000; do" as expensive.

I.e. I'm running the first test batch from the old tests.

I have no idea whether it actually needs to run 10..10k times, I
didn't try to grok the actual test code.

>> The old version would also mysteriously fail on systems without
>> /usr/bin/time, there's now a check for that using the test
>> prerequisite facility.
>
> Should this patch be split up?

It all touched the same bits, it'd be nastier to split it up IMO.