Test performance

[Patricia Shanahan <pa...@acm.org>]

I've done a simple survey of the test run times, and the approximately 
20 tests that each took more than 400 seconds on one of my runs 
collectively account for almost 4 hours of the run time.

I plan to look individually at some of those 20 tests to see if there 
are significant savings without reducing the test's effectiveness. I may 
learn some principles that can be applied on a bulk basis to shorter tests.

Patricia

Re: Test performance

[Tom Hobbs <tv...@googlemail.com>]

I couldn't find a central config file - but I'm sure Jonathan knows
more about it than me.

I did to a quick search for all occurrences of "timeout" though.
There is a fair amount of them, I've attached the raw output from this
search just in case it proves interesting/useful.

On Wed, Nov 10, 2010 at 10:08 AM, Jonathan Costers
<jo...@googlemail.com> wrote:
> Hi Patricia
>
> I've noticed a lot of the QA tests take parameters for things as number of
> services to start, sleep times, timeouts, etc. This could be a starting
> point?
> The .td files typically contain these settings, but a lot of them are also
> centrally controlled in a certain config file ... I'll try and dig up what
> that file was again.
>
> Jonathan
>
> 2010/11/9 Patricia Shanahan <pa...@acm.org>
>
>> I've done a simple survey of the test run times, and the approximately 20
>> tests that each took more than 400 seconds on one of my runs collectively
>> account for almost 4 hours of the run time.
>>
>> I plan to look individually at some of those 20 tests to see if there are
>> significant savings without reducing the test's effectiveness. I may learn
>> some principles that can be applied on a bulk basis to shorter tests.
>>
>> Patricia
>>
>

Re: Test performance

[Jonathan Costers <jo...@googlemail.com>]

Hi Patricia

I've noticed a lot of the QA tests take parameters for things as number of
services to start, sleep times, timeouts, etc. This could be a starting
point?
The .td files typically contain these settings, but a lot of them are also
centrally controlled in a certain config file ... I'll try and dig up what
that file was again.

Jonathan

2010/11/9 Patricia Shanahan <pa...@acm.org>

> I've done a simple survey of the test run times, and the approximately 20
> tests that each took more than 400 seconds on one of my runs collectively
> account for almost 4 hours of the run time.
>
> I plan to look individually at some of those 20 tests to see if there are
> significant savings without reducing the test's effectiveness. I may learn
> some principles that can be applied on a bulk basis to shorter tests.
>
> Patricia
>