launchpad-dev team mailing list archive

[Jonathan Lange <jml@xxxxxxxxxxxxx>]

On Tue, Nov 16, 2010 at 5:34 PM, Julian Edwards
<julian.edwards@xxxxxxxxxxxxx> wrote:
> On Tuesday 16 November 2010 17:31:52 Julian Edwards wrote:
>> On Tuesday 16 November 2010 16:11:12 Jonathan Lange wrote:
>> > >  - test suite parallelisation: parallelisation initially, then
>> > >
>> > > parallelisation in our CI/landing robot
>>
>> This, and ...
>>
>> >   * use message queues for something in Launchpad (diff generation
>> >   please)
>>
>> this.  (asynchronous build dispatching please)
>>
>> :)
>
>
> I thought of one more thing:
>
>  * No more freaking sample data
>
> I'm sure that alone will boost our test performance if we don't have to reset
> and load sample data for each test in the DatabaseLayer.
>
> I appreciate that one's quite a lot of work, particularly for Soyuz though :/
>

It is a lot of work. I'd also argue that it's not something that needs
a great deal of architectural direction or planning.

I started a branch in June to remove only the BugBranch sample data. I
used it as a launching point for some productive yak shaving, so it
has not yet landed (although it's now one bald yak). It was
straightforward, tedious, time consuming work. Anyone could do it.

I'm no estimatologist, but I reckon that either parallelizing or a
data mapping layer would have a bigger impact on test run time and
probably be less effort. Once the test suite is faster, removing
sample data becomes *substantially* easier.

jml