launchpad-dev team mailing list archive

[Jonathan Lange <jml@xxxxxxxxxxxxx>]

On Mon, Oct 4, 2010 at 5:47 AM, Michael Hudson
<michael.hudson@xxxxxxxxxxxxx> wrote:
> On Mon, 4 Oct 2010 12:22:57 +1300, Robert Collins <robert.collins@xxxxxxxxxxxxx> wrote:
>> On Mon, Oct 4, 2010 at 11:40 AM, Michael Hudson
>> <michael.hudson@xxxxxxxxxxxxx> wrote:
>> > On Mon, 27 Sep 2010 10:41:51 +1200, Robert Collins <robertc@xxxxxxxxxxxxxxxxx> wrote:
...
>> >> 5) For low-overhead iteration
>> >> * Find an existing external process
>> >> * Must 'know' its config a-priori
>> >> -run tests-
>> >> Â * check the process is running, do per-test isolation
>> >> Â * run a test
>> >
>> > Hm. Â This feels a bit tricky, but I guess you don't need this at the
>> > same time as concurrent testing...
>>
>> Ideally starting things like appservers and the librarian wouldn't
>> take 6-7 seconds. removing zcml will probably fix that (and then we
>> can remove the low-overhead iteration story completely).
>
> Um, I'm not sure what you mean by removing the ZCML?  It was my
> impression (and hey, I might be wrong) that it was the work the ZCML
> implied not the ZCML itself that takes up most of the time of setting up
> the component architecture.  Although I agree -- it would be nice if
> services started really fast.
>

Benji & I did some profiling of this at the last Epic. Sadly, most of
the details escape me. IIRC, the parsing & the work itself both take a
fair chunk of time. The real problem is just how much of it we have.

jml