subunit-dev team mailing list archive
-
subunit-dev team
-
Mailing list archive
-
Message #00024
Re: timing data, the time stamp facility and test activity granularity
On Thu, 2009-08-27 at 11:05 +1000, Jonathan Lange wrote:
> > I think its going to fit somewhere in the gap between
> > tests take care of testing
> > results take care of reporting
> >
> > - in a model of tests being a black box what things do what is internal.
> > - we need some opaque-structured way to expose things that tests do
> > (structured additional data)
> >
>
> I think this is a good analysis.
>
> > Having that structured data be taggable in the same way that the
> > outermost stream is would allow a convention for reporting that may work
> > well, and allow TestResources and other similar things to collaborate.
>
> Sounds good to me.
Some further thoughts:
- test resources can already log events when tearing up and down
(via a ugly hack in python, but still)
- perhaps nesting isn't needed or appropriate, just some events that
are understandable-by-convention (e.g. hide-from-the-ui even if you
don't recognize the directive).
- 'slow setup' isn't very different to 'slow test'. details matter
much more at that level - having separate resources that provide
their own overall time is great, but 'bzr selftest --lsprof-tests'
(this exists) is better (when combined with some analysis logic).
-Rob
Attachment:
signature.asc
Description: This is a digitally signed message part
References