On Thu, Jul 24, 2008 at 02:03:13PM -0700, Dan Price wrote:
> So will you be checking in a baseline?  [Hopefully? :)]

Yes :)

> I was wondering if perhaps the baseline should be in some sort of
> text-tool-processable format?  (I was thinking one line per test).  I
> was thinking that people might want to be able to review changes which
> have been made to it by using 'diff'.  I know that may be more of a pain
> than a 'pkl'.
 
OK, sounds good.  I was planning to do that but just figured I'd get the stuff
up for review before going further.

> For human consumption, I'm wondering if it would be
> helpful just to show the complete path to the test in
> order:
> 
> api.t_catalog.TestCatalog.testcatalogfmris2             pass
> 
> I also have a preference for eliminating "..." sequences, as
> I personally find them more distracting than whitespace.  Maybe
> it's just me.
> 
> Another option would be to do more of a tabular form:
> 
> api.t_action.TestActions:
>     test_action_errors                  pass
>     ...

So this is possible but would require some surgery.  We don't currently have
that level of granularity, but we could if we made each "class" into its own
suite, and then ran that.  If people feel strongly about it'll certainly go
that route, otherwise changing to the suggestion before that (or using Danek's
suggestion) is pretty easy.  Here's what the output looks like now:

api.t_version.py TestVersion.testversionsuccessor9          pass

Is it reasonable to left justify the test name to 60 columns?  That gives a
good whitespace separation for most tests.

> In pkg5unittest.startTest(), make sure to flush the stream... that way
> if the test hangs, we know what test we're running.  There
> may be other places where a flush is also useful, but I'll leave
> it to you to find them.

Done.

> > ... a failing test that normally passes and you just broke ...
> > 
> > testrolllogfiles (api.t_catalog.TestUpdateLog) ... FAIL
> 
> So as we talked about yesterday--  there are four cases.  Can
> we get example output for each?
> 
>            Result               Baseline        Output
>              Pass                   Pass        ??
>              Pass                   Fail        ??
>              Fail                   Pass        ??
>              Fail                   Fail        ??
> 
> > FAILED (failures=1)
 
So I just made that change; does this look ok? (baseline result being in
parens)

api.t_version.py TestVersion.testversionsuccessor9          pass
api.t_version.py TestVersion.testversionsuccessor9          pass (fail)
api.t_version.py TestVersion.testversionsuccessor9          FAIL (pass)
api.t_version.py TestVersion.testversionsuccessor9          FAIL

> Would it be possible to update this "(failures=1)" output to reflect
> the four cases listed above?

Do you have a suggestion on what that would look like?  I'm thinking:
(pass=2, fail=2, baseline_mismatch=2)

Thanks for reviewing, I'll post an updated webrev after the de-pkl'ing.

Thanks,
Brad
_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss

Reply via email to