On 27/09/2007, Rajith Attapattu <[EMAIL PROTECTED]> wrote:
>
> Thank you very much. This kind of cynicism helps this project a lot.
> I give up trying to explain it anymore. We can definitely continue to go
> our
> merry ways.
>
> Regards,
>
> Rajith
>
> On 9/27/07, Robert Godfrey <[EMAIL PROTECTED]> wrote:
> >
> > >
> > > It is a problem as people continue to write and run the wrong kind of
> > > tests
> > > under client and broker.
> >
> >
> >
> > I think we should all be grateful when people contribute tests, and if
> we
> > believe there are gaps (i.e. in micro-tests of individual emthods) then
> > one
> > should write them onself rather than complaining that other people
> haven't
> > :-) :-)



Rajith,

I'm not trying to be cynical... but we've had this discussion several times
before I don't think it does anyone any good to complain about other people
doing the tests "wrong" in the past.  Stating things in absolutes of right
and wrong definitely does none of us any favours.

I do think it is much more productive to actually write more tests.  And
*none* of us have been conspicuous in doing that.  My previous mail was as
much to indicate that I am not sticking my oar in because I would feel a
hypocrite for saying how things ought to be when I have been no more active
in contributing tests than anyone else.

Now, I am all for re-organising the tests if it makes more sense and so on
(they are surely a mess at the moment).  But I think that complaining about
others is extremely counter-productive.  I also worry that once again we
will focus our energies on "re-organising" to the way we individually might
like things, rather than actually moving things forward (OK - this I admit
is cynical).

I know the tests have been "re-organised" once in the past when things were
moved over to maven, and though I was only a witness I know that this
re-organisation caused a lot of people who wrote those tests originally to
get quite annoyed that things had been broken.  I also don't think it helped
anything very much.

The general thrust of this discussion has been good.  And I think I like the
way the discussion is going - however I do *not* want to rush into
anything.  Further there is always a temptation to move the existing tests
"out of the way" to make room for pure unit tests, and to remove the
existing tests from the build.

This would be a *HUGE* mistake.

The first concern at the moment *HAS* to be getting M2 out.  That means
running the existing tests and finding out why they are failing
intermittantly.  Having them fail in a differnt directory isn't exactly
progress :-) <-- note the smiley

I understand people want to separate out the tests into their correct
locations, to properly distinguish between unit, component, system tests
etc... To distinguish between tests which are JMS (not Qpid), tests which
are AMQP (e.g. the Python tests) etc...  We should involve the C++
developers, the .net developers, python developers etc to look at testing in
Qpid as a whole.

However I think the first step should probably be to examine and document
what tests we have at the moment.  We already have people who are possibly
existing tests that are already there.  The problem is that such a
documentation task isn't high up on anybody's list of fun tasks... and I
doubt that anyone is going to volunteer to do it?  Like I said, I'm a
hypocrite since I'm not going to commit myself to doing this, even though I
know it needs to be done.

Anyway, before anyone actually goes and makes a change on the organisation,
can we *please* try to present a detail design and rationale before doing
so?  And it would be good if it covered the whole of Qpid and not just the
Java parts.  We should be able to leverage tests done in other languages -
surely?

Your Thoughts?

-- Rob




>
> > -- Rob
> >
>

Reply via email to