QA in Squeak (was: Re: [Bug][Fix] Setting
copy/paste-keypreference under Windowsdoesnot work)
Andreas Raab
andreas.raab at gmx.de
Mon Dec 6 23:29:08 UTC 2004
Tim -
> I'm not convinced of that. I'd much prefer to be able to have proper
> reviews by several people whenever possible. I'm trying to keep the VM
> stuff 'clean' as best I can and it's somewhat helped by VMMaker being
> an external package. But the practical issue is getting people's
> eyeball time for reviewing; we're all pretty busy and this is entirely
> volunteer so what can one do? Fire someone?
Certainly not. But just having eye-balls is entirely useless when it comes
to these areas. In fact, I think it's worse to have some random eye-balls
than having no review whatsoever; at least you know that noone knowledgable
has looked at it. The problem with random eye-balls is that it gives you the
feeling that things ought to be fine (it's been reviewed) where in reality
it hasn't.
(and trust me, I am really glad that we have VMMaker as external package - I
won't even try to imagine what might happen to it if you weren't acting as
the person who can place a veto if things get entirely out of hand)
> We all get to make mistakes, no matter how experienced or expert we
> are. I didn't spot Andreas' mistakes in the 'isArray' VM changes
> introduced in 3.6 and thus we had to bugfix later, a classic example of
> trusting an experienced expert and not doing enough cross checking.
> When I was at ParcPlace I got to correct a few bugs by Peter Deutsch -
> Peter for ghu's sake - in the translator.
We all make mistakes. That's okay. And -quite honestly- I wouldn't have been
so upset if there would have been no reviews of that change - the only
person to blame then is myself for not looking at it. But the problem starts
when you look at these reviews and the process behind it. The reviews were
essentially all of the kind "works for me". Then, a request for inclusion
was made and some harvester - based on the CS, the reviews and the request -
made the decision to approve and post the change. As you can see, the entire
process looks perfectly fine, there is noone to really blame for what
happened, except that things should not have gotten as far without someone
looking at it who understands about the wider implications of that change.
We are talking about a language change for god's sake! The only one that
actually made it into Squeak in ten years - based entirely on "works for me"
reviews.
Cheers,
- Andreas
More information about the Squeak-dev
mailing list
|