Success report on Linux/PPC, small Ada problem

[mike stump]

> To: mrs@windriver.com, rth@redhat.com
> Date: Tue,  5 Feb 2002 22:36:44 -0500 (EST)
> From: dewar@gnat.com (Robert Dewar)

> The changes to the FSF repository cannot possibly meet this
> criterion, and in any case I don't think they should have to.

:-(

I think it is reasonable to solicit companies with large proprietary
testsuites to run them on the changing top of tree, and provide us
quality feedback on those areas that we may be breaking.  Let's call
it black box testing.

I think that as we mature, we have to do this, and that this would be
good for the project.  I think it allows us to attain a level of
quality that we would otherwise be hard pressed to meet.

I would love to see you guys feed such information back into the top
of the tree so that we can try and achieve the level of quality you
would like.  Certainly my company's use of the tree fall into the life
critical and beyond applications.  It is advantageous to us, if gcc
meets a high quality standard, not just every once in a while (say, at
release time), but on a continuous basis.

My experience is, if you give developers the information on quality
early and often, they can usualy ramp up to meet the mandate of no new
regressions and stay there most of the time within 8 months.  Though,
at first, as you allude to, they can't possibly meet this.

As machines become faster and faster, and disk space, cheaper and
cheaper, and humans, more and more expensive, it makes more and more
sense to do more and more testing in an automated fashion to the
developers.