On Tue, 2018-02-27 at 15:41 +0000, Daniel P. Berrangé wrote:
> > Running syntax-check in all 5 scenarios isn't buying us
anything,
> > as the syntax-check rules don't depend on what is installed in the
> > host. IOW, running syntax-check in 1 scenario is sufficient to get
> > us the coverage we need.
>
> Okay, fair enough. The change still "obfuscates" the Travis
> configuration though, because now you can't just look at a single
> script entry but you have to explode the matrix in your head and
> convince yourself you're covering all bases, so I'm not too happy
> with it.
I don't think we've got so many different scenarios here that understanding
it is a real problem
It's not a massive hurdle, but it's still cognitive load that
I'd rather not have to take on. See my first reply for a way of
achieving the same result in a much more explicit and easy to
grasp manner.
> Moreover, there was a whole thing about just dropping support
for
> precise (as Canonical already did) and making our lives easier
> later in the mail, but you snipped it without replying...
Opps, I'm not in favour of dropping precise, because I think it is useful
to get coverage on older distros. Travis is what I use for testing complex
patch series before submission, so I like it to have a useful mix of vintage
OSs, not only the very latest that is largely the same as what I build on
locally already.
Support for precise is going to be dropped by Travis in two months
either way:
https://blog.travis-ci.com/2017-08-31-trusty-as-default-status
So we can keep it around for the time being if you want, but we're
going to have this very same conversation again pretty soon :)
--
Andrea Bolognani / Red Hat / Virtualization