[Linaro-validation] What can we learn from Travis CI

Michael Hudson-Doyle michael.hudson at linaro.org
Tue Mar 6 02:07:34 UTC 2012


Finally getting back to some old mail!

On Mon, 27 Feb 2012 17:36:50 +0100, Zygmunt Krynicki <zygmunt.krynicki at linaro.org> wrote:
> Hi
> 
> I've looked at this blog announcement [1] and found it interesting.
> 
> Travis seems to be a build service optimized to work with, and
> understand a load of dynamic languages. What I really like about it
> how it makes the effort required to add CI to $my_pet_project very
> very low. I like how they provide HTML snippets that show if your
> project builds or not. I like how they display stats [2]. I like how
> they are integrated with github (it probably makes sense for them but
> the idea is that you loose another "config" mess - no separate login,
> no separate repo setup). I like their UI [3]. They seem to support
> building and testing branches of the single repo from github which is
> also nice [4].
> 
> I know they are focused on CI (build and yes-or-no testing) but is
> there anything else we could learn from them?

I don't know, it's hard to say.  Their use case is a lot simpler than
ours.  They also don't seem to do _anything_ on the interpretation side;
you can see if the tests passed or failed and that's it.

I sort of feel like they are tackling the other end of the problem to us
in some sense: making it possible to run any simple test, where we are
more about running a more fixed set of tests, but with more
depth/complexity in how they are run and how to interpret the results.
Does that seem fair?

That said, they do what they do quite nicely.  It's easy to see what's
going on from just poking around.  It might make sense to think about
that if we were to go back to our front page...

Cheers,
mwh



More information about the linaro-validation mailing list