Re: [Boost-users] Using static code checkers against the Boost code base

Subject: Re: [Boost-users] Using static code checkers against the Boost code base
From: Niall Douglas (s_sourceforge_at_[hidden])
Date: 2014-12-06 09:07:06

On 5 Dec 2014 at 18:19, Richard wrote:

> >Setting this stuff up is not free, including renting the CI testing 
> >resources.
> 
> I've setup cppcheck to run on travis-ci.org from my github repo.  That
> approach is completely free.

Travis is okay for Linux only toy CI. It is better than nothing, but 
it's very toy - no results recording, no soak testing, no scheduled 
testing, no automated fork reconciliation, test and pushing etc etc.

It doesn't even approach remotely a proper CI setup. Even my Jenkins 
CI dashboard for AFIO (https://boostgsoc13.github.io/boost.afio/) 
isn't a patch on a professional job, and I have four separate 
analyser passes in there. What I really need in there is historical 
performance regression analysis and bisection soak testing, but I 
just don't have the time.

Also, with Boost.Thread at least, a large chunk of the unit test 
suite would need upgrading to output results which can be consumed by 
a CI. Right now a large chunk doesn't even use Boost.Test, and the 
unit test suite is woefully incomplete from what it should be. And 
that's just one small Boost library.

Good test engineers get paid more than good developers in recent 
years. I agree with that market assessment.

Niall

-- 
ned Productions Limited Consulting
http://www.nedproductions.biz/ 
http://ie.linkedin.com/in/nialldouglas/