Boost Testing :
Subject: Re: [Boost-testing] New tester
From: Beman Dawes (bdawes_at_[hidden])
Date: 2009-09-03 10:37:35
On Thu, Sep 3, 2009 at 3:32 AM, Christopher Jefferson <chris_at_[hidden]
> I'm hoping to start regression tester, in particular testing Boost.MPI, on
> Mac OS X, both with the system compiler and a more up-to-date gcc which
> supports C++0X. As a test I am submitting caj-darwin-mpi.
Thanks. You might want to just call the tests caj-mpi, since "Darwin" is
reported automatically, and as long as you are testing a single compiler,
screen width usage is a concern.
Thanks! It would also be a help if you keep an eye on the test results, and
let developers on the main list know about MPI related failures.
> A couple of questions:
> 1) For my more up-to-date compiler, I may try following gcc svn. Is that
> sensible, and would it upset the regression tester (as obviously fails might
> come from gcc).
I'd prefer to only publicly test released compilers, so that time isn't
wasted tracking down gcc snapshot problems.
> 2) Would our time be better spent on release or trunk? It looks to me like
> release could do with more testing, and we only ever use releases of boost
> in practice.
For release testing, we prefer a specifically targeted set of platforms, and
test runners who have demonstrated their ability to run tests reliably. So
please test for awhile on trunk, and only switch to release if your tests
are cycling smoothly once every 24 hours on an automated basis.
The release platform you could think about contributing would be the
bleeding-edge gcc compiler (currently 4.4.1) running with gnu++0x enabled.
So much the better if it runs MPI enabled.