commons-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Phil Steitz <phil.ste...@gmail.com>
Subject Re: [math] adding a profil for Ekstasi test optimizer
Date Mon, 15 Dec 2014 23:49:44 GMT
On 12/15/14 2:50 PM, Luc Maisonobe wrote:
> Hi all,
>
> A fex weeks ago, Milos Gligoric proposed to add his Ekstazi project to
> improve tests runs during development time (see
> <http://markmail.org/message/7a4ayov3gsvtsm5g>). The rationale is that
> after a first (long) and complete test run, some fingerprints are stored
> and only a subset of test is rerun if only a few classes have changed.
> The project home is <http://www.ekstazi.org/>.
>
> I have tested it and the latest version worked pretty well for me.
> Typically the latest changes from today between commits 59fe593 and
> 809f0f8 did change 31 files, but in fact most changes were javadoc and
> only one class really changed (LevyDistribution with an added
> constructor). This was detected and only the Levy distribution tests
> were run, which is a huge speedup for a project with as many tests as we
> have in [math].
>
> In order to activate this, the only thing required is to add a profile
> in our pom which will pull a maven plugin. Milos did prepare the patch
> for us here: <http://mir.cs.illinois.edu/gliga/tmp/commons-math.patch>.
> Once the patch is applied, if we run tests with the profile activated
> (i.e. with -Pekstasi in the maven command line), then tests are run
> according to the profile. The first time it is done all tests are run
> and a .eksazi folder is created to hold the fingerprints (we should
> probably also add it to .gitignore). If we don't use the profile, all
> tests are run as usual.
>
> Accroding to what I have seen on the project homepage, they intend to
> publish the project under Apache V2 license, but the source code seems
> not yet available (anyway, it is a build tool, not a code dependency).
>
> Do you agree with adding this profile to our pom?

I don't have a problem with it.  I rely heavily on -Dtest=... to
limit execution when I am working on stuff, but then before a push I
run the full suite (actually running it now ;). I can see the value
of what you are proposing and might grow to trust it (or at least
use it instead of -Dtest).  So no objections from me.

Phil
>
> best regards,
> Luc
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
> For additional commands, e-mail: dev-help@commons.apache.org
>
>



---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@commons.apache.org
For additional commands, e-mail: dev-help@commons.apache.org


Mime
View raw message