[Bioperl-l] Smoke test aggregator - Buildbot + Smolder ?

[Jay Hannah]

On May 5, 2010, at 7:12 AM, Chris Fields wrote:
> I think this would be great!  Would you stick with trunk only, or other bioperl dists (run, for example)?

I would definitely start with trunk and see how it goes.

Last night I tried to smoke all our old $work[0] tags and failed impressively. Our tests were (and probably still are) too reliant on 3rd party black boxes being online and responsive, and servers tend to move and get reconfigured over the years. Presumably BioPerl and Moose and more self-contained (unless external deps are explicitly enabled), so perhaps historical smoking would work fairly well.

In Moose land the request is that I smoke not only Moose, but everything on CPAN that *depends on Moose*:

   export MOOSE_TEST_MD=1; prove xt/test-my-dependents.t

Which should be ... educational.  :)   While exciting, I don't think that concept translates to the BioPerl monolith.

If I'm the only one smoking, you'll get a very limited number of architecture + perl version combinations reported. Which begs the question of how to harness a broader tester pool. It's great that 342 systems smoked our latest CPAN upload:

   http://static.cpantesters.org/distro/B/bioperl.html

But the crazy I'm embarking on would mean several smokes each day (every svn/git commit?), compared to the cpantesters who haven't had a new CPAN release to smoke since Sep 2009 (1.6.1). Maybe I'd just do one or two a day or something? 

Whoever wanted to could report into our central Smolder server using their architectures + perl versions. A volunteer would just install Smolder from CPAN and run this in their bioperl-live directory:

prove -I . --recurse --archive test_run.tar.gz
smolder_smoke_signal --server smolder.jays.net \
   --username MyUserName --password MyPass \
   --file test_run.tar.gz --project bioperl-live --tags trunk

Deep ponderings,

Jay Hannah
http://clab.ist.unomaha.edu/CLAB/index.php/User:Jhannah