Folks,
In rev. r5770, I've updated the Ant build so that there are two
different ways to run rubysecs from rubinius project:
stable and unstable/latest specs.
Stable ones are in known good state, and all known JRuby failures are excluded.
The idea is that the stable specs runs must be clean, no failures. And any new
failure must mean the regressions in your upcoming fixes. Basically,
once your new
fix/change is ready, it's OK to execute "ant test" and "ant spec" and
verify that there are no failures.
Stable specs are downloaded as a snapshot of particular revision.
The revision for the stable specs is being set in rubyspecs.revision file.
The build is smart enough to detect when the new revision is specified
and to download the updated specs when needed.
NOTE: rubyspecs are not actively tested/maintained on Windows, so
there could be test failures or other problems when runnnig on Windows.
But they are actively ran on Linux/MacOSX.
Unstable specs are the very latest, and might have new failures.
Unstable specs are downloaded as a git repo.
(That's the original approach we've been using for some time now).
The problem with it is that rubyspecs are actively updated on a day-by-day
basis, and it was hard to have an always-changing target.
So, basically, there are two important commands:
ant spec - runs the "stable" specs.
ant spec-latest - runs the latest specs.
Also, the build is modified in such a way that "ant clean" does not remove
the specs (in order not to waste bandwidth downloading them over and over).
Use ant clean-all to remove everything including the specs.
And, I've updated the most important targets descriptions, so that
ant -p produces more useful list of important/public targets the users
might be interesting to run.
Thanks,
--Vladimir
---------------------------------------------------------------------
To unsubscribe from this list please visit:
http://xircles.codehaus.org/manage_email
