On Sat, Mar 07, 2009 at 08:02:07PM -0800, Allison Randal wrote: > Patrick R. Michaud wrote: >> >> I think a check needs to be done much closer to the release date than >> "a couple of weeks". Either that or we need to do some >> level of "code freeze" between the time of checking with project leads >> and the release. Otherwise we continue to have a good potential for >> the situation that gave rise to this issue in the first place. > > A couple of days? The longer time-frame was to give language devs time > to report failures so we could fix them.
Aye, we probably need to do that as well. > Daily or even weekly smokes of a language against parrot trunk would be > ideal, but I'm balancing that against reasonable human effort. Is one > run mid-month, and one just before the release manageable? I think that a run mid-month and again just before release is manageable. And I don't think the criteria ought to be "does the language pass all of its tests" -- it simply needs to be "do the HLL project leads have any critical blockers for the current trunk that is about to be released?" In Rakudo's case, I don't necessarily need Parrot trunk to build Rakudo without any test failures; but I do need to know that the about-to-be-released-trunk doesn't contain a fatal flaw that will prevent Rakudo from building altogether, regardless of any fixes/updates we might be able to apply to Rakudo. Pm _______________________________________________ http://lists.parrot.org/mailman/listinfo/parrot-dev
