[PREP] Last Rounds for Parrot 0.9.1

[Andy Dougherty]

On Sun, 8 Mar 2009, Reini Urban wrote:

> 2009/3/8 Allison Randal <allison at parrot.org>:
> > Reini Urban wrote:
> >>
> >> Skipped tests are less useful than passing todo tests.
> >
> > No, a release should never have test failures or passing todo tests.
> 
> Sure. But for making a release I have to know in advance which tests
> are passing and which not. To be able to fix the code or adjust the
> tests.
> Without test results on foreign platforms this is impossible.

> In order to get a release with no passing native_pbc TODO's and
> failing tests the platform matrix has to be tested and adjusted
> accordingly.

Yes, but this only makes sense if the tests are testing what will be
shipped in the release.  (See TT #357 "Enable meaningful testing of
t/native_pbc/*.t).  Currently, in the release manager guide, step 2g
instructs the release manager to change all the t/native_pbc/*pbc files
as part of the release process.  Thus the files shipped with the release
are not the ones that have been tested.

> People are only sending smokes on trunk and people are not producing
> native_pbc files without begging them.

In my case, that's because the process for producing them didn't work.

> Andy e.g. only tested releases.

No.  That's simply not true.

-- 
    Andy Dougherty		doughera at lafayette.edu