[Gluster-devel] regressions on release-3.7 ?

Vijay Bellur vbellur at redhat.com
Mon Apr 20 17:36:51 UTC 2015


On 04/20/2015 10:39 PM, Jeff Darcy wrote:
>> Are you meaning deleting the tests temporarily (to let other stuff
>> pass without being held up by it), or permanently?
>
> Kind of both.  I think the tests should be deleted so they don't
> interfere with other work, but that means we'll have features without
> tests.  Lack of a test should be considered a release blocker just as
> surely as lack of the feature itself.  Therefore, either fixing the old
> tests or developing new ones should be a very high priority.
>

I propose that we don't drop test units but provide an ack to patches 
that have known regression failures. We should also ensure that all 
regression tests pass as a gating factor for the release of 3.7.0. This 
would help us in achieving both aspects outlined here - not block 
ongoing work & ensure that we add new tests or fix broken ones.

With our recent improvements to run-tests.sh, this would need us to 
maintain a list of known test units that fail and continue running tests 
as long as we hit a failure from the list.

>> That makes sense to me... as long as any temporarily-deleted-tests
>> have their root cause(s) found and fixed before we release 3.7. :)
>
> Exactly.  We can't release until we can test.  Whether we can't test
> because the tests are broken or because the tests are absent, the result
> is the same.  New/fixed tests should be required before we can release.
>

+1.

-Vijay



More information about the Gluster-devel mailing list