Open Stack

On 06/11/2013 05:54 AM, Mark McLoughlin wrote:
> On Tue, 2013-06-11 at 08:42 -0400, David Shrewsbury wrote:
>> On Jun 11, 2013, at 6:34 AM, Mark McLoughlin <markmc at redhat.com> wrote:
>>> At this point, I'd suggest we delete the newer versions from pypi and
>>> roll back to 0.5.11 ... if that's possible.
>>
>> That's not really ideal since earlier versions have a bug (that 0.5.12 was
>> attempting to fix) where required python packages were not being installed.
>> You wouldn't notice it unless you were trying to install from a fresh Python
>> environment.
> 
> Yes, we'd have to do a follow up release with a non-regression-causing
> fix for that.
> 
> https://review.openstack.org/32124 is approved, so this is all academic
> at this point. I hope.
> 
> Point is that if we find we've introduced a serious regression, we
> should revert the regression-causing change and approach the issue again
> without the mistakes-causing pressure of "OMG, everything's broken, fix
> it quick".

I agree in general. In this case, the regression didn't surface in a way
we understood for a while, so reverting would have been equally non-trivial.

We have identified new test scenarios that we missed before, so we'll be
implementing those before we land any further pbr changes.

Thanks to everyone for continuing to work through this one. It was not fun.

Monty