Excerpts from Chris Dent's message of 2015-06-12 03:47:02 -0700:
On Fri, 12 Jun 2015, Joe Gordon wrote:
Glad to see us catch these issues early.
Yes! CI is doing exactly the job it is supposed to be doing here. It
is finding bugs in code. When that happens we should fix the bugs,
not revert. Even if it stalls other stuff.
I'd like to offer an alternative path. Please do remember that people
deploy OpenStack from trunk, and we want them to do that, because if an
organization is focused and prepared, continuous deployment will produce
a vibrant OpenStack experience for users, and will keep those operators
in a position to contribute.
We also have hundreds of developers who need the system to work well.
There are maybe only a handful who will work on this issue. It does not
make any sense to me to stall hundreds of developers while a handful
work through something that may take weeks.
So, consider reverting quickly, landing a test that fails at least more
reliably with the current patch, and then have that handful of users who
are capable of fixing it work in their own branch until the problem is
resolved to a level that it won't stand in the way of progress for CD
operators and the development community at large.