Open Stack

On 02/07/2013 06:55 AM, Daniel P. Berrange wrote:
> On Thu, Feb 07, 2013 at 07:40:13AM -0500, Sean Dague wrote:
>> It's all fun and games... until someone has to do the work. :)
>>
>> The entry_point drum beat has been around for while, at least the
>> last 2 design summits. But when you look at Nova, there are at least
>> 3 different ways we load things today, so this ends up being in the
>> "Tedious Task" realm to get us there.
>>
>> Monty's tried a couple of times, but the first came late in Folsom,
>> so the change was way to destabilizing. The second was early in
>> Grizzly, had a huge impact on the unit test run time, which never
>> got addressed, and I think got Abandoned.

The run time issue has now been addressed - Doug added a memoization
feature into stevedore. But then I got sucked in to the testr stuff.

>> I'd say right now we're too late in Grizzly for this. But a real
>> plan for Havana is probably a good one.
>>
>>  * Converting our existing oslo importer to entry points
>>  * Converting the LazyPluggable loaded stuff to entry points
>>  * Making nova extensions actually entry points
>>
>> With it will come some substantial deprecation again, but that's
>> probably fine. Something like this should drop by havana-1 or
>> havana-2, by the time we get into havana-3 it's just too late for
>> such changes.
>
> Agreed, that kind of timeframe does look like the best option
> given the potential for trouble the change may create.

Agree. It's actually not hard work to get it slogged through - but it is
a slog-through-it thing.

Also - the console-scripts entry point script hits an error that I've
never been able to sort out...

Monty