On Wed, 24 Jun 2015, Sean Dague wrote:
On 06/24/2015 07:57 AM, Chris Dent wrote:
If the primary reason is so that we can rely on the console_scripts
entry point to handle getting the application somewhere useful then
that too feels a bit crufty, in the sense of "that's a hack".
[snip]
The reason I want this is so that the upgrade process for keystone is:
pip install ./keystone
And not have to also have knowledge about the contents of the keystone
source directory. Today a lot of installation and upgrade logic for
packages is "left as an exercise for the reader", which means devstack,
ansible, rpms, debs all end up doing a bunch of work beyond pip install.
Minimizing that by bringing more of this back into what pip install does
will make for a more common base moving forward for everyone.
Makes sense. My only concern would be the somewhat of a mismatch in
semantics between console_scripts and what is not a console script.
Does anyone who is more familiar with entry points know of an
alternative that will accomplish the same sort of thing?
If there's not, then awesome, let's do it.
--
Chris Dent tw:@anticdent freenode:cdent
https://tank.peermore.com/tanks/cdent
__________________________________________________________________________
OpenStack Development Mailing List (not for usage questions)
Unsubscribe: [email protected]?subject:unsubscribe
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
