Open Stack

Excerpts from Clint Byrum's message of 2015-07-07 09:13:39 -0700:
> Excerpts from Victor Stinner's message of 2015-07-07 08:22:26 -0700:
> > Hi,
> > 
> > Oslo libraries become more and more stable, cool :-) Sometimes, we have 
> > to modify APIs, to fix bugs, to deprecate functions, to add new function 
> > parameters, etc. It would be cool to _document_ these changes. Sphinx 
> > provides nice markups to document API changes:
> > 
> >     http://sphinx-doc.org/markup/para.html
> > 
> > We can try to ensure that API changes are documented in reviews, or 
> > maybe also document old changes.
> > 
> > Examples:
> > 
> > .. versionadded:: 2.5
> >     The *spam* parameter.
> > 
> > .. versionadded:: 2.6
> >     The function now returns an integer.
> > 
> > .. deprecated:: 3.1
> >     Use :func:`spam` instead.
> > 
> > I noticed that the deprecated markup is already used in oslo.utils doc:
> > 
> > http://docs.openstack.org/developer/oslo.utils/api/timeutils.html#oslo_utils.timeutils.isotime
> > 
> 
> This is great information, thank you.
> 
> Any time we have versions appearing in code or docs, I cringe, because
> versions and releasing are separate from coding and documenting.
> If I'm adding a parameter today in the latest commit after 1.0.0, I
> really don't know if the next version released will be 1.1.0 or 2.0.0
> (I do know it won't be 1.0.1 because I have changed the API!).
> 
> Or do we just make a rule that you _do_ get to choose the version bump
> in the commit that changes the API, and if you say 1.1.0, and it becomes
> 2.0.0 before release, thats ok because the versionadded is still accurate
> even if 1.1.0 didn't release?

I expect that most of the time incrementing the minor version is
going to be accurate (we've so far only had backwards-incompatible
versions > 1.0 for removing the oslo namespace). However, if we
find that a versionadded note references a version that was never actually
released, we can treat that as a doc bug and go back and update it.

Doug