[Python-Dev] feature freeze questions

[Georg Brandl]

Am 26.07.2012 10:24, schrieb Eli Bendersky:
>>> I have a few questions about feature freeze:
>>>
>>> (1) Is increasing test coverage acceptable during feature freeze, even
>>> if it does not involve fixing a bug?
>>>
>>> (2) When adding new tests (e.g. in the course of fixing a bug or
>>> increasing test coverage), are we allowed to refactor other tests so
>>> that supporting test code can be shared?  Or should the tests be added
>>> in a less DRY fashion and refactored only after the branch goes back
>>> to pre-alpha?
>>
>> You can do basically anything you want to tests just as long as you
>> don't make them less stable.
>>
>>>
>>> (3) What types of documentation changes are allowed during feature
>>> freeze?  For example, are we only allowed to fix incorrect
>>> information, or is it acceptable to improve or add to the information
>>> about existing functionality?
>>
>> All documentation changes are accepted.
> 
> A follow-up question: during the RC stage, do documentation & test
> changes require code review by an additional core dev?

Test changes: yes, doc changes: only if you think it's necessary.

The thing about the docs is: we can fix them basically at any time,
since most people use the online version anyway, which is rebuilt
daily.

> This also leads me to think about minor (= bugfix) releases. For
> example, current commits to the 3.2 branch will (presumably) go into
> 3.2.4 once it's out. Should these commits have the same restrictions
> as commits to an upcoming release that's in RC stage? Common sense
> would suggest they should (since any change can break existing
> functionality and lead to regressions in a previously stable release),

But that can also happen for 3.2 -> 3.3.

> but in practice I don't think this is followed. At least not to the
> letter, with the devguide stating:
> 
>     You cannot skip the peer review during an RC, no matter how small!
> Even if it is a simple copy-and-paste change, everything requires peer
> review from a core developer.

That is only for RC phase, yes.  Reviews are always useful, of course
(calling them "peer review" sounds funny to me though), and should be
always done for larger patches -- which we are quite used to nowadays.

Georg