On Tue, 18 Jul 2017 at 16:06 Terry Reedy <tjreedy@udel.edu> wrote:
On 7/18/2017 2:31 PM, Brett Cannon wrote:

> Once again, glad the goals are panning out. :) Key thing has always been
> to increase our bandwidth and part of that was always to push more on to
> contributors so they can be more self-servicing.
>     * most contributors create backports (using cherry-pick) themself.
>     Before, this painful/error-prone task was usually done by a single
>     developer without any review, without pre-commit tests, etc.

Backports affecting the same file should be done in the same order, with
the same 'before' state.

   Real example: I merged PR A for file x.py and updated my clone to the
result of the merge.  I then prepared and merged PR B for the same file.
  Someone 'helpfully' prepared a backport of PR B, call it PR B3.6.  It
passed CI but was wrong.  Fortunately, I checked the 'after' state of
the file on a side-by-side diff. I prepared and merged PR A3.6, updated
the 3.6 branck of my clone, and then prepared a new and correct backport
of PR B.

Correct backporting is most easily assured if backports are done
immediately.  I currently do so myself instead of requesting and waiting
for a contributor to do so (who likely is not immediately available).
Even better would be for the backport to happen automatically.

> My wishlist that I don't think anyone is actively working on ATM is:

Just so people know, I'm not going to reply to any comments about the wishlist since worrying about design details for something that doesn't even have someone motivated enough to work on it is I think too premature to worry about. I put this list here basically to see if something suddenly garnered someone's interest to implement or to see if there was an unexpected groundswell of interest to push it up the priority list. If any of these ideas to get someone behind them then the discussion will end up on the core-workflow mailing list.