[gmx-developers] impending 5.1.4 and 2016.1, and new bugfix policy

[Mark Abraham]

Hi,

We have a few scientific-correctness fixes in release-5-1 that we should
ship. So I will do that on Wednesday.

It's disappointing that Berk's straightforward fix for Parrinello-Rahman
with nstpcouple != 1 was up in gerrit for three weeks (now merged by me),
and only I found time to look at it. Naturally, many of you can't have the
time or expertise to look over such things, but otherwise that means that
people are voting with their feet that they want long-lived bug-fix
branches only if someone else does the work of 2-person review (and then
the review of the merges).

Since this kind of thing is typical of the last few years, for the next
period I suggest we'll adopt a policy (which Roland suggested to me
privately a few weeks back) where all bug fixes go onto release-2016
branch, and get their primary review there. Anybody has discretion to
cherry-pick a scientific-correctness bug fix back to release-5-1 for
possible future release. We need to change something, because there's
clearly not enough available manpower to justify our current workflow of
* find mdrun bug while cleaning up master branch
* fix bug on release-5-1
* wait for 2-person review on release-5-1
* do merge to release-2016
* wait for 2-person review on merge to release-2016
* do merge to master
* wait for 2-person review on merge to master
* resume work on master branch several months later

That will likely mean fewer things actually get fixed in release-5-1, but
that seems to be how people actually want to spend their time.

This will also mean that come next release phase, we will close
release-2016 for the duration of the release phase. Once we release 2017,
probably we'll have bug fixes that make sense to backport to release-2016,
but clearly we can't do more than one thing at a time efficiently.

After 5.1.4, I'll then be travelling for 10 days, but we should plan to do
2016.1 in late-ish September, so please get your fixes and reviews in :-).

Mark
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-developers/attachments/20160902/8eed65e5/attachment.html>