Installing bug fixes: Trunk or Emacs 22 branch? (was: [david.reitter@gmail.com: reftex visits .rel files])

[Reiner Steib <reinersteib+gmane@imap.cc>]

[ I'd suggest to shift this to emacs-devel. ]

On Mon, Jul 23 2007, David Kastrup wrote on auctex-devel:

> Reiner Steib <reinersteib+gmane@imap.cc> writes:
>> Bug fixes should go into Emacs 22, I think.  (And Richard explicitely
>> requested to install the fix in the Emacs 22 branch.)
   [ Later Richard added: "Someone will probably merge it into the
   trunk, but if you have time to do so yourself, please do.  That is
   more reliable." ]

> But I think they usually percolate from trunk to branch.  The problem
> is that people are not really working with the branch, so putting the
> stuff into the branch first guarantees more or less that it is not
> tested.  And that is not a good idea.

IMHO, installing only in the trunk bears the risk that the fix does
not get installed in Emacs 22 because people tend to forget about
it[1] so that it might not be included in Emacs 22.2.  The problem is
that (to my knowledge) no way to mark a patch as "to be installed in
the stable branch as well after some testing" so that it will not be
forgotten.  Installing in branch and trunk at the same time would be
preferable.  But that makes it put some extra burden on the person who
commits the patch.  For patches that are not urgent, it's okay to wait
for Miles to sync it from the stable branch to the trunk.

Let me explain how we handle it for Gnus...

In Gnus, we install all bug fixes either
  (a) in the trunk _and_ in the stable branch
or
  (b) only in the stable branch.

In case (b), Miles syncs the fix from stable branch to the trunk
(usually within a couple of days).  When done in this way, we make
sure that all bug fixes are installed in stable _and_ trunk (syncing
is done semi-automatically) and are tested as good as possible (modulo
Miles' delay).

Only if a bug fix is considered as risky, we install it only on the
trunk and sync it manually to the stable branch after some testing
(c).  Some developers aren't aware of the stable branch (or forget
about it) and install fixes only in the trunk.  In this case, I
usually ask the developer if there's a reason not to install it in the
stable branch (sometimes the ChangeLog entry doesn't make it clear
whether or not it's a bug fix) or I install it in the stable branch
myself (doc fixes, typos, ...).

Bye, Reiner.

[1] See e.g. the thread "PCL-CVS: `A' opens wrong ChangeLog file"
    http://thread.gmane.org/v9r6n3knik.fsf%40marauder.physik.uni-ulm.de
-- 
       ,,,
      (o o)
---ooO-(_)-Ooo---  |  PGP key available  |  http://rsteib.home.pages.de/