[Monetdb-developers] Branches, Checkins & Propagation

Stefan Manegold Stefan.Manegold at cwi.nl
Fri Nov 10 00:39:08 CET 2006


Esteemed fellow developers,

having two tightly related branches --- the release branch and the
development trunk --- give us the opportunity to clearly and conveniently
isolate new "progressive" developments on the development trunk from
"conservative" code consolidation and bug fixing on the release branch.

Unfortunately, also this medal has a second side: to make the development
trunk benefit from the code consolidation and bug fixes on the release
branch either (1) *each* of the respective changes has to be checked in
twice (by *each* developer), or (2) the checkins to the branch are *in bulk*
(aka. "automa[tgn]ically") propagated from the release branch to the
development trunk on an (ir?)regular basis or on need/request.

(1) has the advantage that all bug fixes are "instantly" available in the
development trunk, too,
but comes at the cost of double checkins and bears the risk of "forgetting"
one.
(2) keep all developers from extra work and -responsibility, but has the
disadvantage that the availability of bug fixes in the development trunk is
delayed.

Both strategies bear the risk of conflicts in case development has modified
code before a bug fix is applied. That's obviously unavoidable and such
conflicts need to be resolved "by hand".

Provided only one of the two strategies is strictly applied these should be
the only conflicts.

Mixing the two strategies, however, is IMHO no option, as it will/does/did
introduce extra conflicts:
Say developer A applies a bug fix X to both the release branch and the
development trunk. 
Later developer B applies once bug fix Y that (partly) changes the code that
was also modified by bug fix X.
Finally, when developer C tries to perform the bulk propagation of bug fixes
since the last propagation (i.e., X & Y) from branch to the trunk, a
conflict occurs since X had already been applied on the trunk.

Hence, mixing both strategies forces us to spend even more of our precious
time and energy on fixing these *avoidable* conflicts by hand.
I guess, I can assume that you all agree that this is not desirable.


Since we started with the first release branch a couple of years ago, we
have been using strategy (2), summarized as 

 * all bug-fixes (and only these) must go to the new release branches 
   (and only there)
 * all new features must go to the development trunks (and only there)


However, since the above (and similar) situation of mixing the strategies
with all its consequences has just occurred, I'm wondering whether you're
all still "satisfied" with strategy (2), or whether you'd prefer to switch
(completely!) to strategy (1).

I'm in favor of sticking to (2), but I'm perfectly fine if we agree on
*consequently* switching to (1).

Hence, please hit your keyboard, comment, and vote for either strategy.

Looking forward to hearing your verdict!

Good night,

Stefan

-- 
| Dr. Stefan Manegold | mailto:Stefan.Manegold at cwi.nl |
| CWI,  P.O.Box 94079 | http://www.cwi.nl/~manegold/  |
| 1090 GB Amsterdam   | Tel.: +31 (20) 592-4212       |
| The Netherlands     | Fax : +31 (20) 592-4312       |




More information about the developers-list mailing list