[asdf-devel] GIT branching [was Re: [Bug 587889] Re: bad fasls can poison future build]
On 6/16/10 Jun 16 -11:13 AM, Faré wrote: ....
Robert: any thoughts re: branching?
I am so far from being a git expert, that I will not venture any suggestions. A while ago I read the following article which proposes a branching policy for use with git. It seemed, to my limited understanding, reasonable. Perhaps we could adopt some variant of this policy? http://www.newartisans.com/2009/10/branch-policies-with-git.html He seems to have many more branches than I would have thought necessary. I don't know that we need a "released," "stable devel," and "bleeding edge" as he does. Perhaps some sort of variant where we have a maint branch --- 2.0 with patches for bug fixes devel branch --- moving towards 2.1 ? I'm inclined to think that we could further minimize/simplify by having the devel branch be master. We could also have topic branches, like the one I built in order to do the first TRAVERSE mods for module dependencies. Does that sound reasonable? Feel free to say "no," since I'm making this up as I go along based on half-remembered blog posts. best, r
I agree we don't want to multiply unnecessary branches or slow down the process with bureaucracy. I propose we have: 1- a "release" branch, where we drop off 2.0. The head of that branch should always be the properly tagged latest release, which will someday be 2.1, etc. We don't (currently) plan to support multiple old version branches, so just one branch is enough. 2- the "master" branch, where development happens. If possible one should not push to this branch without passing proper tests, though shit may happen when some "obviously right" change breaks things on some unintended obscure case. 3- various transient topic branches, when applicable, for experimental development. Note that in my rush to release 2.0, I haven't stuck to any rule such as "any feature should have a test". Maybe we should institute this rule for any future feature or bug fix. [ François-René ÐVB Rideau | Reflection&Cybernethics | http://fare.tunes.org ] There is one rule for the industrialist and that is: Make the best quality of goods possible at the lowest cost possible, paying the highest wages possible. — Henry Ford On 16 June 2010 20:16, Robert Goldman <rpgoldman@sift.info> wrote:
On 6/16/10 Jun 16 -11:13 AM, Faré wrote: ....
Robert: any thoughts re: branching?
I am so far from being a git expert, that I will not venture any suggestions.
A while ago I read the following article which proposes a branching policy for use with git. It seemed, to my limited understanding, reasonable. Perhaps we could adopt some variant of this policy?
http://www.newartisans.com/2009/10/branch-policies-with-git.html
He seems to have many more branches than I would have thought necessary. I don't know that we need a "released," "stable devel," and "bleeding edge" as he does.
Perhaps some sort of variant where we have a
maint branch --- 2.0 with patches for bug fixes
devel branch --- moving towards 2.1
?
I'm inclined to think that we could further minimize/simplify by having the devel branch be master.
We could also have topic branches, like the one I built in order to do the first TRAVERSE mods for module dependencies.
Does that sound reasonable?
Feel free to say "no," since I'm making this up as I go along based on half-remembered blog posts.
best, r
I did create that release branch in the end, but by a manipulation mistake, I ended up pushing the current HEAD instead of just the current release tag. Oh well. There is no code difference, happily, only enhanced documentation and a test cleanup. But from now on, I'll be using separate checkouts so there is less room for blunder. Next time someone makes a code change in HEAD that is not going to be released as part of 2.0, I propose he should bump the version to 2.100. When the 2.100 branch has matured enough, we can release 2.200 or 2.500 or 3.000 or whatever. Hopefully, the project remains simple enough that we don't need multiple dots in version names as in 2.6.27.47, but come what may. In any case, the master branch is open for commits. But please don't break it - test before you push. [ François-René ÐVB Rideau | Reflection&Cybernethics | http://fare.tunes.org ] As the Chinese say, 1001 words is worth more than a picture. — John McCarthy
participants (2)
-
Faré -
Robert Goldman