Le 21/04/2015 02:26, Adam Heath a écrit :
On 04/20/2015 04:12 AM, Jacques Le Roux wrote:
Like Adrian and mostly for the same reasons, I don't believe we need Git.
But there is one other major reason which has already been discussed in the other common ASF MLs. As Taher exulted, it's possible to create local
branches. So people are able to do a lot of work alone without exchanging before committing or submitting. It will certainly not help to have this
possibility. Remember our recent discussion on the lack or core commits reviews. With Git you end with commits bursts or big patches and it's then
hard to review and too late to share ideas.
I take exception to this; I believe you are referring to my commit bursts, in
times past. Aka, 10-15 commits get added to svn in under a minute.
Git allows me to create a new feature, initially as a single large commit(or several large commits). Then, using the rebase feature, I pull out
very small changes, that are easy to understand, and digest. I then might commit those as completely separate fixes/feature-additions, which then
get reviewed. I then wait a bit, and rebase my big work on top of the new trunk.
Eventually, I get the history to such a point that I feel that the series of commits are an easy to understand progression. At that point, I commit
the entire set to svn.
Note, that I run the entire set of test cases(as they exist at that point)
against each and every single commit, before I send them off.
Having each commit separate, allows for each change to be looked at in isolation. If they were all combined into one, then it would be much more
difficult to review.
If the number of commits at once is the problem, then I don't follow. Spreading each commit out over several hours still won't cause anything to
get reviewed.
I can see your point and it's difficult to not agree. I though have still a feeling that with Git spirit less things will be shared and discussed
before being committed
Jacques
