Java开发网 - 用subversion在做rebase是的疑问？

Java开发网

您没有登录

» Java开发网 » Design Pattern & UML

打印话题 寄给朋友 订阅主题

作者

用subversion在做rebase是的疑问？

dissip

BigCat

发贴: 240
积分: 60

于 2006-03-13 19:58

如图所示：
create branch (@2)from trunk(@1)
rebase from trunk to branch 2 times(from@3 to @5 and from @6 to @8)
finally when i want to merge the change from branch(@9 ) to trunk (@10)
how to do it in subversion.

it seems because subversion treat branch as copy, rebasing will always cause collision when finally try to merge branch to trunk.
Is that true? any better solutions than just merge and solve conflict for all files?

trunk    1-------->3------->6---10
            |cp      |rb      |rb     ^
            v         v         v       |merge
branch    2-->4-->5->7-->8-->9

dissip edited on 2006-03-13 20:04

To live is to fight.

作者

Re:用subversion在做rebase是的疑问？ [Re:dissip]

floater

Java Jedi

总版主

发贴: 3233
积分: 421

于 2006-03-26 11:58

I am not aware of. If the conflicts happen more often, we should think about the code entangling at the design level(facing code and facing requirement sides), need to seperate responsibilities further. In my exp, it usually happens a couple of times a year.

Another entirely different approach is branch-and-die, i.e., never merge back from official/release versions. Usually, this is for highly incompatible versions.

I've worked in different companies with different version control stategies, merge vs die, or layout on top of the head of branch. It works, but not perfect.

I think/believe the ultimate solution is to have a change control system on top of version controls, such as clear quest/sablime/testtrack, etc. So you can have a precise image of the source code base(branch/head are just particular examples of a snapshot of the base) because we have to controls over the changes(via change requests, or CRs in short).

A CR system normally handles conflict better than at the SCCS level because it brings in all related parties together. It doesn't eliminate the conflicts, but with all related parties, it should be handled gracefully. Of course, the price to pay is the extra system.

my two cents

"Any fool can write code that a computer can understand. Good programmers write code that humans can understand."
- Martin Fowler, Refactoring - Improving the Design of Existing Code

作者

Re:用subversion在做rebase是的疑问？ [Re:dissip]

dissip

BigCat

发贴: 240
积分: 60

于 2006-03-27 16:52

3x for your comments. but

in my rebase diagram:
at point 5(after rebase 1), code changes in trunk(from 1->3) are merged.
at point 8(after rebase 2), code changes in trunk(from 3->6) are merged.
at point 9, when i try to merge all changes on branch to trunk, it's hard for me to distinguish which change is made on branch(this is what we need to merge) and which come from trunk by previous rebase(already have on trunk).
So i compare the diff from point 9 to point 2, and merge the diff to trunk. this will cause the changes that's rebased from trunk to be merge back to trunk again. it will cause collision.

To live is to fight.

作者	Re:用subversion在做rebase是的疑问？ [Re:dissip]
floater Java Jedi 总版主发贴: 3233 积分: 421	于 2006-03-28 00:16 My guess is that you should compare the diff between 6 and 9, rather than 2 and 9. "Any fool can write code that a computer can understand. Good programmers write code that humans can understand." - Martin Fowler, Refactoring - Improving the Design of Existing Code

已读帖子

新的帖子

被删除的帖子