On 10/01/2008 10:35 AM, Adrian Brock wrote:
On Wed, 2008-10-01 at 16:32 +0100, Kabir Khan wrote:
>> On 10/01/2008 05:05 AM, Adrian Brock wrote:
>>> A branch will be created for the broken build
>>> before rolling back so the commit(s) can be fixed
>>> and tested before re-merging.
>> I vote against this, if my vote counts.
What does voting have to do with it?
You either have a persuasive, reasoned argument or you don't.
"I don't like it" isn't an argument. :-)
Good thing I didn't make that argument then. :-)
> Just roll it back, you
>> don't need a branch.
The purpose of the branch is so that
the person that broke the build can get help when they
don't understand why it broke.
Or when concurrent commits conflicted with each other.
Well I still think that isn't necessary.
Bob: Hey Fred, your commit 28374 broke the build, so I reverted it with 29102.
Fred: Ok. I need help figuring this out, can you look at it for me?
(Bob types "svn up -r28374" and does some testing)
Bob: You forgot to frobnicate the blorgsnaffle.
Fred: Oh, ok. I can fix that.
(Fred types "svn merge -r29102:29101 ." to undo the revert, or at his
option, "svn merge -r28373:28374" to re-do their commit, and subsequently
frobnicates the blorgsnaffle)
Fred: Ok, fixed in 29152.
It's really upto the person that needs help with the commit
whether they branch from the rolled back revision
or create a branch from the most recent revision and
re-apply their patch.
OK, I'd say it's up to the fixer to create the branch if needed.
The important part is that don't just keep recommitting it
to trunk or a stable branch
which potentially stops everybody else from working.
OK, I agree with this point.
> The original committer can use "svn merge -
>> rXX:YY" to re-merge the change in their local tree. We already have
>> an awful lot of branches as is.
You can't get somebody else to look at the problem if it only
exists in your local tree.
Well they can easily do the same thing though. Or roll back to the broken
version. The thing I want to emphasize is that the burden should be on the
breaker, not the one who reverts the patch, to do the extra branching and
testing and so on.