I really don't mean to create/prolong a debate here. Personally, I find it
easier to stay out of downmerge problems with a branch, when there are
parallel changes afoot, and I like to have local history of my efforts.
However, chacun a son gout.


On Sat, Oct 3, 2009 at 6:28 PM, Grant Ingersoll <[email protected]> wrote:

>
> On Oct 3, 2009, at 12:10 PM, Benson Margulies wrote:
>
>  On Sat, Oct 3, 2009 at 11:19 AM, Grant Ingersoll <[email protected]
>> >wrote:
>>
>>
>>> On Oct 3, 2009, at 10:11 AM, Benson Margulies wrote:
>>>
>>> There is one further level of fun available here.
>>>
>>>>
>>>> Step 1: committer makes svn branch.
>>>>
>>>> Step 2: verify/arrange that branch is included in git clone.
>>>>
>>>> Step 3: set up served git clone of that branch at your place with
>>>> gitosis
>>>> or
>>>> whatever. Note that git svn doesn't do 'bare', but it doesn't turn out
>>>> to
>>>> interfere.
>>>>
>>>> Step 4: students actually commit to that clone.
>>>>
>>>> Step 5: committer pushes that up to ASF svn on the branch with light
>>>> supervision.
>>>>
>>>>
>>>>  I'm not sure the ASF treats branches any differently from trunk when it
>>> comes to the supervision involved.
>>>
>>>
>> Let me clarify. Anything that goes into the real ASF repo has to come from
>> someone with an iCLA, and a committer has at least to be confident that
>> there isn't someone else's IP involved. Beyond that, we're in the
>> territory
>> of the particular's community's taste in patch control. Piling up a bunch
>> of
>> contributions in a branch is, in my view, just another alternative to
>> stacking them up in JIRAs.
>>
>
> Patches can be updated individually and managed individually.  If a branch
> gets to far out of whack, even in Git, it becomes harder and harder to merge
> back in and will require a massive merge unless it is done frequently, in
> which case you might as well use patch.
>
> I don't know why the whole patch thing is viewed as so complex:
>
> Steps:
> Make changes in code
> In the checked out dir: svn diff > ../MAHOUT-XXX.patch
> Upload to JIRA.
>
> Application:
> Download patch
> patch -p 0 -i ./MAHOUT-XXX.patch (do a --dry-run first)
> Review changes.
>
> Besides the fact, that JIRA should still be used extensively b/c that is
> how we track changes for releases.
>

Reply via email to