Rebase?

[Arran Cudbard-Bell]

On 1 Nov 2013, at 17:09, John Dennis <jdennis at redhat.com> wrote:

> On 11/01/2013 12:26 PM, Arran Cudbard-Bell wrote:
>> 
>> On 1 Nov 2013, at 15:30, Phil Mayers <p.mayers at imperial.ac.uk> wrote:
>> 
>>> Did someone rebase master? I just failed a ff-merge on my local copy (no changes) and some commits appear to have changed IDs e.g.
>>> 
>>> -commit 287390887d81e3c4a1379dd11c2b176d45cb5a91
>>> +commit a3bab0eafc07e3f32c311c3f412f39560e8b6183
>>> Author: Arran Cudbard-Bell <a.cudbardb at freeradius.org>
>>> Date:   Thu Oct 17 17:32:07 2013 +0100
>>> 
>>>    Typo
>>> 
>>> ...and about 10 more.
>> 
>> Alan was unhappy about the Mavericks commits I made. They were removed from both branches and redone by him.
>> 
>> If you rewind before the point of the divergence and pull it should be fine. 
>> 
>> git reset --hard HEAD~5
>> git pull
> 
> Rewriting history on a shared git repository is considered a no-no for
> exactly this reason (it breaks the trees of cloned repos).

No. One or more branches of the repository diverge, they do not break.

> A better
> approach would have been to make a new commit that corrected the prior
> commit with a clear commit message indicating what happened (including
> referencing the commit id being "repaired")

There are multiple ways you can fix diverged repos, if you don't have 
commits of your own you can rewind a few commits back and pull, if you
do have commits you can use git rebase -i, dig out commits back to the 
point of divergence and git pull --rebase.

No, rewriting history is not a good thing to do, but neither is it some
kind of apocalyptic repository destroying event.

Arran Cudbard-Bell <a.cudbardb at freeradius.org>
FreeRADIUS Development Team