Re: glibc-ports-merge branch available for testing

[Russ Allbery <rra at stanford dot edu>]

Thomas Schwinge <thomas@codesourcery.com> writes:
> On Sun, 01 Jul 2012 10:24:33 -0700, Russ Allbery <rra@stanford.edu> wrote:
>> "Joseph S. Myers" <joseph@codesourcery.com> writes:

>>> I should add: I believe

>>> git push origin glibc-ports-merge:refs/heads/master

>> This does an implicit fast-forward merge of your local glibc-ports-merge
>> branch to master in the remote repository.  It's okay to do it this way,
>> but it's mildly confusing (to me at least).  I would instead do:

>>     git checkout master
>>     git merge glibc-ports-merge

>> locally, ensure that Git says that the merge is a fast-forward merge, and,
>> if it does, push your master to the remote repository the way that you
>> normally would.

> But that's always going to give the exactly same results, in my
> understanding?  What distinction are you out for?

Either approach is basically functionally equivalent other than what it
does to your local repository.  If the explicit push makes more sense to
you, then by all means use it.

The things that make it confusing to me is that it doesn't follow the
normal flow of making a change in your local repository first and then
pushing that change, and it results in master on the remote repository
being ahead of master on your local repository even though it was a change
that you just made.  On the other hand, it's probably slightly safer,
since the push will refuse to do a non-fast-forward merge (with a normal
remote repository configuration).

-- 
Russ Allbery (rra@stanford.edu)             <http://www.eyrie.org/~eagle/>