From: Junio C Hamano <gitster@pobox.com>
Date: Tue, 21 Jun 2016 18:31:04 GMT
Subject: Re: How to find commits unique to a branch
Message-ID: <xmqq7fdiqtqv.fsf@gitster.mtv.corp.google.com>
In-Reply-To: <87oa6ubeq4.fsf@thinkpad.rath.org>

Nikolaus Rath <Nikolaus@rath.org> writes:

> On Jun 20 2016, Nikolaus Rath <Nikolaus@rath.org> wrote:
>> On Jun 20 2016, Junio C Hamano <gitster@pobox.com> wrote:
>>> Nikolaus Rath <Nikolaus@rath.org> writes:
>>>
>>>> What's the best way to find all commits in a branch A that have not been
>>>> cherry-picked from (or to) another branch B?
>>>>
>>>> I think I could format-patch all commits in every branch into separate
>>>> files, hash the Author and Date of each files, and then compare the two
>>>> lists. But I'm hoping there's a way to instead have git do the
>>>> heavy-lifting?
>>>
>>> "git cherry" perhaps?
>>
>> That seems to work only the "wrong way around". I have a tag
>> fuse_3_0_start, which is the common ancestor to "master" and
>> "fuse_2_9_bugfix". I'd like to find all the commits from fuse_3_0_start
>> to master that have not been cherry-picked into fuse_2_9_bugfix.

Hmm, so the topology roughly would look like:

            A'--B'--D' 2fix
           /
          o---A---B---C---D---E---F master
	3start

And you want to find commits in 3start..master that do not have
equivalent in 3start..2fix

"git cherry --help" starts like this:

    NAME
           git-cherry - Find commits yet to be applied to upstream

    SYNOPSIS
           git cherry [-v] [<upstream> [<head> [<limit>]]]

    DESCRIPTION
           Determine whether there are commits in <head>..<upstream>
           that are equivalent to those in the range <limit>..<head>.

Applying that to our picture, we want to find commits yet to be
applied to 2fix, and do so by comparing the commits between
3start..master and 3start..2fix.

I find that the first sentence of the description is fuzzy
("Determine whether" would imply that you would get "Yes/No" but
what we want is "here are the commits that do not have counterpart
in 2fix"), but we already know <upstream> corresponds to 2fix
(i.e. we are finding ones yet to be applied to there, which can be
inferred from the NAME line), so <head> must be 'master' That means
that <limit> corresponds to 3start, and we will be comparing commits
in two ranges:

    master..2fix (i e. <head>..<upstream>, which is the same thing as 3start..2fix)
    3start..master (i.e. <limit>..<head>)

So perhaps "git cherry -v 2fix master 3start"?