Re: [PATCH 2/2] push: Use "last one wins" convention for --recurse-submodules

From: Stefan Beller <sbeller@google.com>
Date: Dec 10, 2015, 23:44 UTC
Message-ID: <CAGZ79kaehf+o9qwznTZyG743OrO2EerEO0PAWois4L2G6=933w@mail.gmail.com>
In-Reply-To: <xmqq8u51ho27.fsf@gitster.mtv.corp.google.com>

On Thu, Dec 10, 2015 at 3:38 PM, Junio C Hamano <gitster@pobox.com> wrote:

Show 15 quoted lines

> Stefan Beller <sbeller@google.com> writes:
>
>>> +               git push --recurse-submodules=on-demand --no-recurse-submodules ../pub.git master &&
>>> +               # Check that the submodule commit did not get there
>>
>> Do we want to check here that the supermodule commit did get there,
>> instead of only checking the submodule?
>
> Hmm, your point is that when the push succeeds, (1) the command
> should return with 0 status, (2) the branch in the superproject
> should update to the right commit, and (3) none of the submodule
> should be affected, and the current test does not check the second
> one?
>
> I think that makes sense, in somewhat a paranoid way ;-).

I was just comparing to the case before, where we had

> +               # Check that the supermodule commit did get there
> +               git fetch ../pub.git &&
> +               git diff --quiet FETCH_HEAD master &&

which I just skimmed over and mistakenly thought it would be the same check as before (checking the superproject did *not* get there).

So looking at it, the superprojects history would need no update, the commit stays the same, so no need check for it to stay the same.

So, sorry for the noise.